PT2591600E

PT2591600E - Adapting the set of possible frequency transforms based on block size and intra mode

Info

Publication number: PT2591600E
Application number: PT117337634T
Authority: PT
Original assignee: Qualcomm Inc
Priority date: 2010-07-09
Filing date: 2011-07-08
Publication date: 2015-01-14
Also published as: ES2526053T3

Description

DESCRIPTION "ADAPT THE ASSEMBLY OF POSSIBLE FREQUENCY TRANSFORMATIONS BASED ON BLOCK SIZE AND INTRA MODE"

TECHNICAL FIELD

This disclosure concerns video encoding.

BACKGROUND

Digital video capabilities can be incorporated into a wide variety of devices, including digital televisions, digital direct broadcast systems, wireless transmission systems, personal digital assistant (PDAs), laptop or desktop computers, digital cameras, digital recording devices, digital video players, video game devices, video game consoles, cellular or satellite radio phones, video teleconferencing devices, and the like. Digital video devices implement video compression techniques, such as those described in standards defined by MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.264 / MPEG-4, Part Advanced Video Coding (AVC), the next High Efficiency Video Coding (HEVC) standard (also referred to as H.265), and extensions of such standards, to transmit and receive digital video information more efficiently. Video compression techniques perform spatial prediction and / or temporal prediction to reduce or eliminate redundancy inherent in video sequences. For block-based video encoding, a frame or video segment can be broken down into macro blocks. Each macro block can be further apportioned. Macro blocks in a frame or intra-coded segment (I) are encoded using spatial prediction in relation to neighboring macro blocks. Macro blocks in a frame or inter-coded segment can use spatial prediction in relation to neighboring macro blocks. Macro blocks in a frame or inter-coded segment (P or B) may use spatial prediction relative to other macro blocks in the same frame or segment or temporal prediction relative to the other reference frames. Reference is made to the following prior art documents: DAVIES (BBC) T ET AL: "Suggestion for a Test Model", l. JCT- VC MEETING; 4/15/2010 - 4/23/2010; DRESDEN; (JOINTCOLLABORATIVE TEAM ON VIDEO CODING OF ISO / IEC JTC1 / SC29 / WG11 AND ITU-TSG.16); URL: HTTP://WFTP3.ITU.INT/AV-ARCH/JCTVC-SITE/, May 7, 2010 (2010-05-07) discloses a suggestion for a Test Model for the next HEVC where a different number of transformations is provided depending on the size of a block. XIN ZHAO ET AL: "Rate-distortion optimized transform for intra-frame coding", ACOUSTICS SPEECH AND SIGNAL PROCESSING (ICASSP), 2010 IEEE INTERNATIONAL CONFERENCE ON, IEEE, PISCATAWAY, NJ, USA, March 14, 2010 -14), pages 1414-1417, ISBN: 978-1-4244-4295-9, discloses associating multiple transformations for each intra-prediction mode by selecting one of them using a criterion of the distortion rate and indicating the transformation selected for the decoder .

SUMMARY The present invention relates to a method of decoding video data in accordance with the appended claim 1, an apparatus for decoding video data in accordance with the attached claim 7, a computer program product according to claim 9, a method of encoding video data according to the appended claim 10, an apparatus for encoding video data according to the appended claim 13 and a computer program product according to the appended claim 15 .

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram illustrating an example of a video encoding and decoding system that can utilize techniques for encoding and decoding transform units of an encoding unit. FIG. 2 is a block diagram illustrating an example of a video encoder that can implement one or all of the techniques for encoding video data transformation units described in this disclosure. FIG. 3 is a block diagram illustrating an example of a video decoder, which decodes an encoded video sequence. FIG. 4 is a conceptual diagram illustrating a graph describing a set of examples of forecasting directions associated with various intra-forecasting modes. FIG. 5 is a conceptual diagram illustrating various intra-prediction modes specified by the ITU-T H.264 / AVC standard. FIG. 6 is a conceptual diagram illustrating a zig-zag search for a block of transformation coefficients. FIGS. 7A and 7B are conceptual diagrams illustrating an example of quadtree and a corresponding larger coding unit (LCU). FIG. 8 is a flowchart illustrating an example of a method for selecting a transformation and a search for applying to a block based on a selected intra-forecast mode for the block. FIG. 9 is a flowchart illustrating another example method for selecting a transformation and search to apply to a block based on an intra-forecast mode selected for the block. FIG. 10 is a flowchart illustrating a sample method for applying an intra-prediction and transformation mode to special-size sub-CUs. FIG. 11 is a flowchart illustrating a sample method for performing an adaptive search of transformation coefficients based on a selected transformation applied to a block of residual values. FIG. 12 is a flowchart illustrating a sample method for selecting a context model to use when searching and entropy encoded syntax elements describing adaptively searched coefficients. FIG. 13 is a flowchart illustrating a sample method for decoding a transform unit that has been encoded using one or more of the techniques of this disclosure. FIG. 14 is a flow chart illustrating a sample method for selecting a transformation to apply to an intra-coded block including a threshold for which the intra prediction mode DC is signaled.

DETAILED DESCRIPTION

In general, this disclosure describes techniques for encoding video data. More specifically, this disclosure describes techniques that relate to transforming residual data and searching for transformation coefficients during a video encoding process. Encoded video data may include prediction data and residual data. A video encoder can produce the forecast data during an intra-forecast mode or an inter-forecast mode. In general prediction involves predicting a block of an image relative to the neighborhood, previously encoded blocks of the same image. Inter prediction generally implies predicting a block of an image relative to the data of a previously encoded image.

Following intra- or inter-prediction, a video encoder can calculate a residual value for the block. The residual value in general corresponds to the difference between the predicted data for the block and the true value of the block. To further compress the residual value of a block, the residual value can be transformed into a set of transform coefficients that compact data (also referred to as "energy") as possible in as few coefficients as possible. The transformation coefficients correspond to a two-dimensional array of coefficients that is the same size as the original block. In other words, there are only as many transformation coefficients as pixels in the original block. However, because of the transform, many of the transformation coefficients can have values equal to zero.

In some cases a secondary transformation, such as a rotation transformation, may be applied to a subset of the transformation coefficients generated by the first transformation. For example, after transforming a 16x16 residual block into a 16x16 transformation coefficient matrix, a rotation transformation can be applied to the lower frequency transformation coefficient block 8x8. While this example describes a rotation transformation as a secondary transformation, other secondary transformations (eg, KLTs, DCTs, and the like) can also be applied as secondary transformations. Such secondary transformations can also be selected based on an intra-predicted mode signaled to the block.

References to "DCT transformations" should be understood to include both fixed-point implementations and floating-point implementations. That is, an implementation of a DCT transformation can actually comprise an approximation of a DCT, such that DCT transformation has integer coefficients (i.e., fixed point coefficients) instead of rational number coefficients.

In some examples, a transformation may comprise a non-separable transformation. Non-separable transformations are typically computationally expensive, and therefore, video encoding devices may instead apply separate transformations. In general, separate transformations include a horizontal component applied to the block lines and a vertical component applied to the block columns. In this manner, a separate transformation may have one transformation component line and one transformation component column, also referred to as two orthogonal transformation components. Two matrices can be used to define a separate transformation, each of the matrices corresponds to one of the orthogonal transformation components. A non-separable transformation may include only a matrix which, when applied, produces a similar result conceptually for applying the transformation separately, but by relatively more intensive calculations.

Transforming a block of residual data produces a set of transformation coefficients for the block. The video encoder can then quantify the transform coefficients to further compress the video data. Quantification in general implies mapping values in a relatively large set to values in a relatively small set, thereby reducing the amount of data required to represent the quantized transformation coefficients. Following quantification, the video encoder can search the transformation coefficients, producing a one-dimensional array vector of the two dimensions including the quantized transformation coefficients. The video encoder can zero certain coefficients before or after the search, eg, all except the upper left corner of the array or all coefficients in the series from the N position to the end of the series. The video encoder can then encode the resulting series entropy, even to further compress the data. In some examples, the video encoder may be configured to use variable length codes (VLCs) to represent various possible quantized transform coefficients of the series, eg, using adaptive context variable length (CAVLC) encoding. In other examples, the video encoder may be configured to use binary arithmetic coding to encode the resulting quantized coefficients, eg, using adaptive context binary arithmetic encoding (CABAC).

This disclosure describes various techniques related to transforming, quantifying, searching, and encoding entropy of residual values during a video encoding process. The techniques may be applied by both video encoding and decoding units, including video encoders / decoders (CODECs) and processing units configured to perform video encoding and / or decoding. References to "video encoding units" or "video encoding devices" should be understood to refer to units or devices capable of encoding, decoding, or both encoding and decoding video data.

Efforts are currently under development to develop a new video encoding standard, now called High Efficiency Video Encoding (HEVC). The next standard is also referred to as H.265. Standardization efforts are based on a model of a video coding device termed as the HEVC Test Model (HM). The HM assumes various capabilities of video encoding devices through devices according to, eg, ITU-T H.264 / AVC. For example, while H.264 provides nine intra-prediction coding modes, HM provides up to thirty-four intra-prediction coding modes. HM refers to a video data block as a coding unit (CU), which may include one or more predictive units (PUs) and / or one or more transformation units (TUs). Syntax data in a bit stream may define a larger encoding unit (LCU), which is a larger encoding unit in terms of pixel numbers. In general, a CU has a similar purpose for a H.264 macro block, except that a CU does not have a size distinction. Thus, a CU can be divided into sub-CUs. In general, references in this disclosure for a CU may refer to a larger coding unit of an image or a sub-CU of an LCU. An LCU can be divided into sub-CUs, and each sub-CU can be further divided into sub-CUs. Syntax data for a bit stream can define a maximum number of times that an LCU can be divided, termed CU depth. Therefore, a bit stream can also define a smaller encoding unit (SCU). This disclosure also uses the term "block" to refer to either a CU, PU, or TU.

An LCU can be associated with a quadtree data structure. In general, a quadtree data structure includes a node by CU, where a root corresponds to the LCU. If a CU is divided into four sub-CUs, the node corresponding to CU includes four sheets, each corresponding to one of the sub-CUs. Each node of the quadtree data structure can provide syntax data for the corresponding CU. For example, a node in the quadtree may include a division flag, indicating whether the CU corresponding to the node is divided into sub-CUs. Syntax elements for a CU can be defined recursively, and may depend on whether the CU is divided into sub-CUs. If a CU is not further divided, it is termed as a CU-sheet. In this disclosure, 4 sub-CUs of a CU-sheet will also be referred to as CU-sheets although there is no explicit division of the original CU-sheet. For example if a CU of 16x16 size is no longer divided, the four 8x8 sub-CUs will also be termed CU-sheets even though the CU 16x16 has never been split.

In addition, TUs of CU-sheets can also be associated with their quadtree data structures. That is, a CU sheet can include a quadtree indicating how the CU sheet is apportioned into TUs. This disclosure refers to the quadtree indicating how an LCU is apportioned as a quadtree CU and the quadtree indicating how a CU-sheet is apportioned in TUs as a TU quadtree. The root of a TU quadtree generally corresponds to a CU-sheet, while the root of a quadtree CU generally corresponds to an LCU. TUs from TU quadtree that are not divided are called TU-sheets.

A CU sheet may include one or more predictive units (PUs). In general, a PU represents all or a portion of the corresponding CU, and may include data for retrieving a reference sample for the PU. For example, when the PU is inter-coded, the PU may include data defining a motion vector for the PU. The data defining the motion vector can describe, for example, a horizontal component of the motion vector, a vertical component of the motion vector, a resolution for the motion vector (eg, a quarter pixel accuracy or an eighth of the pixel), a reference frame to where the motion vector points, and / or a reference list (eg, list 0 or list 1) for the motion vector. Data for the CU sheet defining the PU (s) may also describe, for example, division of the CU into one or more PUs. Split modes may differ depending on whether the CU is not encoded, encoded intra-predictive mode, or encoded inter-prediction mode. For intra coding, a PU can be treated in the same manner as a sheet transformation unit described below.

A CU sheet may include one or more transformation units (TUs). Transformation units can be specified using a quadtree TU structure, as discussed above. That is, a division flag can indicate whether a CU sheet is divided into four transformation units. Then, each transformation unit can be further divided into 4 sub TUs. When a TU is no longer divided, it can be termed as a TU-sheet. In general, for intra-coding, all TU sheets belonging to a CU sheet share the same intra-forecasting mode. That is, the same intra-prediction mode is generally applied to calculate predicted values for all TUs of a CU-sheet. For intra coding, a video encoder can calculate a residual value for each TU-sheet using the intra-prediction mode as a difference between the portion of the prediction values corresponding to TU and the original block. The residual value can be transformed, quantified, and researched. For inter coding, a video encoder can perform prediction at the PU level and can calculate a residual for each PU. Residual values corresponding to a CU-sheet can be transformed, quantified, and searched. For inter-coding, a TU-sheet may be larger or smaller than a PU. For intra coding, a PU can be placed with a corresponding TU sheet. In some examples, the maximum size of a TU-sheet may be the corresponding CU-sheet size.

Generally, this disclosure uses the terms CU and TU to refer to the CU-sheet and TU-sheet, respectively, unless otherwise indicated. In general, the techniques of this disclosure are concerned with transforming, quantifying, searching, and encoding data from a CU. As for example, the techniques of this disclosure include selecting a transformation to be used to transform a residual value from an intra predicted block based on an intra prediction mode used to predict the block. This disclosure also uses the term "directional transformation," or "projected transformation," to refer to such a transformation that depends on the direction of intra-forecasting mode. That is, a video encoder can select a directional transformation to apply to a transform unit (TU). As previously noted, intra prediction includes predicting a TU of a current CU of an image of previously encoded CUs and TUs of the same image. More specifically, a video encoder can intra-predict a current TU of an image using a specific intra-prediction mode.

Techniques of this disclosure include associating certain transformations with intra-predictive modes. Thus, there may be a one-to-one correspondence between intra-prediction modes and transformations according to the techniques of this disclosure. In some examples, there may be a many-to-one match between intra-predictive modes and transformations thereof. For example, a large set of intra-forecast modes can be mapped to a smaller set of intra-forecast modes, and each of the smaller sets of intra-forecast modes can be mapped one-to-one for their respective transformations.

Transformations can also be mapped to their search models. In some examples, intra-forecasting modes can be mapped to both transformations and polls, while in other examples, intra-forecasting modes can be mapped into transformations, and transformations can be mapped into polls. In many instances, various combinations of transformations and coefficient searches can be used. For example, intra-predictive modes can be mapped to mode-dependent directional transformations, and a zig-zag search can be used in all cases.

In some examples, instead of mapping intra prediction modes in transformation and / or search models, video encoder 20 can be configured to signal a combination of one or more transforms and a search model to apply. Likewise, video decoder 30 may be configured to determine a transformation and search model to apply based on a received indication, rather than a mapping between an intra-prediction mode and the transformation and search models.

Transformations may include a discrete cosine transform (DCT) and eight directional transformations, also called Karhunen-Loève Transforms (KLTs). The DCT is generally a sum of cosine functions having different frequencies, where the functions are applied to the residual values. Each KLT generally includes two matrices. Each matrix in the KLT has the same size as the residual block to be transformed. The KLTs can be derived from the training set data or derived analytically by assuming a model for the video frames and / or residual predictions.

An HM encoder can be configured with thirty-four intra-prediction modes for certain block sizes. Therefore, to support a one-to-one mapping between directional intra-prediction modes and directional transformations, HM encoders and decoders would need to store up to 68 arrays for each supported transformation size. In addition, the block sizes for which all thirty-four intra-prediction modes that are supported may be relatively large blocks, eg, 16x16 pixels, 32x32 pixels, or even larger.

In some examples, this disclosure provides techniques for reducing the number of directional transformations that encoders and decoders need to support. That is, encoders and decoders can withstand less directional transformations than the number of available intra-prediction modes. An encoder according to one of these techniques, for example, can map a relatively large set of intra-prediction modes to a subset of the intra-prediction modes. Each of the intra prediction modes in the subset can be associated with a directional transformation. That is, intra-forecasting modes in the subset can have a one-to-one correspondence with a set of directional transformations. Therefore, intra-forecasting modes in the large set may have many-to-one matching with the set of directional transformations.

For example, each of the 34 HM directional intra-prediction modes can be mapped to one of eight H.264 directional intra-prediction modes. The video encoder can therefore select a directional predictive mode to intra-predict a value for a current TU, determine an intra-prediction mode of the subset for the selected mode to be mapped, then use the mapped directional transformation to the intra-prediction mode of the subset to transform the current TU. Additionally, each of the directional transformations can be associated with a respective search model. In this way, the encoder can perform the search associated with the directional transformation to produce a vector of transformation coefficients that can then be quantified. In addition, the encoder can be configured with a maximum size for the vector. That is, the encoder can stop searching the transformation coefficients when reaching the maximum size, if it stops or not, the next coefficient to be searched is different from zero.

