WO2012044118A2 - Method and apparatus for encoding / decoding video using error compensation - Google Patents

Abstract

According to the present invention, a method for decoding a video in a skip mode comprises the steps of: deriving a pixel value of an estimation block for a current block; deriving an error compensation value for the current block; and deriving a pixel value of a final prediction block using the pixel value of the prediction block and the error compensation value. According to the present invention, the amount of transmitted information is minimized, and the efficiency of video encoding / decoding is improved.

Description

Method and apparatus for image coding / decoding using error compensation

The present invention relates to image processing, and more particularly, to an image encoding / decoding method and apparatus.

Recently, as broadcasting services having high definition (HD) resolution have been expanded not only in Korea but also in the world, many users are accustomed to high resolution and high quality images, and many organizations are accelerating the development of next generation video equipment. In addition, as interest in Ultra High Definition (UHD), which has four times the resolution of HDTV, is increasing along with HDTV, a compression technology for higher resolution and higher quality images is required.

Image compression techniques include inter prediction techniques for predicting pixel values included in a current picture from previous and / or subsequent pictures in time, and intra predicting pixel values included in a current picture using pixel information in the current picture. (intra) prediction technique, weight prediction technique to prevent deterioration of image quality due to lighting change, entropy encoding technique to assign a short code to a high frequency symbol and a long code to a low frequency symbol, etc. There is this. In particular, when prediction is performed on a current block in a skip mode, a prediction block is