In applying the techniques described above, the encoder does not need to signal the transformation used for a specific TU when the techniques described above are used. That is, the encoder and decoder can each be configured with the many-to-one mapping of the intra-prediction modes of the large set to sub-set predictive modes, and the one-to-one mapping of the intra- prediction of the subset for directional transformations. Thus, by signaling the intra-prediction mode of a large set, the decoder can obtain the transformation used to transform the block. In addition, these techniques can be implemented by older devices that have limited memory that can be allocated to store the arrays for the various directional transformations.

An HM encoder may be configured such that the available set of intra-prediction modes for a block differ based on the block size. That is, the size of a CU can determine the number of intra-prediction modes available for the CU, from which the coder can select an intra-prediction mode to predict values used to calculate TU coefficients. Table 1 below illustrates an example

of a match between CU sizes and the number of intra-forecast modes available for CUs of that size. In this disclosure, 4 sub-CUs of a CU-sheet are also referred to as CU-sheets, although there is no explicit division of the original CU-sheet. If the CU-sheet has the smaller CU size, these 4 sub-CUs can select different intra-forecasting modes. Thus, the table has an entry for CU 4x4 size. TABLE 1

In general, a video encoder may signal a prediction direction for a block, so that a video decoder correctly decodes the block. In some examples, a video encoder may be configured to determine a single prediction direction for a CU that can be applied to all TUs belonging to CU. However, as noted earlier in Table 1, certain block sizes have fewer available intra-prediction modes compared to other block sizes. In these cases, it can be solved by allowing the number of prediction directions in the size of the CU block to be used in the TU block sizes. Alternatively, the intra-predicate modes of a larger set can be mapped to intra-prediction modes of a smaller set, eg, a subset. As discussed earlier, there may be a many-to-one relationship between intra-forecast modes of the larger set and intra-forecast modes of a smaller set.

Quadtree TU structures may lead to the decomposition of a larger block (CU) into a smaller block (TUs). The spatial prediction mode of the root block (for CU) can be explicitly signaled in the bit stream. The result of smaller quadtree blocks of TU (TUs) may inherit their prediction modes from that TU quadtree root block (which corresponds to CU). However, the number of spatial prediction directions supported by smaller blocks (TUs) may be different from this root block (CU). This can be solved by allowing more prediction directions for smaller blocks (TUs). Alternatively, smaller block prediction (TUs) modes may be derived from these root blocks (CU) by a many-to-one or one-to-one mapping according to a predetermined criterion such as minimizing the difference of the angle of the prediction direction between the direction of the intra prediction for the CU and the prediction directions supported in the smaller block. Directional transformations and search models can be selected based on this mapping.

In this way, the video encoder can signal an intra-forecasting direction once for a CU. Assuming that the CU includes a TU of a size that does not support the direction of intra predicted signaling, the video encoder can determine the intra prediction mode for TU based on the mapping. That is, the video encoder may intra-predict a predicted block used to calculate a TU using the intra-prediction mode of the smaller set so that the signaled intra-prediction mode of the larger set is mapped. Likewise, a video decoder may include the same configuration, such that video decoder can determine intra-prediction modes for each TU of an received CU. Alternatively, the number of prediction modes for a TU can be increased to match the number of prediction modes for the corresponding CU.

In some examples, for some intra-forecasting modes, multiple transformations may be possible for TUs of specific sizes. In such cases, a video decoder may not be able to obtain the transformation to apply to TU solely from the intra predictive mode. In this way, the video encoder may need to signal the transformation to be used for the TUs of sizes so that multiple transformations are possible. Instead of signaling a transformation for each such TU, this information can be flagged at the CU level. In such cases, this transformation may apply to all TUs contained in CU. For TUs of sizes so that only one transformation is mapped to the signaled intra-forecast mode, the mapped transformation can be used.

In addition, the syntax specifying the transformation needs only to be present if the CU includes a TU of a size so that multiple transformations are possible. For TUs for which only one transformation is possible, the video encoder and decoder can determine the transformation to use based on the selected intra-forecasting mode. However, for TUs of a size for multiple transformations to be possible, the video encoder can explicitly signal the transformation to be used on all TUs of similar size in CU, eg, by flagging the transformation to be used in TU's quadtree root for ASS.

In this way, if a video decoder encounters a TU of a size so that multiple transformations are possible based on the intraprediction mode for the corresponding CU for the TU, the decoder can determine the transformation to be applied based on the explicit signaling. For other TUs, the video decoder can use the transformation associated with the predicted intra-prediction mode for CU.

In some examples, a video encoder may apply more than one transformation (eg, more than a separate transform) to a residual value for a CU. For example, the video encoder can transform a CU TU once by using a first transformation, producing a first set of transformation coefficients, and then applying a second transformation to the first set of transform coefficients, producing a second set of transformation coefficients. This process of applying two or more transformations to a TU can be termed as a cascade transformation. In some examples, the second transformation can be applied only to a subset of coefficients produced by the first transformation. It will be understood that the second transformation may comprise a second, separate transformation, while the first transformation may comprise a first separate transformation. Thus, cascade transformation can be applied by applying four arrays in total to the coefficients: two for the first separate transformation and two for the second separate transformation.

In some examples, the second transformation (i.e., the second transformation separately) may correspond to a rotation transformation (ROT). A rotation transformation can generally be considered to change the coordination system of the transformation base. For example, a video encoder can first apply a directional transformation, and then a spin transformation, to a TU. Another example, the video encoder can first apply a DCT to a TU, then apply a spin transformation to the TU. The video encoder can be configured with multiple rotation transformations. The video encoder may further be configured to apply a rotation transformation following certain directional transformations and / or in conjunction with certain intra-predictive modes. That is, the video encoder can be configured to apply a rotation transformation to certain combinations of directional transformations and certain intra-prediction modes. The different rotation transformations can be indexed by a certain value, eg, the angle of rotation. In some examples, not all coefficients are transformed using a spin transformation. For example, a video encoder may be configured to only rotationally transform low frequency transformation coefficients of a TU.

In addition, in some examples, the techniques of this disclosure include intra-mode predicting a TU having the boundary detected within the TU. For example, a video encoding unit may detect the presence of a boundary on a neighboring block, and then determine that the boundary continues in the current TU. Limit delivery forecast modes can be provided to intra predict such TU. The video encoder can determine whether to predict the TU using the boundary-based predictive mode or other directional intra-prediction mode. When a threshold is determined to exist in the TU, and when the predictive mode based on the threshold is selected, a value indicative of the DC prediction mode can be used to signal the intra prediction mode used but due to the determination of the existence of the this value can be interpreted to indicate the predictive mode of end manipulation. In addition, the angle of the boundary can be determined and mapped to a directional transformation in a similar way to mapping the directional intra-prediction modes for directional transformations discussed earlier. Similarly, a research model mapped to directional transformation can also be used in this example.

This disclosure also provides techniques for researching transformation coefficients so as to produce a uni dimensional vector which can then be coded for entropy. According to these techniques, a video encoder can be configured to select a fixed search model based on a number of factors or to perform an adaptive search. For example, a video encoder may include a set of fixed search models. The video encoder can select one of the fixed search models based on the various criteria, such as for example an intra prediction mode, a transformation selected for a TU, either TU is transformed using a cascade transformation, a transformation of rotation selected for TU, or any combination thereof. For example, the video encoder can select one of the sets of predefined searches based on an intra-forecast mode, a secondary transformation, or a combination thereof. In some examples, the video encoder can select a search index based on one or more of the previously discussed factors, where the search index may match either a fixed or an adaptive search.

In some examples, a video encoder can be configured to adaptively search for transformation coefficients. The video encoder can store an initial and fixed search model. Because the video encoder encodes blocks of an image, the video encoder can update the search model adaptively. The video encoder can, for example, collect indicative statistics if coefficients at locations tend to have values at zero, and if a coefficient at a specific location generally has values at zero, the video encoder can elect to search those coefficients later than other coefficients that generally do not have zero values. In addition, the video encoder can store separate fixed searches and / or search statistics for various combinations of factors, such as, for example, an intra prediction mode, a transformation selected for a TU, if the TU is transformed using a cascade transformation, a selected spin transformation to TU, or any combination thereof. In some examples, a video encoder can store separate statistics for each combination of the cascade transformation, eg, a first transformation followed by a rotation transformation. In some examples, the video encoder may use an adaptive search when the video encoder applies a cascading transformation, and a fixed search when the video encoder applies a single transform.

As noted earlier, searching a two-dimensional array of transformation coefficients produces a one-dimensional vector that can then be coded for entropy. In some examples, a video encoder may encode the transform coefficients using adaptive context binary arithmetic encoding (CABAC). The video encoder can also encode entropy syntax elements such as, for example, a flag of the significant coefficient and a flag of the last coefficient. When adaptively researching transformation coefficients, a video encoder can set the flag value of the significant coefficient to indicate whether the coefficient is significant or not. The video encoder may, for example, be configured to determine that a coefficient is significant when the coefficient value is nonzero. The video encoder can also set the flag value of the last coefficient to indicate the last coefficient in the vector produced by the adaptive search. A video decoder can use these syntax elements to update locally stored statistics in order to reverse adaptive search of the coded entropy coefficients. This disclosure provides techniques for selecting a context model when performing CABAC to encode such syntax elements. The video encoder can select the context-based model, for example, an intra-prediction mode for the CU to be encoded, among other elements. FIG. 1 is a block diagram illustrating an example of video coding and decoding system 10 which may utilize techniques for encoding and decoding processing units of an encoding unit. As shown in FIG. 1, system 10 includes a source device 12 which transmits encoded videos to a destination device 14 via a communication channel 16. Source device 12 and destination device 14 may comprise any of a wide variety of devices. In some cases, source device 12 and destination device 14 may comprise wireless communication devices, such as wireless mobile phones, cellular radio or satellite calls, or any wireless devices that can communicate video information through a communication channel 16, in the case where the communication channel 16 is wireless.

The techniques of this disclosure, however, with respect to encoding and decoding processing units, are not necessarily limited to wireless applications or definitions. For example, these techniques may apply to air-television communications, cable television transmissions, satellite television transmissions, Internet video transmissions, digital coded video that is encoded in a memory device, or other scenarios. Accordingly, communication channel 16 may comprise any combination of the wired or wireless means suitable for transmitting or storing encoded video data.

In the example of FIG. 1, source device 12 includes a video source 18, video encoder 20, a modulator / demodulator (modem) 22 and a transmitter 24. Destination device 14 includes a receiver 26, a modem 28, a video decoder 30 , and a display device 32. According to this disclosure, video encoder 20 of the source device 12 may be configured to apply the techniques for coding and decoding the transforming units of this disclosure. In other examples, a source device and a destination device may include other components or arrangements. For example, source device 12 may receive video data from an external video source 18, such as an external camera. In the same way, target device 14 may interface with an external display device, rather than include an integrated display device. The system 10 shown in FIG. 1 is merely an example. Techniques for coding and decoding processing units may be performed by any digital video encoding and / or decoding device. While the techniques of this disclosure are generally performed by a video encoding device or a video decoding device, the techniques may also be performed by a video encoder / decoder, typically referred to as a "CODEC." Source device 12 and destination device 14 are merely examples of such coding devices in which the source device 12 generates encoded video data for transmitting to destination device 14. In some examples, devices 12, 14 may operate in a substantially such that each device 12 and 14 includes video encoding and decoding components. Thus, system 10 may support one or two way video transmission between video devices 12 and 14, eg, for video streaming, video playback, video streaming, or video telephony.

Video source 18 of the source device 12 may include a video capture device, such as a video camera, a video file containing previously captured video, and / or a video feed from a video content provider. With an additional alternative, video source 18 may generate computer graphics based on the data as a source video, or a combination of live video, archived video, and computer generated video. In some cases, if video source 18 is a video camera, source device 12 and destination device 14 may form known camera phones or video intercom. As mentioned above, however, the techniques described in this disclosure may be applicable for video coding in general, and may be applied for wireless and / or wired applications. In each case, the captured, pre-captured, or computer generated video may be encoded by the video encoder 20. The encoded video information may then be modulated by the modem 22 in accordance with a communication standard, and transmitted to the device by the transmitter 24. Modem 22 may include a plurality of mixers, filters, amplifiers, or other components designed for signal modulation. Transmitter 24 may include circuits designed to transmit data, including amplifiers, filters, and one or more antennas.

Receiver 26 of destination device 14 receives information from channel 16, and modem 28 demodulates information. Again, the video encoding process may implement one or more of the techniques described herein for encoding and decoding transformation units. The information communicated from the channel 16 may include syntax information defined by the video encoder 20, which is also used by the video decoder 30, which includes syntax elements describing characteristics and / or processing of the encoding units or other data units of encoded video, eg, group of images (GOPs), parts, frames and the like. A quadtree CU data structure may form part of the syntax information for a larger coding unit. That is, each LCU can include syntax information in the form of a quadtree CU, which can describe how the LCU is divided into sub-Cus, as well as signaling information such as the LCU and sub-CUs are encoded. Likewise, quadtree TU data structures may be part of the LCU-CU syntax information, which can describe how their CU sheets are divided into TUs.

Video decoder 30 can use quadtree CU and TU quadtrees to determine how to decode CUs from an incoming image, including TUs of CUs. Video decoder 30 may then decode the CUs and send decoded video data to display device 32. Display device 32 shows the video data decoded to a user, and may comprise any of a variety of display devices such as a tube cathode ray tube (CRT), liquid crystal display (LCU), plasma monitor, organic light emitting diode (OLED) monitor, or other display device.

In the example of FIG. 1, communication channel 16 may comprise any wireless or wireless communication means, such as a radio frequency (RF) spectrum or one or more physical transmission lines, or any combination of wired and wireless communication means. Communication channel 16 may form part of a packet-based network, such as a local area network, a wide area network, or a worldwide network such as the Internet. Communication channel 16 generally represents any suitable communication means, or collection of different communication means, for transmitting video data from source device 12 to destination device 14, including any suitable combination of the communication means with or wireless. Communication channel 16 may include routers, switches, base station, or any other equipment that may be useful to facilitate communication from source device 12 to destination device 14.

Video encoder 20 and video decoder 30 may operate in accordance with a video compression standard, such as ITU-T H.264, alternatively referred to as MPEG-4, Part 10, Advanced Video Encoding (AVC). Another example, video encoder 20 and video decoder 30 may operate in accordance with the High Efficiency Video Coding (HEVC) standard, and may conform to the HEVC (HM) Test Model. The techniques of this disclosure, however, are not limited to any particular coding model. Other examples include MPEG-2 and ITU-T H.263. Although not in FIG. 1, in some respects, video encoder 20 and video decoder 30 may each be integrated with an audio encoder and decoder, and may include appropriate MUX-DEMUX units, or other hardware and software, to handle both audio encoding and video in a single data stream or separate stream of data. If applicable, MUX-DEMUX units may conform to the ITU H.223 multiplexer protocol, or other protocols such as non-connection oriented (UDP) transport protocol. The ITU-T H.264 / MPEG-4 (AVC) standard was formulated by the ITU-T Video Coding Expert Group (VCEG) in conjunction with the ISO / IEC Motion Picture the product of a collective partnership known as the Joint Video Team (JVT). In some aspects, the techniques described in this disclosure may be applied to devices which generally conform to the H.264 standard. The H.264 standard is described in the ITU-T Recommendation H.264, Advanced Video Coding for Generic Audiovisual Services, by the ITU-T Study Group, and dated March 2005, which may be referred to herein as specification of the H.264 or H.264 standard, or the H.264 / AVC standard or specification. The Joint Video Team (JVT) continues to work on the extensions for H.264 / MPEG-4 AVC.

Video encoder 20 and video decoder 30 each may be implemented as any of a variety of suitable encoder circuit, such as one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuit (ASICs), network programmable logic gates (FPGAs), discrete logic, software, hardware, firmware or any combinations thereof. When techniques are implemented in software, a device may store instructions for the software in an appropriate, non-transitory, computer readable medium and execute the instructions using one or more processors to perform the techniques of this disclosure. Each video encoder 20 and video decoder 30 may be included in one or more encoders or decoders, either may be integrated as part of a codec / decoder combination (CODEC) into a respective camera, computer, mobile device, notification device , transmission device, set-top box, server, or the like.

A video sequence typically includes a series of video frames. An Image Group (GOP) generally comprises a series of one or more video frames. A GOP may include syntax data in a GOP header, a header of one or more GOP frames, or elsewhere, which describes a number of frames included in the GOP. Each frame may include a syntax data frame describing a coding mode for the respective frame. Video encoder 20 typically operates on encoding units in individual video frames in order to encode the video data. A coding unit may correspond to an LCU or a sub-CU, and the term CU may refer to an LCU or a sub-CU. Header information for an LCU can describe the size of the LCU, the number of times the LCU can be divided (referred to as CU depth in this disclosure), and other information. Each video frame may include a plurality of portions, and each portion may include a plurality of LCUs.

In some forecasting examples can be realized for various sizes of CU. The size of an LCU can be defined by syntax information. Assuming that the size of a specific CU sheet is 2Nx2N, intra-forecast sizes may include 2Nx2N or NxN in some examples, and symmetric inter-forecast sizes may include 2Nx2N, 2NxN, Nx2N, or NxN. In some examples, asymmetric separation can be used for inter prediction with sizes of 2NxnU, 2NxnD, nLx2N, and nRx2N. In asymmetric separation, one direction of a CU is not divided, while the other direction is divided into 25% and 75%. Whose portion of the CU is 25% divided is indicated by an "n" followed by an indication of "Up", "Down,""Left," or "Right," Thus, for example, "2NxnU" refers to a CU 2Nx2N that is

horizontally divided with a 2Nx5N PU on top and a 2Nxl.5N PU on the bottom.

In this disclosure, "NxN" and "N by N" can be used interchangeably to refer to the pixel dimensions of a block (eg, CU, PU, or TU) in terms of vertical and horizontal size, eg, 16x16 pixels or pixels 16 by 16. In general, a 16x16 block will have 16 pixels in a vertical direction (y = 16) and 16 pixels in a horizontal (x = 16) direction. Likewise, an NxN block generally has N pixels in a vertical direction and N pixels in a horizontal direction, where N represents an integer value greater than zero. The pixels in a block can be arranged in rows and columns. In addition, blocks need not necessarily have the same number of pixels in the horizontal direction as in the vertical direction. For example, blocks may comprise NxM pixels, where M is not necessarily equal to N. PUs of a CU may comprise pixel of data in spatial domains (also referred to as the domain of pixels), while TUs of CU may be transformed to produce coefficients in a transformation domain, eg, following a subsequent patent application a transformation such as a DCT, an integer transformation, a wave transformation, or a conceptually transformed similar to residual video data. The residual data generally represent different pixels between values of a PU and the values of the pixels placed and decoded from the video data input. The coefficients can still be quantified. The transformed TU coefficients can be said to be in the frequency domain.

Video encoder 20 may implement one or all of the techniques of this disclosure to improve encoding of transform units of a coding unit. In the same way, video decoder 30 may implement one or all of these techniques to improve decoding of the processing units of a coding unit. In general the techniques of this disclosure are directed to a transformation of coefficients of transformation units following calculations of the coefficients based on the intra-forecast mode. However, certain aspects of this disclosure may also be implemented in relation to inter-forecasting coding. For purposes of example, these techniques are described in relation to the intra prediction coding of TUs. It should be understood that certain aspects of these techniques may also be performed in conjunction with inter-prediction coding.

Video encoder 20 may receive an LCU and determine whether to divide the LCU into four quadrants, each comprising a sub-CU, or whether to encode the undivided LCU. Following a decision to divide an LCU into sub-CUs, video encoder 20 may determine whether to divide each sub-CU into four quadrants, each comprising a sub-CU. Video encoder 20 may continue to determine recursively whether to divide a CU, with a maximum number of divisions indicated by the LCU depth. Video encoder 20 may provide a quadtree data structure of the CU indicative of the division of an LCU and sub-CUs of the LCU. The LCU can match to a quadtree LCU root. Each quadtree CU node can correspond to an LCU CU. In addition, each node may include a value indicative of the division flag if the corresponding CU is divided.

If the LCU is split, for example, video encoder 20 can set the value of the split flag at the root to indicate that the LCU is divided. Then, video encoder 20 may define values of the root child nodes to indicate that, if any, the LCU sub-CUs are divided. A CU, which is not split may correspond to a sheet of the quadtree CU data structure, where the sheet does not have child nodes. In addition, each CU sheet may include one or more TUs, as indicated by a quadtree of TU for the CU sheet.

Video encoder 20 may encode each sub-CU of the corresponding LCU to a sheet in the quadtree data structure. For purposes of example, this disclosure describes the techniques regarding intra-prediction coding of the corresponding TUs for the CU sheet. In the intra coding mode, video coder 20 may form prediction units (PUs) for each corresponding TU for a sheet in the quadtree TU data structure. In some examples, video encoder 20 may select one of thirty-four different intra-forecast modes for the CU and signal the selected intra-prediction mode in the quadtree root of TU. Starting with a first, larger TU (equal in size to the CU sheet in the quadtree CU), video encoder 20 can determine whether to divide the larger TU, and, recursively, to divide sub-TUs of the relative TU. Video encoder 20 may further signal an intra-predictive mode in the quadtree CU of the sheet to the CU including the quadtree of the TU, where the predicted intra-prediction mode can describe the intra-predictive mode to be used to calculate predicted values for each of the TUs in TU quadtree corresponding to CU. Video encoder 20 retrieves the prediction data of the neighbors' TUs, previously encoded video data, according to the selected intra-forecast mode. In this way, PUs of a predicted CU using an intra prediction mode are the same size as CU TUs.

According to the techniques of this disclosure, if the selected intra-prediction mode is not available for the current CU, eg, because of its size, video encoder 20 may select an intra-prediction mode for which the mode signaled at the root of the quadtree is mapped. That is, video encoder 20 may include information that maps each of the modes from a large set of mode modes to a smaller set, eg, a subset of the large set, in a many-to-one match. Video encoder 20 may then intra-predict one or more PUs to CU using the intra-prediction mode of the smaller set. In this manner, video encoder 20 only needs to signal an intra-prediction mode to the LCU, although the video encoder 20 may use multiple modes to intra-predict LCU sub-CUs without explicitly signaling each mode and sub-CUs for which modes are used. Therefore, multiple intra-prediction modes can be used without increasing the amount of information included in the bit stream, thus reducing the overhead. In another embodiment, a greater number of prediction directions may be allowed at the CU level to allow using the same intra-predictive mode for the LCU independent of the CU sub-sizes or PU sizes.

Video encoder 20 may further be configured with threshold-based prediction mode to predict TUs in a CU which video encoder 20 determines to include a threshold. In general, a threshold corresponds to a high frequency shift along a straight line through the TU. For example, a boundary may occur along the edge of an object represented in TU in contrast to a background also represented in TU. To detect a boundary on a TU, video encoder 20 can calculate gradients for pixels in the TU and determine if the gradients identify a line in spite of TU. After determining that a current TU includes a threshold, video encoder 20 can determine whether to use predictive mode based on the threshold. If such a threshold is detected, and when the boundary-based prediction mode is selected, video encoder 20 may signal the use of the boundary-based predictive mode using a value which otherwise indicates the use of the DC predictive mode . That is, after detecting the presence of a limit in a current block, video encoder 20 may select an intra-forecast mode from a set including boundary-based predictive mode and other directional predictive mode (but excluding DC mode) , and when the boundary-based forecast mode is selected, signal the use of the boundary-based forecast mode as if you are using the DC-forecast mode.

In the sequence of intra prediction and prediction coding to produce predicted data for a TU of a CU, video coder 20 may calculate residual data, comprising TU coefficients representing different pixel-by-pixel values between the predicted data and the original data for the YOU. Video encoder 20 may form one or more TUs including residual data for the CU in this manner. Video encoder 20 can then transform the TUs. According to the techniques of this disclosure, video encoder 20 may select a transformation to apply to a TU with an intra-prediction mode used for intra-mode predictive data for TU.

In some examples, video encoder 20 may include configuration data that provides a many-to-one mapping between a large set of intra-predictive modes, and a smaller set of intra-predictive modes. For example, video encoder 20 may include configuration data that provides a mapping between the HM's intra-predictive modes and the H.264 intra-predictive modes. In addition, video encoder 20 may include configuration data that provides a mapping between the smaller set of intra-prediction modes and directional transformations. The set of directional transformations may be the same size as the smaller set of intra-prediction modes, such that it is a one-to-one mapping between the smaller set of intra-prediction modes and the set of directional transformations. In this way, the configuration data for video encoder 20 may provide an indirect, many-to-one mapping between the large set of intra-prediction modes and the set of directional transformations. Alternatively, in some examples, there may be a one-to-one mapping of the large set of directional transformations to a large set of directional transformations or other designed transformations, such as discrete cosine transform, discrete sine transform, or other transformations conceptually similar. However, using the mapping, video encoder 20 can select a transformation for each TU based on the intraprediction mode selected for a CU including TU.

In some examples, there may be multiple possible directional transformations for a TU of a given size. In some examples, video encoder 20 may signal a selected intra-prediction mode (eg, a selected intra-forecast direction) at the root of a quadtree data structure of the TU corresponding to a CU (i.e., a CU sheet in quadtree CU ), and the selected intra-forecast mode may apply to all TUs of CU. If all TUs in CU have size for which only one transformation is possible, then video encoder 20 may proceed according to the previous example, in which the transformation can be derived from an intra predicted mode signaled to the LCU. If at least one TU in CU is of a size for multiple transformations to be possible, however, then video encoder 20 can select one of the possible transformations and signal a selected transformation in the quadtree root of TU. Accordingly, video encoder 20 may use the signaling transformation to transform each TU into the CU having a size associated with multiple possible transformations. In this way, video encoder 20 may explicitly signal a transformation, without consuming excess excess bandwidth.

In some examples, when video encoder 20 applies an intrapredicate mode based threshold to predict a value for a TU, video encoder 20 may select a transform to apply to the TU based on a threshold angle. As discussed previously, video encoder 20 may determine that a threshold is present in a current TU based on the direction of a boundary on neighboring TUs that share an edge with the current TU. According to the techniques of this disclosure, video encoder 20 may calculate a relative angle of the boundary and use the angle of the boundary to select a directional transformation, similarly to select a directional transformation for an intra prediction mode. For example, video encoder 20 may compare the angle of the threshold for angles for the directional intra-prediction modes, determine a directional intra-prediction mode having an angle α which is close to the angle of the threshold, and then transform the predictive mode TU based on the end using the transformation that is mapped to the given intra-forecast mode.

In some examples, video encoder 20 may be configured to further apply a transformation to a TU, that the present disclosure relates to a cascade transformation. For example, the first transformation may correspond to a discrete cosine transform (DCT) or a Karhunen-Loève transform (KLT), also generally referred to as a directional transformation. When a directional transformation is selected based on an intrapredicate mode mapped to the directional transformation, the transformation can be referred to as a dependent mode directional transformation (MDDT). This disclosure also relates to a transformation selected on the basis of an intra prediction mode as a projected transformation, which may include directional transformations, discrete cosine transform, discrete sine transform, or other conceptually similar transformations specifically selected for a predictive mode . The second transformation may correspond to a rotation transformation. In some examples, video encoder 20 may be configured with multiple rotation transformations. Video encoder 20 can select one of the rotation transformations to apply by means of cost calculation of the Distortion Rate for each of the rotation transformations, in some examples. Video encoder 20 can be configured to apply a spin transformation to a number of coefficients smaller than the first transformation. According to the techniques of this disclosure, video encoder 20 may include configuration data for dependent mode rotation transformations (MDROT), including an array of transforming column and the array of transforming line. Intra-prediction modes can be mapped both to early transformations, eg, one of the MDDTs, as well as one of the rotation transformations, eg, one of the MDROTs. Thus, an intra predicted mode signaled to a CU may also provide an indication of a first transformation to apply to an LCU TU and a second transformation to apply to the TU. Although MDROTs are described as examples, it should be understood that the second transformation may comprise other transformations, such as directional transformations,

When transforming coefficients of a TU, video encoder produces an array of transformation coefficients. This array is the same size as TU. In general, a transformation process prepares the residual data for quantification, which compresses the data further. Quantification in general refers to a process in which the transformation coefficients are quantized to possibly reduce the amount of data used to represent the coefficients. The quantization process can reduce the bit depth associated with some or all coefficients. For example, an n-bit value can be rounded down to a m-bit value during quantization, where n is greater than m.

In some examples, video encoder 20 may use a preferred search order to search for quantized transform coefficients to produce a vector that can be encoded by entropy. For example, following a conventional transformation or a dependent mode transformation, video encoder 20 may be configured to perform a zig-zag search. Video encoder 20 may also be configured to apply a search based on an intra-predictive mode and / or one or more transforms applied to the block. In some examples, video encoder 20 may perform an adaptive search followed by transformation and quantification of the coefficients of a TU. In some examples, video encoder 20 may comprise configuration data defining different search schemes for each possible transformation scheme. For example, video encoder 20 may include configuration data comprising a one-to-one mapping between a set of directional transformations and a set of predefined search models. Research models can be defined based on empirical research tests following a specific directional transformation to optimize the placement of transformation coefficients in the vector following the corresponding directional transformation. Alternatively, video encoder 20 may include configuration data defining search indexes for which intra-prediction modes (or transform schemes) can be mapped, where search indices can indicate either predefined searches or adaptive searches.

Consequently, each directional transformation may have an associated search model that is relatively optimized for this directional transformation, based on empirical tests. As noted previously, video encoder 20 need not signal the directional transformation or search model used for a particular TU, assuming that there is a mapping between an intra prediction mode signaled a quadtree from TU to a CU including TU and directional transformation and research model. In several examples, search models may be dependent on a first selected transformation (eg, DCT or MDDT), second selected transformation (eg, MDROT, DCT, or other separate secondary transformation), or a combination of both. In some examples, one of the two cascade transformations may comprise a projected transformation applied in a particular direction (eg, horizontal or vertical), and video encoder 20 may select a search order generally corresponding to the same direction or a orthogonal direction, based on the configuration data.

In examples where video encoder 20 applies a cascade transformation to a TU, video encoder 20 can adaptively search for resulting coefficients of those resulting from the cascade transformation. To perform an adaptive search, video encoder 20 can generally track statistics indicatives if a particular position in the transformation coefficient matrix is more or less likely to be significant (eg, non-zero). Video encoder 20 can adapt the search model over time in such a way that search model matches these statistical probabilities. That is, the adaptive search model can attempt to ensure that transformation coefficients having a relatively greater probability of being significant (eg, non-zero) are searched before transformation coefficients have a relatively minor probability of being significant. Alternatively, video encoder 20 may select a search index for the cascade transformation to be mapped.

Video Encoder 20 can track search statistics for each cascade transformation possible separately. For example, the possibility that a location of the specific coefficient in a transformation matrix may differ based on the first and second transformations applied during the cascade transformation. Therefore, video encoder 20 can separately crawl independent sets of statistics for each possible cascade transformation. As an example, assuming intrapredictive modes are mapped to both an MDDT and an MDROT (or other separate secondary transformations), video encoder 20 can track independent statistics for each combination of MDDT and MDROT (or other secondary transformations) applied to TUs . Another example, video encoder 20 may be configured to perform cascade transformation only when video encoder 20 applies a DCT to a TU. Thus, video encoder 20 can perform adaptive search, and trace independent statistics to perform adaptive search, based on a selected MDROT (or other separate secondary transform) for the TU applied following DCT.

In some examples, if using an adaptive search or a predetermined search, video encoder 20 may zero out coefficients in the series after the search. That is, video encoder 20 can set values for coefficients at N positions to the end of the series equal to zero. The value of N may relate to the size of the CU and / or to the size of the TU. In some examples, video encoder 20 may zero transformation coefficients in the array prior to being searched, eg, all coefficients in the array except coefficients in the upper left corner of the array.

After searching the transformation matrix to form a one-dimensional vector, video encoder 20 may encode the one-dimensional vector, eg, according to content-adaptive variable-length coding (CAVLC), context-adaptive binary arithmetic coding (CABAC) binary arithmetic coding of adaptive context-based syntax (SBAC), or other coding entropy methodology.

To perform CAVLC, video encoder 20 may select a variable length code for a symbol to be transmitted. Code words in VLC can be constructed such that relatively shorter codes correspond to more probable symbols, while longer codes correspond to less probable symbols. In this sense, use or VLC can save a bit through, for example, using code words of equal length for each symbol to be transmitted.

To perform CABAC, video encoder 20 may select a context model to apply to a certain context to encode symbols to be transmitted. The context may relate to, for example, whether neighboring values are different from zero or not. Video encoder 20 can also encode entropy elements of syntax, such as a flag of the significant coefficient and a flag of the last coefficient produced when performing an adaptive search. According to the techniques of this disclosure, video encoder 20 may select the context model used to encode these syntax elements based, for example, in an intra prediction direction, a search position of the coefficient corresponding to the syntax elements, types block, and / or types of transform, among other factors used for context model selection.

Video decoder 30 may operate essentially symmetrically for such video encoder 20. For example, video decoder 30 may receive coded entropy data representative of an encoded CU, including encoded TU data. This received data may include information indicative of an intra-predictive mode used to encode the PU data, assuming the CU had intra-predicted coding. Video decoder 30 may reverse entropy code the received data, forming coded quantization coefficients. When video encoder 20 encodes data by using a variable-length code algorithm, video decoder 30 may use one or more VLC tables to determine a symbol corresponding to a received word code. When video encoder 20 encodes data by using an arithmetic code algorithm, video decoder 30 may use a context model to decode the data, which may correspond to the same context model used by the video encoder 20 to encode the data .

Video decoder 30 can then reverse the search for decoded coefficients by using a search inverter that mirrors the search used by video encoder 20. To reverse adaptive search the coefficients, video decoder 30 may decode syntax elements including significant coefficient flag and flag the last coefficients to regenerate the statistics used by the video encoder 20 to perform the adaptive search. Video decoder 30 may thus form a two-dimensional array of the one-dimensional vector resulting from the entropy decoding process. Subsequently, video decoder 30 may invert quantify the coefficients in the two-dimensional array produced by the search inverter. Video decoder 30 can then apply one or more inverse transformations to the two-dimensional array. The inverse transformations may correspond to the transformations applied by the video encoder 20. Video decoder 30 can determine the inverse transformations to be applied based, for example, on the intra-prediction mode used to calculate coefficients for the TU, and whether multiple transformations are available for a TU of a certain size, information flagged in the root of a TU quadtree corresponding to CU including TU currently being decoded. In this manner, video decoder 30 may select one or more inverse transformations to be applied to invert quantized coefficients for a TU to reproduce the TU, based on an intra predicted mode signaled to the TU. In addition, video decoder 30 may calculate a predicted TU value using an intra predictive mode corresponding to a signaled indication of the intra predictive mode, eg, in TU quadtree. In some examples, video decoder 30 may determine that a threshold is present in the TU, based on a threshold detected in the neighboring TU, and, when a DC mode is signaled, instead of using an end-based mode to predict a value for TU.

Video encoder 20 and video decoder 30 each may be implemented as any of a variety of suitable encoder or decoder circuit, as the case may be, such as one or more microprocessors, digital signal processors (DSPs), specific application integrated circuit (ASICs), programmable logic gates (FPGAs), discrete logic circuits, software, hardware, firmware or any combination thereof. Each video encoder 20 and video decoder 30 may be included in one or more encoders or decoders, the one of which may be integrated as part of a combination of video encoder / decoder (CODEC). An apparatus including video encoder 20 and / or video decoder 30 may comprise an integrated circuit, a microprocessor, and / or a wireless communication device, such as a cellular telephone. FIG. 2 is a block diagram illustrating an example of video encoder 20 which may implement one or all of the techniques for encoding video data transformation units described in this disclosure. Video encoder 20 can perform intra and inter-encoding of CUs in video frames. Intra coding depends on spatial prediction to reduce or remove spatial redundancy in the video in a given video frame. Inter encoding depends on the temporal prediction to reduce or remove temporal redundancy between a current frame and previously encoded frames of a video clip. Intra mode (I-mode) can refer to multi-spatial variability based on compression modes and inter modes such as unit-directional prediction (P-mode) or bi-directional prediction (B-mode) to any of the various time base compression modes.

As shown in FIG. 2, video encoder 20 receives a current video block within a video frame to be encoded. In the example of FIG. 2, video encoder 20 includes motion compensation unit 44, motion estimation unit 42, intra prediction unit 46, store reference frame 64, adder 50, transformation unit 52, quantizer unit 54, and coding unit of entropy 56. Transforming unit 52 shown in FIG. 2 is the unit that performs the current transformation, not to be confused with a TU, of a CU. For reconstruction of the video block, video encoder 20 also includes Reverse Quantitation Unit 58, Reverse Transformation Unit 60, and Adder 62. Additionally, video encoder 20 may include configuration data, such as mapping data 66. A filter (not in FIG. 2) may also be included for filter block limits to remove reconstructed video blocking artifacts. If desired, the unlock filter would typically be mottling filter of adder 62.

During the encoding process, video encoder 20 receives a frame or video segment to be encoded. The frame or segment can be divided into multiple video blocks, eg, larger encoding unit (LCUs). Motion estimation unit 42 and motion compensation unit 44 perform inter-prediction coding of the received video block relative to one or more blocks in one or more reference frames to provide time compression. Intra-prediction unit 46 may perform intra-prediction coding of the received video block relative to one or more neighboring blocks in the same frame or segment as the block to be encoded to provide spatial compression.

Mode select unit 40 may select one of the intra- or inter-coding modes eg, based on error results (sometimes termed as distortion), and provides the resultant of the intra- or inter-coded block for the adder 50 to generate residual data from the block and adder 62 to reconstruct the coded block for use in a reference frame. Some video frames may be referred to as I-frames, where all blocks in the I-frame are encoded in an intra-prediction mode. In some instances, intra prediction unit 46 may perform intra-prediction coding of a block in a P- or B-frame, eg, when motion search performed by motion estimation unit 42 does not result in a sufficient block prediction.

Motion estimation unit 42 and motion compensation unit 44 may be highly integrated, but are shown separately for conceptual purposes. Motion estimation is the process of generating motion vectors, which calculates motion for video blocks. A motion vector, for example, may indicate the displacement of a prediction unit in a current frame relative to the reference sample of a reference frame. The reference sample is a block that is found to strictly match the portion of the CU including the PU being coded in terms of pixel difference, which can be determined by the sum of the absolute difference (SAD), sum of the square difference (SSD) , or other different metrics. Motion compensation, performed by the motion compensation unit 44, may involve reaching or generating values for the prediction unit based on the motion vector determined by the motion estimate. Again, motion estimation unit 42 and motion compensation unit 44 may be functionally integrated, in some examples.

Motion estimating unit 42 calculates a motion vector for a prediction unit of an inter-coded frame by comparing the prediction unit for reference samples of a reference frame stored in the store of reference frame 64. In some examples, encoder of video pixel positions 20 can calculate values for positions of the sub-integer pixel of the reference frames stored in the reference frame store 64. For example, video encoder 20 can calculate values of the one-quarter pixel positions, one-eighth pixel positions , or other positions of other pixel fractions of the reference frame. Therefore, motion estimation unit 42 may perform a motion search relative to the positions of the full pixel and positions of other pixel fractions and sends a motion vector with fractional pixel precision.

Motion estimation unit 42 sends the calculated motion vector to entropy coding unit 56 and motion compensation unit 44. The portion of the reference frame identified by a motion vector may be referred to as a reference sample. Motion compensation unit 44 may calculate a prediction value for a prediction unit of a current CU, eg, by retrieving the reference sample identified by a motion vector for the PU.

Intra-prediction unit 46 may intra-predict coding the received block as an alternative to prediction performed by the motion estimation unit 42 and motion compensation unit 44. Intra-prediction unit 46 may encode the received block relative to the neighborhood, blocks previously encoded, eg, blocks on top, on top on the right, on top is left, or left of the current block, assuming a coding order from left to right, top to bottom for blocks (such as a sort scanning). Intra-prediction unit 46 can be configured with a variety of different intra-forecasting modes. For example, intra prediction unit 46 may be configured with a certain number of directional prediction modes, eg, directional prediction modes, based on the size of the CU to be encoded.

Intra-prediction unit 46 can select an intra-prediction mode by, for example, calculating error values for various intra-forecasting modes and selecting a mode that produces the smaller error value. Directional predictive mode may include functions for combining neighboring spatial pixel values and applying the combined values to one or more pixel positions in a PU. Once the values for all pixel positions in the PU have been calculated, intra prediction unit 46 may calculate a different pixel based prediction mode error value between the PU and the received block to be coded.

Intra-prediction unit 46 may continue to test intra-predictive modes until an intra-prediction mode that produces an acceptable error value is encountered. Intra-prediction unit 46 may then send the PU to adder 50.

According to the techniques of this disclosure, intra-prediction unit 46 may be configured to predict a block including a threshold using a threshold-based prediction mode. In particular, intra-prediction unit 46 may analyze neighboring pixels, previously encoded blocks to determine if a threshold is detected in at least one of the neighboring blocks, and if the boundary intersects a boundary between the previously encoded block and the current block. To detect the boundary, intra prediction unit 46 can calculate gradients for pixels in neighbors, blocks previously encoded in both horizontal and vertical directions. When the gradients for a plurality of pixels in one of the neighboring blocks are relatively perpendicular to a common line crossing a boundary between the neighboring block and the current block, intra prediction unit 46 may determine that the current block also includes a limit particularly, a limit along the line detected as described above). It is to be understood that the term "boundary" in this context refers to a high frequency shift along a straight line within a block of pixels, and not to the boundary or edge between separately encoded blocks.

Accordingly, when a threshold is detected in the block, intra prediction unit 46 may determine whether to predict the block using a boundary-based predictive mode or a directional intra-prediction mode. When intra-forecast unit 46 selects the boundary-based prediction mode, intra-forecast unit 46 may signal which DC forecast mode was used to predict the block, in order to avoid an increase in the number of values needed to signal the intra-forecasting modes. This, as discussed in more detail with respect to FIG. 3, a video decoder such as video decoder 30 may be configured to interpret an indication (eg, a signaling or syntax information) that the DC prediction mode has been used to predict a block, has an indication of a prediction mode based on the threshold when the video decoder determines that a threshold is present in the block.

Video encoder 20 forms a residual block by subtracting the prediction data computed by the motion compensation unit 44 or intra prediction unit 46 of the original video block to be encoded. Adder 50 represents the component or components that perform this subtraction operation. The residual block may correspond to a two-dimensional array of values, where the number of values in the residual block is the same as the number of pixels in the PU corresponding to the residual block. The values in the residual block may correspond to the differences between pixels placed in the PU and in the original block to be coded.

Processing unit 52 may form one or more transformation units (TUs) of the residual block. Transforming unit 52 applies a transformation, such as a discrete cosine transform (DCT), directional transformations, or a conceptually similar transformation, to TU, producing a video block comprising transform coefficients. According to the techniques of this disclosure, intra-prediction unit 46 may send an indication of the selected intra-forecast mode of TU to the processing unit 52, eg, by signaling the mode in a node or a quadtree of TU corresponding to TU. Accordingly, transformation unit 52 may select a transformation to apply to TU based on the indication of the intra-forecast mode received from the intra-prediction unit 46.

In some examples, transformation unit 52 is configured to select a transformation, such as a directional transformation, to apply to TU based on the intra-prediction mode used to predict TU. That is, mapping data 66 may store configuration data describing a transformation to be applied based on the intra-prediction mode used to predict TU. In this manner, transformation unit 52 may transform a residual block using a mapped transform to an intra-prediction mode used to calculate the residual block. Likewise, mapping data 66 may map an intraprediction mode, a transformation, or both, to a specific search index, which can be used to select a search to be applied to quantized transformation coefficients of the block.

In some examples, mapping data 66 may store configuration data including in a many-to-one mapping between a large set of intra-forecasting modes and a smaller set of intra-forecasting modes. The smaller set may comprise a subset of the intra-forecasting modes. In some examples, the large set may comprise intra prediction modes supported by the HEVC Test Model, while the smaller set may comprise intraprediction modes supported by H.264. Mapping data 66 may also include a mapping, such as a one-to-one mapping, between the smaller set of intra-prediction modes and transformations. The smaller set of intra-prediction modes can be mapped one-to-one for directional transformations that are designed to empirically provide the best transformation results for the corresponding intra predictive mode.

In some examples, intra-forecasting modes of the large set and the smaller set can be associated with their respective prediction angles. The mapping between the large set of the prediction mode and the smaller set can therefore be determined by selecting a prediction angle from one of the intra-forecast modes in the smaller set that approximates the prediction angle of one of the intra-forecast modes of the large set. If α represents the prediction angle of one of the intra-forecasting modes of the large set, and ββ represents the prediction angle of the intra-forecast mode of the smaller set. To map the intra-prediction mode to one of the intra-prediction modes of the smaller set, video encoder 20 may, given a, find pi such that the following equation 1 is satisfied: argjpi ·. min (min (abs (a - pi), abs (- <x - βί))) (1)

Transformation unit 52 may receive the indication of the selected intra-prediction mode from the intra-prediction unit 46, then queries mapping data 66 to determine a transformation to be used to transform a TU including the residual values received from the adder 50. If TU is of a size so that multiple possible transformations are possible, transformation unit 52 or other video encoder unit 20 may select a transformation for TUs of that size, such that transformation unit 52 may apply the same transformation to all TUs of that size in same CU. In some examples, transformation unit 52 may further be configured to perform a rotational cascade transform of the first transformation. That is, following the first transformation, transformation unit 52 can select and apply a rotation transformation to the transformation coefficients. The transformation unit 52 may select a rotation transformation based, for example, from the intra-prediction mode used to predict the PU to the current TU.

As discussed earlier, in some examples, intra prediction unit 46 is configured to determine if a block includes a threshold (i.e., a high frequency shift between pixels within the block). When a threshold is detected, intra prediction unit 46 may select either a boundary-based predictive mode or a conventional directional intra-predictive mode. That is, intra-forecast unit 46 may override DC forecast mode for the boundary-based forecast mode. In some examples, when the block is predicted using the boundary-based prediction mode, transformation unit 52 selects a transformation (such as a directional transformation) mapped to the direction of an intra-prediction mode having an angle approaching an angle of the limit. That is, transformation unit 52 in some examples determines an angle of the boundary within the block and selects a transformation that is mapped to a directional intraprediction mode having an angle (eg, has a relative minimum distance) of the angle of the limit. In some examples, transformation unit 52 is configured to select a transformation that is mapped to a directional intra-prediction mode having an angle approaching the angle of a boundary detected in a block.

Video Encoder 20 can also signal the use of the predictive mode based on the threshold using a value that would otherwise be used to signal the use of the DC predictive mode. Thus, although DC predictive mode is signaled, the boundary-based predictive mode can be used to predict the block. Likewise, although the boundary-based predictive mode may be used to predict the block, transforming unit 52 may use the mapped transformation to the intra-prediction mode having an angle approaching the angle of the boundary detected in the block.

Mapping data 66 may provide configuration data which indicates when a threshold is detected in a block for which DC predictive mode is signaled, transforming unit 52 is for selecting a transform having an angle approaching the angle of the threshold. In addition, as previously discussed, mapping data 66 may include mapping of intra-prediction modes and angles for intra-prediction modes (which may define intra-prediction modes) for directional transformations. Accordingly, transformation unit 52 may query mapping data 66 to determine an intra-prediction mode having an angle approaching the angle of a boundary in a block, as well as to determine the transformation that is mapped to the particular intra-prediction mode.

In this manner, video encoder 20 is an example of a video encoder configured to compute a residual block for a video data block based on a predicted block formed using an intra prediction mode, and transform the residual block using a mapped transform for intra-forecasting mode. Video encoder 20 is also an example of a video encoder configured to receive an indication of a first intra-prediction mode in a first set of intra-prediction modes for a video data block, determining a second intra-prediction mode of a set of intra-predictive modes so that the first intra-prediction mode is mapped, determine a directional transformation so that the second intra-prediction mode is mapped, and apply the directional transformation to residual data of the block.

In addition, video encoder 20 is also an example of a video encoder configured to select an intra-predictive mode to use for encoding a video data block, determining whether the block includes a sub-block of a size so that multiple transformations are possible based on the sub-block size and the selected intra-prediction mode when the block includes the sub-block of the size so that multiple transformations are possible based on the sub-block size and the selected intra-forecast mode , select one of multiple possible transformations, transform the sub-block using the selected one of the multiple possible transformations, and provide an indication of the selected of the multiple transformations possible for the block size.

In addition, video encoder 20 is an example of a video encoder configured to determine that a block to be intra-predicted coding contains a boundary within the block, calculating a residual block for the block based on a predicted value calculated using a transforming the residual block using a directional transformation mapped to a directional intra-prediction mode having an angle approaching a boundary angle, and sending information representative of the transformed residual block and information indicating that the block was predicted using an intra-DC prediction mode.

Processing unit 52 may send the resulting transform coefficients to the quantizing unit 54. The quantizing unit 54 may then quantify the transformation coefficients. In some examples, quantizing unit 54 may then perform a search of the array including the quantized transformation coefficients. Alternatively, entropy coding unit 56 may perform the search. This disclosure describes the entropy coding unit 56 carried out the research, although it is understood that in other examples, other processing units, such as quantitation unit 54, can perform the search.

In some examples, entropy coding unit 56 may receive an indication of the selected intra-prediction mode of the intra-prediction unit 46 or of the transform unit 52. The entropy coding unit 56 may select a search to apply to the array of coefficients of transform, to convert the two-dimensional array into a one-dimensional vector In some examples, entropy-coding unit 56 selects a search from a predetermined set of searches. Mapping data 66 can map the smallest set of intra-forecasting modes to the predetermined set of searches. Entropy coding unit 56 can select the search based on various features of the current TU, such as, for example, block type (inter or intra), intra-prediction mode (assuming an intra-coded block), and / or a type of transformation applied to TU (eg, DCT or KLT).

In some examples, entropy coding unit 56 may be configured to perform an adaptive search. Initially (eg, for a first TU of a current frame), entropy coding unit 56 may use a predetermined search model. Over time, entropy coding unit 56 can update the search model to perform the adaptive search. In general, the goal of adaptive research is to determine a probability that a specific transformation coefficient will be different from zero. Then, the search order generally follows from coefficients with the highest probability being nonzero for the lowest probability being nonzero. Entropy coding unit 56 can determine these probabilities over time using various statistics and calculations. In addition, entropy coding unit 56 can track separate statistics for each mode of intra prediction, transformation, cascade transformation, or any combination thereof.

Entropy coding unit 56 can use a high dynamic range table and dynamic update lookup tables to determine the probabilities of transform coefficients being nonzero, and to determine the order of the adaptive search. Assuming a TU NxN, each of these tables can be NxN tables with values corresponding to TU transformation coefficients. The high dynamic range table can be a fixed, predetermined table providing probabilities that each transformation coefficient is nonzero. This table can be calculated based on a set of training data. In addition, this table can be used to provide the starting point for the order of the adaptive search.

Entropy coding unit 56 can update the dynamic update lookup table over time to reflect recently determined statistics for transform coefficients. Entropy coding unit 56 can keep a count of the number of times that each coefficient at a specific location in the NxN transformation matrix is nonzero. That is, for each TU of a current frame, entropy coding unit 56 can increment values in the dynamic update lookup table corresponding to nonzero coefficients in the current transform block, specifically, in the dynamic update lookup table associated with the intra-prediction mode, transformations, and / or cascade transformation for the current CU. For example, if the transformation coefficient in row 2 and column 1 is nonzero, entropy coding unit 56 can add one to the value in the dynamic update lookup table in row 2 and column 1. Entropy coding unit 56 can also periodically normalize the values in the dynamic update lookup table to avoid values exceeding a maximum value.

To perform the adaptive search for a first TU of a current frame, entropy coding unit 56 can search based on the elevated dynamic range table alone. Entropy Encoding Unit 56 can also initialize the dynamic update lookup table by, eg, set all values in the dynamic update lookup table to zero. For each non-zero coefficient in the transformation block, entropy coding unit 56 may add one to the value placed in the dynamic update lookup table associated with the intra-prediction and transformation mode or cascade to the current TU. For subsequent TUs using the same mode of intra prediction and transformation or cascade transformation, entropy coding unit 56 can first refer to the dynamic update lookup table to determine which of the transformation coefficients is most likely to be nonzero, then search in order to decrease the probability of the coefficients being nonzero. In some cases, two or more values in the dynamic update lookup table may be the same. In such a case, entropy coding unit 56 refers to the high dynamic range table to determine which coefficient to search next. In this way, entropy coding unit 56 may perform an adaptive search for each intra-prediction, transformation, or cascade transformation mode (or any combination thereof) based on a combination of a high dynamic range table and a lookup table dynamic update.

By investigating the two-dimensional array of the transformation coefficients, entropy coding unit 56 can produce a series of one dimension including the transformation coefficients. Entropy coding unit 56 can then search the TU to form a series and quantify the transformation coefficients in the series following the search, additionally reduce the bit rate. The quantization process may reduce the bit depth associated with some or all of the coefficients. The quantification level can be coded by adjusting a quantification parameter.

Entropy coding unit 56 can also encode entropy syntax elements for the matrix coefficients before or during adaptive searching. The syntax elements may include a significant coefficient flag that indicates whether a specific coefficient is significant (eg, other than zero) and a flag of the last coefficient that indicates whether a specific coefficient is the last coefficient searched in the adaptive search. A video decoder can use these syntax elements to reconstruct the dynamic update lookup table, such that video decoder can reverse the search of the coefficients encoded by the entropy coding unit 56.

To encode the entropy elements, the entropy coding unit 56 can perform CABAC and select contextual model based, for example, on the number of significant coefficients in the N coefficients previously searched, where N is an integer value that can be related to the size of the block to be searched. Entropy coding unit 56 can also select the context model based on a prediction mode used to calculate residual data that has been transformed into the transform coefficient block, and a type of transformation used to transform the residual data into the coefficient block of transformation. When the corresponding prediction data has been predicted using an intra prediction mode, entropy coding unit 56 may further substantiate the selection of the context model in the direction of the intra prediction mode.

In this manner, video encoder 20 represents an example of a video encoder configured to transform residual data to a video data block using a first transformation to produce an intermediate, two-dimensional block of transform coefficients, transforming the intermediate block two-dimensional transformation coefficients using a spin transformation to produce a two-dimensional block of transformation coefficients, selecting a set of statistics associated with at least one of the first transformations and the spin transformation, wherein the set of statistics provides possibilities that the locations in the two-dimensional block of transformation coefficients will be different from zero, and will adaptively search the two-dimensional block of transformation coefficients based on the selected set of statistics.

Video encoder 20 also represents an example of a video encoder configured to search a two-dimensional block of transform coefficients to produce a one-dimensional vector of transform coefficients, determine values indicative if the transform coefficients in the one-dimensional vector are significant, and encoding entropy at least one of the values using a selected context model based on at least a percentage of significant coefficients in a predetermined number of the encoded values before at least one of the values.

In some examples, processing unit 52 may be configured to zero certain transformation coefficients (i.e., transformation coefficients in certain locations). For example, transformation unit 52 may be configured to reset all transform coefficients outside the upper left quadrant of the TU following the transformation. Another example, entropy coding unit 56 may be configured to zero transform coefficients in the series following a certain position in the series. In some examples, entropy coding unit 56 can quantify a two-dimensional array, and entropy coding unit 56 can perform the search. In any case, video encoder 20 may be configured to zero a certain portion of the transform coefficients, eg, before or after the search. The phrase "zeroing" is used to mean setting coefficients equal to zero, but not necessarily ignoring or discarding coefficients.

Following quantification, entropy coding unit entropy 56 encodes the quantized transformation coefficients. For example, entropy coding unit 56 may perform content adaptive variable length coding (CAVLC), context-adaptive binary arithmetic coding (CABAC), or other entropy coding technique. Following the entropy coding unit by the entropy coding unit 56, the encoded video can be transmitted to another device or archived for later transmission or retrieval. In the case of adaptive binary arithmetic coding of the context, context may be based on the neighboring blocks.

In some cases, entropy encoding unit 56 or other video encoder unit 20 may be configured to perform other encoding functions in addition to entropy encoding. For example, entropy coding unit 56 may be configured to determine values of coded block (CBP) models for the blocks. Also, in some cases, entropy coding unit 56 may perform the coding of run length of the coefficients in a block.

Inverse quantization unit 58 and inverse transform unit 60 apply inverse quantization and inverse transformation, respectively, to reconstruct the residual block in the pixel domain, eg, for later use with a reference block. Motion compensation unit 44 may calculate a reference block by adding the residual block to a prediction block to one of the reference frame storage frames 64.

Motion Compensation Unit 44 may also apply one or more interpolation filters to the reconstructed residual block to calculate values of the sub-integer pixel for use in motion estimation. Adder 62 adds the reconstructed residual block to the compensated motion prediction block produced by the motion compensation unit 44 to produce a reconstructed video block for storing in the reference frame storage 64. The reconstructed video block can be used by the unit motion estimation unit 42 and motion compensation unit 44 as a reference block for inter-encoding a block in a next video frame. FIG. 3 is a block diagram illustrating an example of the video decoder 30, which decodes an encoded video sequence. In the example of FIG. 3, video decoder 30 includes an entropy decoding unit 70, motion compensation unit 72, intra prediction unit 74, inverse quantization unit 76, inverse transform unit 78, reference frame 82 and adder 80 storage. Video decoder 30 may, in some examples, perform a generally reciprocal decoding passage for the described coding passage relative to the video encoder 20 (FIG 2). Motion compensation unit 72 can generate prediction data based on motion vectors received from the entropy decoding unit 70. Intra-prediction unit 74 may generate prediction data for a current block of a current frame based on an intra mode predicted signal and data from previously decoded blocks of the current frame.

In some examples, entropy decoding unit 70 or inverse quantizing unit 76 may search the received values using a search mirroring used by the video encoder 20. In the example of FIG. 3, video decoder 30 includes mapping data 84, which may include similar or identical data of the mapping data 66. Consequently, video decoder 30 may select a search based, for example, on an indication of an intra-coding mode for a current block (eg, presented in the root of a quadtree for the LCU including the current block), a transformation to the current block, a cascade transformation to the current block, or the other factors used by the video encoder 20 to select the search. Likewise, video decoder 30 may be configured to perform an adaptive search, or to select a predetermined search based on these factors. In this manner, video decoder 30 may produce a two-dimensional array of transform coefficients quantized from a received, one-dimensional array of coefficients.

Inverse quantization unit 76 inverts quantizations, ie, de-quantifies, the quantized transform coefficients provided in the bit stream and decoded by the entropy decoding unit 70. The inverse quantization process may include a conventional process, eg, as defined by H.264 decoding standard or HEVC. The inverse quantization process may include the use of a quantization, QPy parameter calculated by the video encoder 20 for the CU to determine a level of quantification, and likewise a level of inverse quantization to be applied.

Inverse transformation unit 58 applies an inverse transformation, eg, a reverse DCT, a reverse integer transformer, a reverse rotation transformation, or a reverse directional transformation. In some examples, inverse transform unit 78 may determine an inverse transformation based on an in-prediction mode signaled to a received intra-predicted coding block. If the block is of a size for more than one transformation is possible, based on the intra-predictive mode, then inverse transformations unit 78 can determine a transformation to apply to the current block based on the transformation signaled in the root of a quadtree to an LCU including the current block. In some examples, inverse transformation unit 78 may apply a reverse cascade transformation, eg, first a reverse rotation transformation followed by a reverse directional transformation.

In some examples, eg, where the predicted intra-prediction mode is predictive mode DC, inverse transform unit 58 (or other video decoder unit 30) can determine whether a threshold is present in the current block. Inverse transformation unit 58 may determine whether the limit is present using techniques corresponding substantially to those described with respect to the video encoder 20 of FIG. 2. If a boundary is present in the current block, inverse transform unit 78 can determine an angle of the boundary within the block and select an inverse transformation that is mapped to an intra-prediction mode having an angle approaching the angle of the boundary.

As discussed previously, mapping data 84 may provide angles for intra-forecasting modes and mapping between intra-forecasting and reverse transformation modes. Consequently, reverse transformation unit 78 may query mapping data 84 to determine a reverse transformation mapped to an intra prediction mode having an angle approaching the angle of the boundary, when predictive mode DC is signaled. In addition, intra prediction unit 74 may apply a prediction mode based on the threshold to predict the block, rather than the DC prediction mode as signaled to the block, when the threshold is detected on the block. Mapping data 84 may also provide a mapping of an intra-prediction mode, a secondary transformation such as a rotation transformation, or a combination thereof, to a search index, to select a reverse lookup for received quantized transformation coefficients.

Motion compensation unit 72 produces offset motion blocks, possibly performing interpolation based on interpolation filters. Identifiers for interpolation filters to be used for motion estimation with sub-pixel precision can be included in the syntax elements. Motion compensation unit 72 may use interpolation filters as used by video encoder 20 during video block coding to calculate interpolated values for sub-integer pixels of a reference block. Motion compensation unit 72 may determine the interpolation filters used by the video encoder 20 in accordance with received syntax information and use the interpolation filters to produce prediction blocks.

Motion compensation unit 72 and intra prediction unit 74 use some of the syntax information (eg, provided by a quadtree} to determine sizes of LCUs used to encode frame (s) of the encoded video sequence, to divide information describing how each CU of a frame of the coded video sequence is divided (and in the same way, as sub-CU are divided), modes indicating how each division is coded (eg, intra or inter prediction, and for intra prediction an intra coding mode one or more reference frames (and / or reference lists containing identifiers for the reference frames) for each inter-coded PU, and other information for decoding the encoded video sequence.

Adder 80 mixes the residual block as the corresponding prediction blocks generated by the motion compensation unit 72 or intra prediction unit 74 to form decoded blocks. If desired, an unlock filter can also be applied to filter the decoded blocks in order to remove blocking artifacts. The decoded video blocks are then stored in the reference frame store 82 which provides reference block to follow motion compensation and also produces decoded video for display on a display device (such as display device 32 of Figure 1) .

In this manner, video decoder 30 is an example of a video decoder configured to determine an intra-predictive mode to be used to predict a video data block, and reverse transform of residual block data transformers using a mapped inverse transform for intra-forecasting mode. Video decoder 30 is also an example of a video decoder configured to receive an indication of a first intra prediction mode in a first set of the intra prediction mode for an encoded video data block, determining a second intra prediction mode from a smaller set of intra-predictive modes for the first intra-prediction mode to be mapped, to determine a reverse directional transformation for the second intra-prediction mode to be mapped, and to apply the inverse directional transformation to transformed residual data of the block.

Video decoder 30 is also an example of a video decoder configured to receive a first indication of an intra-predictive mode for use to decode a video data block, determining whether the block includes a sub-block of a size for that multiple transformations are possible based on the sub-block size and the predicted intra-prediction mode when the block includes the sub-block of the size so that multiple inverse transformations are possible based on the size of the sub-block and the intra-mode -representation, receive a second indication of one of the multiple possible inverse transformations, and inverse transformation of the sub-block using the one of the multiple inverse transformations possible.

Video decoder 30 is also an example of a video decoder configured to receive indicative values if transformation coefficients encoded in a one-dimensional vector received from the encoded transformation coefficients are significant, decoding of at least one of the entropy values using a context model selected with based on at least a percentage of the significant coefficients in a predetermined number of the decoded values before at least one of the values, and invert one-dimensional vector search to produce the two-dimensional block of the transform coefficients.

Video decoder 30 is also an example of a video decoder configured to receive an indication that the residual data for a video data block has been transformed using both a first transformation and a rotation transformation to produce a two-dimensional block of video transformation coefficients, select a set of statistics associated with at least one of the first transformations and the rotation transformations, where the set of statistics provides possibilities that locations in the two-dimensional block of transform coefficients will be different from zero, and adaptively inverts searching for a received one-dimensional vector including a version of the residual data encoded for the block based on the selected set of statistics to produce a two-dimensional array of transform coefficients for the block.

Video decoder 30 is still an example of a video decoder configured to receive information indicating that an intra predictive mode for a video data block is an intra predictive mode DC, determining an angle to a threshold in the data block of video based on the intra-prediction DC mode indication for the block, inverse block transformation using a directional reverse transformation mapped to a directional intra-prediction mode having an angle approaching the boundary angle, and decoding the inverse transformed block. FIG. 4 is a conceptual diagram illustrating a graph 104 which describes an example set of directions associated with intra-predictive modes. In the example of FIG. 4, block 106 can be predicted from neighboring pixels 100A-100AG (neighboring pixels 100) depending on a selected intra-prediction mode. Arrows 102A-102AG (arrows 102) are representative of directions or angles associated with various intra-forecasting modes. The example of FIG. 4 is representative of intra predictive modes provided by HM. However, in other examples, more or less intra-predictive modes may be provided. Although the example of block 106 is an 8x8 pixel block, in general, a block can have any number of pixels, eg, 4x4, 8x8, 16x16, 32x32, 64x64, 128x128, and so on. Although HM provides for square PUs, the techniques of this disclosure can also be applied to other block sizes, eg, NxM blocks, where N is not necessarily equal to M.

An intra-predictive mode may be defined according to an angle of the prediction direction relative to, for example, a horizontal axis which is perpendicular to the vertical sides of the block 106. Thus, each arrow 102 may represent a specific angle of one direction predictive mode. In some examples, an intra-prediction direction mode can be defined by the pair of integers (dx, dy), which can represent the direction, the corresponding intra-prediction mode used for the extrapolation pixel context.

That is, the angle of the intra prediction mode can be calculated as dy / dx. In other words, the angle can be represented according to the horizontal displacement dx and the vertical displacement dy. The value of a pixel at the location (x, y) in block 106 can be determined from one of neighboring pixels 100 through which a line passing through the location (x, y) also passes through an angle of dy / dx.

In some examples, each of the intra-prediction modes corresponding to the angles represented by the arrows 102 may be mapped to a specific transformation. Two or more intra-forecasting modes can be mapped to the same transformation in some examples. Transformations may correspond to directional transformations, KLTs, spin transformations, discrete cosine transform, discrete sine transform, Fourier transformations, or other transformations that are specifically selected for specific intra-forecast modes. Such transformations can be generally termed as "conceived" transformations, in which transformations are selected for specific modes of intra-forecasting.

As discussed previously, in some examples, a set of intra-predicate modes can be mapped by a many-to-one match to a smaller set, eg, a subset, of intra-forecast modes. Stated another way, the angles for the intra prediction modes in a large set of modes can be mapped to the angles of the intra prediction modes in a smaller set of modes. In some examples, the mapping can be performed using a mathematical formula. For example, the formula may provide a mapping that minimizes the absolute prediction angle difference between the current prediction angle direction, referred to herein as a, and directions of prediction angles of a smaller set of intra prediction modes, referred to herein as βζ. Given a direction of prediction angle α, the formula can provide a βζ such that formula (1) above is satisfied. Formula (1) is updated below for convenience: arg {| ji} min (min (abs (a - pi), abs (-a - βί))) (1)

In one example, the smaller set of intra-prediction modes may have angles having the same arrows 102E, 1021, 102M, 102Q, 102U, 102Y, I02AC, and 102AG. Thus, each of the angles of the arrows 102 can be mapped to one of the angles for the arrows 102E, 1021, 102M, 102Q, 102U, 102Y, 102AC, and 102AG. As an example, angles for the arrows 102A-102E can be mapped to the angle of the arrow 102E, angles for the arrows 102F-102I can be mapped to the angle of arrow 1021, angles for arrows 102J-102M can be mapped to the angle of the arrow 102A. arrow 102M, angles for arrows 102N-102Q can be mapped to angle of arrow 102Q, angles for arrows 102R-102U can be mapped to angle of arrow 102U, angles for arrows 102V-102Y can be mapped to angle of the arrow 102Y, angles for arrows 102Z-102AC can be mapped to the angle of arrow 102AC, and angles for arrows 102AD-102AG can be mapped to the angle of arrow 102AG.

Other mappings can also be provided. In some examples, video encoder 20 and video decoder 30 may be configured with a variety of different mappings, and video encoder 20 may provide map information information used for a specific bit stream, eg, in the header data, a sequence of parameter set (SPS), or other flagged data.

As discussed above, in some examples, a video encoding device (such as video encoder 20 or video decoder 30) may be configured to determine whether a threshold is present in a block. For example, the video encoding device may be configured to determine whether a threshold is present in block 106 based on a pixel analysis of one or more neighboring blocks, wherein the neighboring block may include one or more neighboring pixels 100. From a As a general rule, a neighboring, previously coded block may share a border with block 106, where the border may correspond to one or more neighboring pixels 100. For example, a neighboring block on the left for block 106 may include neighboring pixels 100I-100P, which define a boundary between the adjacent block on the left and block 106. The video encoding device may be configured to calculate gradients for pixels in a neighboring, previously encoded block to determine if a boundary is present in neighbors, previously encoded block. The video coding device may further determine whether the border crosses (i.e., intersects) a border between the neighbors, previously encoded block and a current block, such as block 106. Referring to the examples of the neighboring block on the left for block 106 described above, the video coding device may determine whether gradients for pixels in the neighboring block on the left indicate the presence of a border intersecting the boundary between the neighboring block on the left and block 106, where the border is defined by pixels 100I-100P , in this example. When the video coding device determines that pixel gradients in the neighboring block on the left indicate the presence of a boundary and that the border crosses the boundary defined by the pixels 100I-100P, the video encoding device may determine that the boundary proceeds to the block 106, and as such, that block 106 includes a limit.

In some examples, when the video encoding device determines that an in-predicted signal mode for block 106 is DC predictive mode, and that block 106 includes a threshold, the video encoding device may predict block 106 by using a mode intra-forecast based on the limit. In addition, the video encoding device may determine a threshold angle. The video coding device may then determine an angle of a prediction mode, generally indicated by the arrows 102, which most closely approximate the angle of the boundary. The video encoding device may then select a transformation (which may correspond to an inverse transformation upon decoding) which is mapped to the intra prediction mode having the angle that most closely approximates the angle of the boundary, and apply a selected transformation to data During a coding process, the video encoding device may apply a transformation to a TU of block 106, during a decoding process, the video encoding device may apply a reverse transformation to transformed residual data for the block 106. FIG. 5 is a conceptual diagram illustrating intra-predictive modes 110A-110I (intra-predictive modes 110) of H.264. Intra-prediction mode 110C corresponds to an intra-predictive mode DC, and therefore is not necessarily associated with a current angle. The remaining intra predictive modes 110 may be associated with an angle, similar to the angles of the arrows 102 of FIG. 4. For example, the angle of the intra prediction mode 110A corresponds to the arrow 102Y, the angle of the intra prediction mode 110B corresponds to the arrow 1021, the angle of the intra prediction mode 110D corresponds to the arrow 102AG, the angle of the intra- prediction 110E corresponds to arrow 102Q, angle of intra-prediction mode 110F corresponds to arrow 102U, angle of intra-prediction mode 110G corresponds to arrow 102M, angle of intra-prediction mode 110H corresponds to arrow 102AD, and angle of intra predictive mode 1101 corresponds to arrow 102E. Angles of the arrows 102 that do not correspond directly to one of the intra-prediction modes 110 may be mapped to one of the intra-prediction modes 110. For example, the angle to one of the intra-prediction modes 110 that approximates the angle of one of the arrows 102 can correspond to the angle so that one of the arrows 102 is mapped.

Each of the intra prediction modes 110 may be mapped to a specific transformation, eg, with a one-to-one correspondence. For example, a video encoding device, such as video encoder 20 or video decoder 30, may include configuration data that maps intra-prediction mode 110C to a DCT, and each of the other intra-prediction modes 110 to one specific directional transformation, eg, a KLT. Accordingly, angles for each of the intra-prediction modes associated with the arrows 102 (FIG.4) can be mapped to intra-forecast mode angles 110 (FIG.5). Intra-prediction modes 110 can, in turn, be mapped to transformations, eg, directional transformations. In this manner, angles for each of the intra-prediction modes associated with arrows 102 (FIG.4) can be mapped to directional transformations. Accordingly, video encoder 20 and video decoder 30 may determine a directional transformation to apply to a TU based on an intraprediction mode selected for a PU corresponding to TU. FIG. 6 is a conceptual diagram illustrating an example of a zig-zag search of coefficients 120A-120P (coefficients 20). Coefficients 120 generally correspond to quantified transformation coefficients, resulting from the transformation and quantification of the pixels or a TU. Video encoder 20 may be configured to search a block of coefficients using the zig-zag search of FIG. 6 below, eg, patent application of a DCT for a residual block. In this example, the zig-zag query starts at coefficients 120A, then proceeds to coefficients 120B, then to coefficients 120E, then to coefficients 1201, then to coefficients 120F, then to coefficients 120C, then to coefficients 120D, then to coefficients 120G, then for coefficients 120J, then for coefficients 120M, then for coefficients 120N, then for coefficients 120K, then for coefficients 120H, then for coefficients 120L, then for coefficients 120L, then for coefficients 120P.

By performing this search, the arrangement of the two dimensions of the coefficients for pixels may be converted into a one-dimensional series including values for each of the coefficients 120. These values may be arranged in the series in the order of the search. For example, the value for coefficients 120A may be the first in the series, followed by the values for coefficients 120B, 120E, 1201, 120F, and so on.

Other predefined search models can also be set for other transformations. For example, each directional transformation can be associated with a search model that is designed to place low frequency coefficients resulting from the directional transformation at the beginning of the table as the coefficients of the highest frequency. One of the directional transformations can cause low frequency coefficients to occur along the leftmost column of a block of transformation coefficients, in which case a corresponding search can be defined that starts at the coefficients 120A, then proceeds to coefficients 120E, then to coefficients 1201, then to coefficients 120M, then to 120B, and so on. A further example, another of the directional transformations may cause low frequency coefficients to occur along the top line of a block of transform coefficients, in which case a corresponding search can be defined that starts at the coefficients 120A, then proceeds to coefficients 120B, then to coefficients 120C, then to coefficients 120D, then to coefficients 120E, and so on.

In some examples, video encoder 20 may be configured to perform an adaptive search, rather than a predefined search. Adaptive research can vary over time on the basis of statistical indicatives if specific coefficients (ie, coefficients corresponding to coefficients 120) are significant. In addition, video encoder 20 may calculate sets of statistics on a stand-alone basis, for example, in an intra-prediction mode selected to predict a block, an index of a rotation transformation to apply following an initial transformation, or other factors .

In some examples, video encoder 20 may include two tables for these statistics: a high dynamic range table and a dynamic update lookup table. Assuming that the block to be searched has NxN coefficients, each of these two tables can also have NxN size. The high dynamic range table can be a fixed, pre-determined table providing probabilities that each transformation coefficient is nonzero. This table can be calculated based on a set of training data. In addition, this table can be used to provide the starting point for the order of the adaptive search. In general, the high dynamic range table can be static (that is, unchanged) for a bit stream. The dynamic update lookup table can be updated over time to reflect recently determined statistics for transformation coefficients. In particular, video encoder 20 may maintain an account of the number of times each coefficient is different from zero. That is, for each transformation block, video encoder 20 can increment values in the dynamic update lookup table corresponding to nonzero coefficients in the current transform block. For example, if a transform coefficient corresponding to coefficient 120E is nonzero, video encoder 20 may add one to the value in the dynamic update lookup table corresponding to coefficient 120E. Values in the dynamic update lookup table can also be normalized periodically to avoid values exceeding a maximum value.

To perform the adaptive search for a first frame processing unit, video encoder 20 can search based on the high dynamic range table alone. Video Encoder 20 can also initialize the dynamic update lookup table by, eg, set all values in the dynamic update lookup table to zero. For each non-zero coefficient in a transform block, video encoder 20 may add one to the value placed in the dynamic update lookup table. For the following blocks, video encoder 20 may first refer to the dynamic update lookup table to determine that the transform coefficients are most likely to be nonzero, then search in order to decrease a probability of the coefficients being different from zero . In some cases, two or more values in the dynamic update lookup table may be the same. In such a case, quantizing unit 54 refers to the high dynamic range table to determine which coefficients to search next. In this way, quantizing unit 54 can perform an adaptive search based on a combination of a high dynamic range table and a dynamic update lookup table. The high dynamic range table can be the same for all adaptive search statistics, in some examples. Thus, video encoder 20 may include specific dynamic update lookup tables for, for example, the selected intra-prediction mode, a rotation transformation index, or a combination thereof. In some examples, video encoder 20 may be configured to select from among the predetermined, static searches when a rotation transformation is not applied, and to perform an adaptive search when a rotation transformation is applied, and further, to select statistics to perform the adaptive search based on one or both of the selected predictive mode and the index of the selected rotation transformation. Alternatively, video encoder 20 may be configured to select a predefined search based on an index of the rotation transformation, an intra prediction mode, or a combination thereof. Video decoder 30 may be similarly configured for video encoder 20 to select an appropriate search. FIGS. 7A and 7B are conceptual diagrams illustrating a quadtree example 150 and a corresponding larger encoding unit 172. FIG. 7A describes a quadtree example 150, which includes nodes arranged in a hierarchical form. Each node in a quadtree, such as quadtree 150, can be a leaf without children, or have four child nodes. In the example of FIG. 7A, quadtree 150 includes root 152. Root 152 has four child nodes, including sheets 156A-156C (sheets 156) and node 154s. Because node 154 is not a sheet, node 154 includes four child nodes, which in this example are sheets 158A-158D (sheets 158). QUadtree 150 may include data describing features of a corresponding larger coding unit (LCU), such as LCU 172 in this example. For example, quadtree 150, by its structure, can describe divisions of the LCU into sub-CUs. Assume that LCU 172 has a size of 2Nx2N. LCU 172, in this example, has four sub-CUs 176A-176C (sub-CU 176) and 174, each of size NxN.

Sub-CU 174 is further divided into four sub-CUs 178A-178D (sub-CU 178), each of size N / 2xN / 2. The quadtree structure 150 corresponds to the division of LCU 172, in this example. That is, root 152 corresponds to LCU 172, leaves 156 correspond to sub-CU 176, node 154 corresponds to sub-CU 174, and leaves 158 correspond to sub-CU 178.

Data for quadtree nodes 150 can describe if the CU corresponding to the node is divided. If the CU is divided, four additional nodes may be present in quadtree 150. In some examples, a quadtree node can be implemented similar to the following pseudo-code: quadtree_node {boolean split_flag (1); // signaling data if (split_flag) {quadtree node childl; quadtree_node child2; quadtree node child3; quadtree_node child4:}} The value of split_úlag can be a value representative of a bit if the CU corresponding to the current node is divided. If the CU is not split, the value of split_ulag can be '0', while if the CU is split, the value of split_blag can be '1'. Relative to quadtree example 150, a series of division flag values may be 101000000.

In some examples, each sub-CU 176 and sub-CU 178 may be intra-predicted coding using the same intra-predictive mode. Accordingly, video encoder 20 may provide an indication of the intra-predictive mode at root 152. In addition, certain sizes of the sub-CUs may have multiple transformations possible for a specific intra-prediction mode. According to the techniques of this disclosure, video encoder 20 may provide an indication of the transformation to be used for such sub-CUs at root 152. For example, N / 2xN / 2 size sub-CUs may have multiple possible transformations available. Video encoder 20 may signal the transformation to use at root 152. Consequently, video decoder 30 may determine the transformation to apply to sub-CU 178 based on the intra predicted mode signaled at root 152 and the signaling transformation at root 152.

As such, video encoder 20 need not signal transforms to apply sub-CU 176 and sub-CU 178 to sheets 156 and sheets 158, but instead may simply signal an intra-prediction mode and, in some examples, a transformation to apply to certain sub-CUs sizes, at root 152, according to the techniques of this disclosure. In this way, these techniques can reduce the overhead of signaling transformation functions for each sub-CU of an LCU, such as LCU 172.

In some examples, intra-predictive modes for sub-CU 176 and / or sub-CU 178 may be different from intra-prediction modes for LCUs. Video encoder 20 and video decoder 30 may be configured with functions mapping a mode predicted intra-prediction mode at root 152 for an intra-prediction mode available for sub-CU 176 and / or sub-CU 178. The function may provide a many-to-one mapping of the intra-prediction modes available for LCU 172 for modes of intra-prediction for sub-CU 176 and / or sub-CU 178.

While FIG. 7A illustrates an example of a quadtree CU, it should be understood that a similar quadtree

can be applied to TUs of a CU sheet. That is, a CU sheet may include a TU quadtree that describes division of TUs into CU. ATU's Quadtree can generally resemble a CU quadtree, except that TU quadtree can signal intra-forecast modes for CU TUs individually.

Fig. 8 is a flowchart illustrating a sample method for selecting a transformation and a search for applying to a block based on a selected intra-forecast mode for the block. While generally being described as embodied by components of the video encoder 20 (FIG. 2) for purposes of explanation, it should be understood that other video encoding units, such as processors, processing units, base encoding units in hardware such as encoder / decoders (CODECs), and the like, may also be configured to perform the method of FIG. 8.

In the example of FIG. 8, transformation unit 52 may initially receive residual data for a current TU 180. Additionally, processing unit 52 may also receive an indication of an intra-forecast mode selected for TU. From this indication, transformation unit 52 may determine a prediction direction of TU (182). For example, transformation unit 52 may determine an angle of the prediction direction for the indicated intra-forecast mode.

In any case, after determining the intra-prediction mode, transformation unit 52 may select a transformation to apply to the residual data based on an intra-prediction mode mapping for the transformation (186). For example, transformation unit 52 may select the transformation to apply by querying mapping data 66 with an intraprediction direction and determining the transformation so that the intraprediction direction is mapped. The transformation may correspond to a discrete cosine transform or directional transformations, such as a directional transformation dependent mode (MDDT). Processing unit 52 can then apply the selected transformation to the residual data to transform the residual data (188). In some examples, mapping data 66 may additionally include an indication that transformation unit 52 must apply two or more transformations, such as a rotation transformation following the first transformation, in which case transformation unit 52 may further apply the indicated rotation transformation .

By transforming the residual data, transformation unit 52 can produce a two-dimensional array of transform coefficients having the same number of coefficients as the residual data. In the example of FIG. 8, quantization unit 54 can then quantify transformation coefficients (190). In some examples, quantizing unit 54 may search the two-dimensional array of coefficients to produce for a one-dimensional array, eg, before or after quantifying the coefficients. Alternatively, entropy coding unit 56 may search the two-dimensional array.

In this example, entropy coding unit 56 may query mapping data 66 to select a search to apply to the quantized transformation coefficients (192). In some examples, mapping data 66 may include data that maps intra-forecasting modes to specific and predefined search models. In some examples, mapping data 66 may include data that maps transformations to predefined search models. In some examples, eg, where mapping data 66 indicates that a rotation transformation is to be applied to transformation coefficients, mapping data 66 may additionally indicate that an adaptive search must be performed, or a predefined search so that the transformation of rotation is mapped. In examples for which an adaptive search is performed, mapping data 66 may additionally include search statistics, eg, a high dynamic range table and a dynamic update lookup table, mapped to the intra predictive mode, an index of the first transformations, an index of the rotation transformation, a combination thereof, and / or other factors.

The entropy coding unit 56 can then search the quantized transformation coefficients using the selected search (194), eg, the default search or the adaptive search based on selected search statistics. In some examples, entropy coding unit 56 may be configured with a search position (which may be less than or equal to the number of transform coefficients) after which the entropy coding unit 56 can zero the values of the coefficients on the Serie. After searching a number of coefficients equal to the search position, entropy coding unit 56 can set the remaining values of the series equal to zero. Zeroing transformation coefficients can occur before or after the search, in some examples.

In some examples, entropy coding unit 56 may then encode entropy coefficients in the series surveyed following the search (196). Alternatively, in some examples, entropy coding unit 56 may encode the entropy coefficients as they are searched. In any case, entropy coding unit 56 may use either CABAC or CAVLC to encode the coefficients entropy.

When using CABAC, and when performing an adaptive search, entropy coding unit 56 may encode entropy syntax elements including significant coefficient flag and last coefficient flag. Entropy coding unit 56 can select contextual model to encode flag entropy of significant coefficients based on a block type (intra or inter), a selected intra-prediction mode (assuming the block is predicted in an intra-mode), and / or a type of applied transformation (eg, DCT or directional / KLT). Entropy coding unit 56 can select the context model to encode the entropy flag of the last coefficient based on index order of adaptive search, block type, spatial prediction direction, and / or a selected transformation.

In this manner, the method of FIG. 8 depicts an example of a method including calculating a residual block for a block of video data based on a predicted block formed using an intra prediction mode, and transforming the residual block using a mapped transformation into the intra prediction mode. FIG. 9 is a flowchart illustrating another example method for selecting a transformation and search to apply to a residual data block. In general, FIG. 9 substantially conforms to FIG. 8. However, in the example of FIG. 9, after receiving the residual data 180 and indicating a selected intra-forecast mode for TU, transformation unit 52 may determine a first prediction direction to predict TU (183). For example, transformation unit 52 may determine an angle of the prediction direction for the indicated intra-forecast mode.

The transformation unit 52 may then determine a second direction mapped to the first prediction direction (184). For example, transformation unit 52 may query mapping data 66 to determine a second intra-prediction mode for the first intra-forecast mode to be mapped. In some examples, processing unit 52 may determine an angle approaching the angle of the indicated intra-prediction mode, and select a second intra-prediction mode corresponding to the given angle. The transformation unit 52 may then select a mapped transformation for the second prediction data (185). After selecting the transformation, which may correspond to select multiple transformations, video encoder 20 generally performs the remaining steps of FIG. 9 in a similar manner to the corresponding steps described with respect to FIG. 8.

In this manner, the method of FIG. 9 is an example of a method which comprises receiving an indication of a first intra-prediction mode in the first set of intra-prediction modes for a video data block, determining a second intra-prediction mode from a smaller set of intra- prediction so that the first intra-forecast mode is mapped, determine a directional transformation so that the second intra-forecast mode is mapped, and apply the directional transformation to the residual data of the block. FIG. 10 is a flow chart illustrating a sample method for applying an intra-forecast mode and transforming into special-size sub-CUs. Although generally described as embodied by components of the video encoder 20 (FIG. 2) for purposes of explanation, it is to be understood that other video encoding units, such as processors, processing units, base encoding units in FIG. hardware such as encoder / decoder (CODEC), and the like, may also be configured to perform the method of FIG. It should also be understood that in other examples, similar methods may include additional or alternative steps to those shown in FIG. 10, or may perform the illustrated steps in a different order, without departing from the techniques described. Techniques for selecting and applying a transformation as described with respect to FIG. 10 may correspond to steps 186 and 188 of FIG. 8. Techniques for applying various intra-prediction modes for blocks of various sizes as described with respect to FIG. 10 may be performed prior to step 180 of FIG. 8.

Intra-prediction unit 46 may receive a block of pixels, eg, an LCU (200). Intra-prediction unit 46 may then determine an intra-predictive mode to apply to the LCU and signal the particular intra-prediction mode to the LCU 201, eg, to a root of a quadtree data structure corresponding to the LCU. Intra-prediction unit 46 may then determine sub-CUs sizes so that only a subset of intra-prediction modes are available (202), intra-prediction unit 46 may additionally divide LCUs into one or more sub-CUs and determine if any of the sub- sub-CUs have a size so that only a subset of intra-forecasting modes are available (204).

If the LCU includes sub-CUs of a size so that only a subset of intra-prediction modes is available ("YES" branch 184), intra-prediction unit 46 may intra-predict sub-CUs using an intra- the intra-prediction mode selected for an LCU is mapped (206). On the other hand, if the LCU does not include any sub-CUs that are of such a size ("NO" branch 184), intra prediction unit 46 may apply the signaled mode to the LCU for all sub-blocks of the LCU 208, .

Video encoder 20 can then calculate residual values for sub-CUs of the LCU. Thereafter, processing unit 32 may determine sub-sizes of the CU so that multiple transformations are possible based on the predicted intra-prediction mode for the LCU 210. The processing unit 52 may further determine if any of the LCU sub-CUs are of a size so that multiple transformations are possible (212). If at least one sub-CU is of a size for multiple transformations to be possible ("YES" branch 212), transformation unit 52 may select and signal a transform to apply to sub-CUs of that size (214). For example, transformation unit 52 may signal the transformation to apply to sub-CUs of that size in the root of quadtree to the LCU. Processing unit 52 can also apply the signaling transformation to all sub-blocks in the LCU of that size (216). On the other hand, if the LCU does not contain any sub-CUs for which multiple transformations are possible ("NO" branch 212), transformation unit 52 may apply transformations to LCU sub-CUs purely in the intra-prediction mode flagged for the LCU (218) in such a way that no signaling on transformations is required.

In this manner, the method of FIG. 10 is an example of a method including selecting an intra-predictive mode to be used to encode a video data block, determining whether the block includes a sub-block of a size so that multiple transformations are possible based on the size of the sub- block and the selected intra-prediction mode, when the block includes the size sub-block so that multiple transformations are possible based on the sub-block size and the selected intra-forecast mode select one of the multiple transformations possible, transform the sub -block using the selected of the multiple possible transformations, and provide an indication of the selected of the multiple possible transformations for the block size. FIG. 11 is a flowchart illustrating a sample method for performing an adaptive search based on a selected transformation applied to a block. Although generally described as embodied by components of video encoder 20 (FIG. 2) for purposes of explanation, it is to be understood that other video encoding units, such as processors, processing units, base encoding units in FIG. hardware such as encoder / decoder (CODEC), and the like, may also be configured to perform the method of FIG. 11.

It is also to be understood that in other examples, similar methods may include additional or alternative steps to those shown in FIG. 11, or may perform the illustrated steps in a different order, without departing from the techniques described. Techniques for adaptive research coefficients following a cascade transformation as shown in FIG. 11 may correspond to steps 192 and 194 of FIG. 8. The techniques for residual adaptive research coefficients of FIG. 11 can be applied to residual data following intra-forecast or inter-forecasting.

Initially, processing unit 52 of the video encoder 20 may receive a residual block 230. The residual block may correspond to residuals following intra-forecast or inter-forecasting of a CU. The residual block may be of the same size or of a different size than a corresponding prediction unit of CU. The processing unit 52 can then transform the residual block 232. In some examples, transformation unit 52 may apply a directional transformation corresponding to an intra prediction mode from a subset of the intra prediction modes, according to techniques of this disclosure. In other examples, processing unit 52 may apply a discrete cosine transform.

The processing unit 52 may then apply a rotation transformation to the transformed block 234. For transformation units (TUs) of sizes 8x8 and larger than the processing unit 52 a DCT applies, processing unit 52 may apply the rotation transformation to the DCT DCT coefficients 8x8 lower frequency. For TUs smaller than 8x8, transformation unit 52 can apply the rotation transformation to the entire TU. If the PU corresponding to TU was intra-predicted coding, transformation unit 52 may select a rotation transformation based on the intra-prediction mode used to predict PU, eg, where a subset of the intra-prediction modes can be mapped to transformations of rotation. These rotational transformations can be referred to as mode-dependent rotation transformations (MDROTs). In some examples, transformation unit 52 may cascade a rotation transformation (or other separate secondary transformation) following a directional transformation dependent mode (MDDT), which may be a KLT.

Following the rotation transformation, quantitation unit 54 can quantify the transformed coefficients in some examples. Next, entropy coding unit 56 may select a set of statistics to use to perform an adaptive search of the transformation coefficients. The statistics set may include a dynamic long-range table (HDR) and a dynamic update table (DU). One or both of the HDR and DU tables can be selected for a specific scenario, eg, if intra-forecast or inter-forecast is used to predict a PU, a specific intra-forecast mode for PU when intra-forecast is used, if a DCT or KLT was applied to the TU corresponding to the PU, the index of the rotation transformation used, or any combination thereof. In this way, entropy coding unit 56 may select the HDR and / or DU tables to use during adaptive searching (236).

As discussed earlier, the HDR table may include a set of predefined data indicating probabilities that coefficients of each location in an array are nonzero. The HDR table can be produced using a set of training data, and can continue the same in every bit stream. Entropy coding unit 56 may collect individual statistics for a frame, part, group of images, or other video data unit to calculate values for the DU table. The DU table may therefore also indicate possibilities that coefficients of each location in the array are nonzero.

To perform the adaptive search, entropy coding unit 56 can first determine the location in the array having the highest probability of

coefficient other than zero using the DU table (238). In some cases, there may be two or more locations in the matrix with equal probabilities of having coefficients other than zero. Thus, entropy coding unit 56 can determine if there are multiple locations in the array with the same probability of the coefficients other than zero (240) included. If there are multiple locations in the array with the same probabilities as non-zero coefficients ("YES" branch 240) included, entropy coding unit 56 can determine the location in the array having the highest probability of including a nonzero coefficient using the table HDR (242).

The entropy coding unit 56 can then search and encode the entropy coefficients at the given location (244). The entropy coding unit 56 can also determine whether the coefficient searched was in fact non-zero and set the value for a significant coefficient flag to indicate whether the coefficient searched was nonzero, and therefore significant. The entropy coding unit 56 can then determine whether all coefficients in the array were searched (246). If not ("NO" branch 246), entropy coding unit 56 can determine the location in the array having the next highest probability of

include a non-zero coefficient using the DU table (or possibly the HDR table), and look up the coefficients at this location.

In addition, in some examples, entropy coding unit 56 can define the value of a flag of the last coefficient for each coefficient to indicate if the corresponding coefficient is the last coefficient in the flag.After determining that all coefficients have been searched ("YES" branch 246), entropy coding unit 56 can set the value for the flag of the last corresponding coefficient for the last coefficient searched by the same as one. Using the techniques of Figure 12 as described below, entropy coding unit 56 may encode syntax elements including the flags of significant coefficient and flags of the last coefficient.

In some examples, following the search (whether adaptive or fixed), video encoder 20 may zero out coefficients in the series produced by the search, eg, all coefficients after the N position in the series, where N is an integer between zero and the length from the series. In other examples, video encoder 20 may zero the coefficient at certain locations of the array following the transformation (s) or quantification. These locations may correspond to the upper left corner of the array, for example. Generally, zeroing these coefficients can result in zeroing the high frequency coefficients, which can improve coding efficiency without much impact on quality.

In this manner, the method of FIG. 11 is an example of a method including transforming residual data to a video data block using a first transformation to produce an intermediate, two-dimensional block of transform coefficients, transforming the intermediate, two-dimensional block of transform coefficients using a rotation transformation to produce a two-dimensional block of transformation coefficients by selecting a set of statistics associated with at least one of the first transformations and the rotation transformation, wherein the set of statistics provides possibilities that locations in the two-dimensional block of transformation coefficients will be nonzero, and will adaptively search the two-dimensional block of transformation coefficients based on the selected set of statistics. FIG. 12 is a flowchart illustrating a sample method for selecting a context model to use when searching and entropy encoded syntax elements describing adaptively searched coefficients.

Although for the purpose of explanation, it is to be understood that other video encoding units, such as processors, processing units, base encoding units, hardware such as encoder / decoder (CODEC), and the like, may also be configured to perform the method of FIG. 12.

It is also to be understood that in other examples, similar methods may include additional or alternative steps to those shown in FIG. 12, or may perform the illustrated steps in a different order without departing from the techniques described. Techniques for selecting a context model to use when searching and entropy encoded syntax elements describing coefficients researched adaptively as shown in FIG. 11 may correspond to steps 192-196 of FIG. 8. The techniques of FIG. 12 can be performed before, during, or after the adaptive search of FIG. 11 to be performed.

Entropy coding unit 56 may receive an array of quantized transformation coefficients (250), eg, from quantitation unit 54. Generally, using the example method of FIG. 12, entropy coding unit 56 may encode syntax elements describing received coefficients. The syntax elements may include, for each coefficient, a flag of the significant coefficient and a flag of the last coefficient. The flag of the significant coefficient can indicate whether the corresponding coefficient is significant, eg, if the corresponding coefficient value is greater than zero. The flag of the last coefficient can indicate whether the corresponding coefficient is the last coefficient of an adaptive search.

Entropy coding unit 56 can determine positions of the significant coefficients in the received matrix. Entropy coding unit 56 may form syntax elements indicating positions of the significant coefficients in the received matrix (252). For example, for each coefficient in the array, entropy coding unit 56 can determine whether the coefficient is greater than zero, in which case it defines a value in a matrix of the syntax element placed with the coefficient equal to one, otherwise unit of entropy coding can set the value placed with the coefficient equal to zero. Entropy coding unit 56 can then update a dynamic update lookup table using the syntax element array (254). For example, entropy coding unit 56 may add the value of the syntax element placed in the array of the syntax element to the current value of each element in the dynamic update lookup table.

The entropy coding unit 56 may then search the first of the syntax elements in the array of the syntax element (256). Entropy coding unit 56 may apply a zig-zag search, such as that shown in FIG. 6A, or a selected search based on a block type (inter- or intra-forecast block), a spatial prediction direction whether the block is an intra-predicted coding block, and / or a type of transformation used (eg, DCT or directional transformation). Then, entropy coding unit 56 may select a context model for encoding the searched syntax element (258). In general, the context model can be selected based on the number of significant (eg, non-zero) coefficients in the previous N coefficient survey, where N is an integer value other than zero. N can be selected based on block size.

After selecting the context model to use to encode the current syntax element, entropy encoding unit 56 may encode the searched syntax element using the selected context template (260). The entropy coding unit 56 may then determine if the encoded syntax element is the last syntax element to be encoded (262). If the syntax element is the last syntax element ("SIM" branch 262), entropy coding unit 56 may stop the search for coefficients. On the other hand, if the syntax element is not the last syntax element ("NO" branch 262), entropy coding unit 56 may search the next syntax element (264), and re-select a context model for encode the searched syntax element, eg, based on a number of significant coefficients in the N coefficients previously searched. The example of FIG. 12 is primarily discussed in relation to the syntax elements describing whether specific coefficients are significant or not. These syntax elements may include, for example, flag of significant coefficients, eg, one-bit flags indicative if corresponding coefficients are significant, eg, non-zero. It should be understood that similar techniques can be applied with respect to syntax elements by describing whether a specific coefficient is the last coefficient in adaptive research. For example, similar techniques may be applied to a flag of the last coefficient. When coding the flag of the last coefficient using CABAC, the context model can be based on the order index in the adaptive search which is based on the block type, spatial forecasting direction, and / or a selected transformation.

The techniques of FIG. 12 may be performed by a video encoding device such as video encoder 20. A video decoder may perform a reverse lookup using the syntax elements encoded according to FIG. 12. For example, video decoder 30 may receive a indication of an intra-prediction mode used to provide a coded block, an indication of a rotation transformation used to transform the coded block, or other data. Video encoder 20 and video decoder 30 can each be configured with the same dynamic long-range table. In the examples where video encoder 20 includes multiple dynamic long-range tables, video encoder 20 and video decoder 30 can each be configured with the same set of dynamic long-range tables. In such examples, video decoder 30 may use received information to select the same dynamic long-range table used by video encoder 20 to perform adaptive search.

As noted previously, video encoder 20 may perform adaptive search based on statistics indicative of the possibility (or probability) that a coefficient at a particular position in a matrix is nonzero. Video Encoder 20 can maintain a dynamic update lookup table that indicates this possibility by updating the dynamic update lookup table for each searched block. By coding syntax elements indicative of which coefficients of a certain block are significant, and which coefficient is the latter in the adaptive search, video encoder 20 may provide video decoder 30 with information that can be used for received reverse lookup coefficients.

For example, video decoder 30 can decode the syntax elements, then update a local version of the dynamic update lookup table using the syntax elements. Video decoder 30 can then decode entropy encoded coefficients and place the decoded coefficients in a corresponding position of a matrix having a next greater probability of being significant (eg, non-zero). In this way, video decoder 30 can reconstruct an array of quantized transformation coefficients of an received vector of entropy-encoded coefficients using an adaptive inverse search.

In this manner, the method of FIG. 12 is an example of a method including searching a two-dimensional block of transformation coefficients to produce a one-dimensional vector of transformation coefficients, determining values indicative if the transformation coefficients in the one-dimensional vector are significant, and encoding entropy at least one of the values using a selected context model based on at least a percentage of significant coefficients in a predetermined number of the encoded values before at least one of the values. FIG. 13 is a flowchart illustrating a sample method for decoding a TU that has been encoded using one or more of the techniques of this disclosure. Although generally described as embodied by the components of the video decoder 30 (FIG.3) for purposes of explanation, it is to be understood that other video decoding units, such as processors, processing units, hardware such as encoder / decoder (CODEC), and the like, may also be configured to perform the method of FIG. It is also to be understood that in other examples, similar methods may include additional or alternative steps to those shown in FIG. 13, or may carry out the illustrated steps in a different order, without departing from the techniques described.

Initially, video decoder 30 may receive encoded residual data (300). In the example of FIG. 13, the residual data correspond to the residual of a CU including one or more PUs predicted in an intra prediction mode, for purposes of illustration. According to the techniques of this disclosure, video decoder 30 may determine a first prediction direction for prediction data associated with the received residual data (302). The forecast direction may correspond to an intra-forecast mode signaled at the root of a quadtree corresponding to CU.

Video decoder 30 may determine a second prediction direction mapped to the first prediction direction (304). For example, mapping data 84 may provide a many-to-one mapping of a set of intra-predicate modes to a smaller set, eg, a subset, of intra-predictive modes. Consequently, video decoder 30 may refer to mapping data 84 to determine the second prediction direction mapped to the first prediction direction. The entropy decoding unit 70 of the video decoder 30 may then begin to decode entropy the received coefficients (306).

Entropy decoding unit 70 may also reverse search of the coefficients during or after decoding entropy (308). In some examples, entropy decoding unit 70 may reverse a fixed search so that the second prediction direction is mapped, eg, as indicated by the mapping data 84. In other examples, eg, when a first transformation is cascaded by a transformation of rotation, entropy decoding unit 70 may invert a dynamic search. As discussed previously, entropy decoding unit 70 may in such examples receive and decode syntax elements, such as flags of the significant coefficient and flags of the last coefficient, such that entropy decoding unit 70 can produce a dynamic update table identical to that used by an encoder, such as video encoder 20, when the encoder adaptively scans the residual data.

However, following the reverse lookup, entropy decoding unit 70 can produce a two-dimensional array including quantized transformation coefficients. Thus, inverse quantizing unit 76 may invert quantify the quantized transformation coefficients of matrix (310). Inverse transformation unit 78 may select a reverse transformation mapped to the second prediction direction (312) and reverse transformation of the transformation coefficients using the selected reverse transformation (314). For example, inverse transform unit 76 may refer to mapping data 84 to select the inverse transform. In some examples, mapping data 84 may both indicate a reverse rotation transformation and another inverse transformation to apply, in the case of reverse transformation unit 78 being able to first apply the reverse rotation transformation and then apply the other inverse transformations to the coefficients of transformation.

According to the techniques of this disclosure, in some examples, there may be a sub-block of transformation coefficients for which multiple reverse transformations are possible. In such examples, video decoder 30 may determine the reverse transform to apply using an indication of the transformation applied by the video encoder. For example, video decoder 30 may receive an indication of the transformation used for the sub-block in the root of a corresponding quadtree for the block including the residual data.

After applying the inverse transform (s), video decoder 30 obtains residual blocks similar to those calculated by the video encoder 20 during the encoding of the video data. Intra-prediction unit 74 may provide a residual data prediction unit for adder 80, which may combine the prediction unit and the residual data to produce a CU decoder (316). Video decoder 30 may join a decoded frame including CU

decoded in the store of the reference frame 82. The decoded frame may then be processed to view and / or used for reference when decoding other frames.

In this manner, the method of FIG. 13 is an example of a method including receiving an indication of a first intra-predictive mode in a first set of intra-predictive modes for an encoded video data block, determining a second intra-prediction mode of a smaller set of intra- prediction so that the first intra prediction mode is mapped, determine a reverse directional transformation so that the second intra prediction mode is mapped, and apply the inverse directional transformation to transformed residual data of the block. The method of FIG. 13 is also an example of a method including receiving a first indication of an intra-predictive mode to use for decoding a video data block, determining whether the block includes a sub-block of a size so that multiple transformations are possible based in the size of the sub-block and the indicated intra-prediction mode, when the block includes the size sub-block so that multiple inverse transformations are possible based on the size of the sub-block and the indicated intra-forecast mode, receive a second one of the many possible inverse transformations, and to reverse the transformation of the sub-block using the one of the multiple inverse transformations possible. The method of FIG. 13 is also an example of a method including receiving an indication that residual data for a video data block has been transformed using both a first transformation of a spin transformation to produce a two-dimensional block of transformation coefficients, selecting a set of statistics associated with at least one of the first transformations and the rotation transformation, wherein the set of statistics provides possibilities that locations in the two-dimensional block of transform coefficients will be different from zero, and adaptively inverts the search for a received one-dimensional vector including a version of the residual data coded for the block based on the selected set of statistics to produce a two-dimensional array of transform coefficients for the block. The method of FIG. 13 is also an example of a method including receiving indicative values if transformation coefficients encoded in a one-dimensional vector received from encoded transformation coefficients are significant, decoding entropy at least one of the values using a selected context model based on at least a percentage of coefficients in a predetermined number of the decoded values before at least one of the values, and invert the search of the one-dimensional vector to produce the two-dimensional block of the transformation coefficients. FIG. 14 is a flow chart illustrating a sample method for selecting a transformation to apply to an intra-coded block including a threshold for the intra-prediction mode DC to be signaled. Although described with respect to the video decoder 30 of FIG. 3, it should be understood that similar (reciprocal) techniques may be applied by the video encoder 20 of FIG. 2, or other video encoding devices.

Video encoder 30 may receive an intra coded block, eg, a TU (180). The block may comprise a block of transform coefficients corresponding to a node in a quadtree of TU. The quadtree of the TU node may include an indication of the intra-forecast mode to be applied to calculate a prediction value for the block. Consequently, video decoder 30 can determine the prediction mode and whether the prediction mode DC is signaled to the block 352. If the DC prediction mode is signaled to the block ("SIM" branch 352), video decoder 30 may further determine if a threshold exists at block 354. For example, as discussed above, video decoder 30 may examine neighbors, previously encoded blocks to determine whether a threshold is detected in previously encoded blocks, and whether the edge intersects an edge between the encoded block thereafter and the current block.

If a threshold is determined to exist in the block ("YES" branch 354), video decoder 30 may calculate a predicted value for the block using a boundary-based predictive mode (356). In addition, video decoder 30 may determine a threshold angle (358) and determine an intra mode with an angle approaching the threshold angle (360). For example, video decoder 30 may calculate differences between angles for one or more of the possible intra-prediction modes and the angle of the threshold, and select the intra-prediction mode having the smallest difference.

Determining this mode of prediction is generally carried out only to determine the transformation that is mapped so that the prediction mode, however, as video decoder 30 generally predicts a value for the block using the predictive mode based in the limit, in this example. That is, video decoder 30 may then select a mapped transformation to determined intra-prediction mode (362), i.e. intra-prediction mode having an angle approaching the angle of the threshold. Video decoder 30 can then transform the block using a selected transformation (364).

On the other hand, if predictive mode DC was not flagged for block ("NO" branch 352), video decoder 30 can predict the block using flagged mode (353). If a threshold is not determined to exist in the block when DC predictive mode is flagged ("NO" branch 354), video decoder 30 can predict the block using DC predictive mode, as signaled (366). Video decoder 30 can also select the mapped transformation to the prediction mode (eg, DC or directional, as flagged) (368) or in some examples a standard transformation, such as a DCT. Video decoder 30 can also transform the block using a selected transformation in this case (364).

After transforming the block, in which in this example corresponds to the inverse transformation of the block, video decoder 30 reproduces a block of residual values in the spatial domains. To decode the block, video decoder 30 may add the residual value block to the predicted block (resulting from step 353, 356, or 366). The steps of adding the residual value to the predicted value are not shown in FIG. 14 for conciseness, but may be performed after step 364.

In this manner, the method of FIG. 14 is an example of a method including receiving information indicating that an intra-predictive mode for a video data block is an intra-predictive mode DC, determining an angle for a threshold in the video data block based on the mode indication of intra-DC prediction for the block, inverse block transformation using a directional reverse transformation mapped to a directional intra-prediction mode having an angle approaching the boundary angle, and decoding the inverse transformed block.

As noted above, a similar method may be performed by, eg, video encoder 20. Such a method may include determining that a block to be intra-predicted encoding contains a boundary within the block, calculates a residual block for the base block in a predicted value calculated using a directed intra-prediction mode boundary, transforming the residual block using a directional transformation mapped to a directional intra-prediction mode having an angle approaching a boundary angle, and sending information representative of the transformed residual block and information indicating that the block was predicted using an intra-DC prediction mode.

In one or more examples, the described functions may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored or transmitted in one, with one or more instructions or code, computer readable medium and executed by a hardware based processing unit. Computer readout communication means may include computer readout storage means, which corresponds to a material carrier such as data storage means, or communication means communication including any means facilitating the transfer of a computer program from one place to another, eg, conform to a communication protocol. In this manner, computational read communication medium may generally correspond to (1) communication means of storing computational read concrete that is non-transient or (2) a communication medium such as a signaling or carrier wave. Data storage medium may be any available means of communication that may be accessed by one or more computers or one or more processors to retrieve instructions, code and / or data structures for implementation as techniques described in this disclosure. A computer program product may include a computer readable medium. By way of example, and not limiting, such computer read storage medium may comprise RAM, ROM, EEPROM, CD-ROM or other magnetic storage devices, magnetic storage disks, or other optical disk storage, disk storage magnetic, or other magnetic storage device, flash memory, or any other means which may be used to store desired program code in the form of instructions or data structures and which may be accessed by a computer. Also, any connection should be appropriately called a computer readable medium. For example, if instructions are transmitted from a web site, server, or other remote sources using a coaxial cable, fiber optic cable, interlaced wire pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio , and micro-wave, then coaxial cable, fiber optic cable, interlaced wire, DSL, or wireless technologies such as infrared, radio, and micro-wave are included in the definition of the medium. It should be understood, however, that computer read storage medium and data storage medium do not include connections, carrier waves, signals, or other transient means, but are instead directed to non-transient, concrete storage medium. Discs and floppy disks, as used herein, include a compact disk (CD), a disk, an optical disc, a versatile digital disk (DVD), a floppy disk, and a Blu-ray disc, where floppy disks normally reproduce magnetic data, while discs reproduce optical data with lasers. Combinations of the above must also be included within the scope of the computational reading means.

Instructions may be performed by one or more processors, such as one or more digital signal processors (DSPs), generic purpose microprocessors, application specific integrated circuit (ASICs), programmed port network (FPGAs), or other equivalent integrated circuits or discrete logic.

Accordingly, the term "processor" as used herein may refer to any of the above-mentioned structure or any other structure suitable for implementation as techniques described herein. In addition, in other respects, the functionality described herein may be provided within dedicated hardware and / or software modules configured for coding and decoding, or incorporated in a codec combination. Also, the techniques can be fully implemented in one or more circuits or logic elements.

The techniques of this disclosure may be implemented in a variety of devices or apparatus, including a wireless handset, an integrated circuit (IC), or a set of ICs (eg, a chip set). Various components, modules, or units are described in this disclosure to emphasize functional aspects of the devices configured to perform the disclosed techniques but do not necessarily need to be performed by different hardware units. Instead, as described above, several units may be combined in a hardware codec unit or provided by an interleaved hardware unit collection, including one or more processors as described above, in conjunction with suitable software and / or firmware.

Lisbon, January 5, 2015

Claims

A method of decoding video data, the method comprising: receiving (201) a first indication of an intra-prediction mode to be used to decode a video data block; determining (212) whether the block includes a sub-block for which multiple transformations are possible, wherein determining is based on the sub-block size and the predicted intra-prediction mode; when the block includes the sub-block so that multiple reverse transformations are possible: receiving (214) a second indication of one of the multiple possible reverse transformations; and reverse transformation (216) of the sub-block using the indicated one of the multiple possible inverse transformations.

The method of claim 1, further comprising, when the block does not include the sub-block for multiple reverse transformations to be possible, inverse block transformation (218) using a reverse transformation associated with the intra-prediction mode indicated for block .

The method of claim 1, further comprising reverse transforming (216) all sub-blocks of the block having the size so that multiple inverse transformations are possible using the indicated of the multiple possible transformations.

The method of claim 1, wherein the intra-predictive mode comprises a first intra-prediction mode, and wherein the sub-block comprises a first sub-block, the method further comprising: determining (204) whether the block includes a second sub-block of a size so that the first intra-forecast mode is not available; when the block includes the second size sub-block for which the first intra-prediction mode is not available: determining (206) a second intra-forecast mode so that the first intra-forecast mode is mapped; and predicting (206) the second sub-block using the second intra-forecast mode.

The method of claim 4, further comprising, when the first intra-prediction mode is available for all sub-blocks of the block, predicting (208) all sub-blocks using the first intra-forecast mode.

The method of claim 1, wherein receiving the first indication comprises receiving the first indication in a root of a quadtree data structure corresponding to the video data block.

Ί. An apparatus for decoding video data, the apparatus comprising: means for receiving a first indication of an intra-predictive mode to be used for decoding a video data block; means for determining whether the block includes a sub-block for which multiple transformations are possible, wherein determining it is based on the size of the sub-block and the indicated intra-forecast mode; means for receiving a second indication of one of multiple possible reverse transformations when the block includes the sub-block so that multiple reverse transformations are possible; and means for inverse transformation of the sub-block using the indicated of the multiple inverse transformations possible when the block includes the sub-block so that multiple inverse transformations are possible.

The apparatus of claim 7, wherein the means for receiving a first indication, the means for determining the means for receiving a second indication, and the means for reverse transformation are incorporated in a video decoder.

A computer program product comprising a computer readable memory device having stored instructions thereon which, when executed, causes the processor to execute the method of any one of claims 1 to 6.

A method of encoding video data, the method comprising: selecting (201) an intra-predictive mode to be used to encode a video data block; determining (212) whether the block includes a sub-block for which multiple transformations are possible, wherein determining it is based on the size of the sub-block and the selected intra-forecast mode; when the block includes the sub-block so that multiple transformations are possible: select (214) one of the many possible transformations; transform (216) the sub-block using the one selected from the multiple possible transformations; and providing (214) an indication of the selected of the multiple possible transformations for the size of the sub-block.

The method of claim 10, wherein the intra-prediction mode comprises a first intra-prediction mode, and wherein the sub-block comprises a first sub-block, the method further comprising: determining (204) whether the block includes a second sub-block of a size for which the first intra-prediction mode is not available; when the block includes the second size sub-block for which the first intra-prediction mode is not available: determining (206) a second intra-forecast mode so that the first intra-forecast mode is mapped; and predicting (206) the second sub-block using the second intra-forecast mode.

The method of claim 10, wherein providing the indication comprises providing the indication at a root of a quadtree data structure corresponding to the video data block.

An apparatus for encoding video data, the apparatus comprising: means for selecting an intra-prediction mode to be used for encoding a video data block; means for determining whether the block includes a sub-block for which multiple transformations are possible, wherein determining it is based on the size of the sub-block and the selected intra-forecast mode; means for selecting one of the multiple transformations possible when the block includes the sub-block so that multiple transformations are possible; means for transforming the sub-block using the one of the multiple transformations possible when the block includes the sub-block so that multiple transformations are possible; and means for providing an indication of the selected of the multiple possible transformations for the size of the sub-block when the block includes the sub-block for multiple transformations to be possible.

The apparatus of claim 13, wherein the means for selecting an intra-prediction mode, the means for determining, the means for selecting one of the multiple possible transformations, the means for transforming, and the means for supplying are incorporated in an encoder of video.

A computer program product comprising a computer readable memory device having stored instructions thereon which, when executed, cause the processor to execute the method of any one of claims 10 to 12. Lisbon, January 5, 2015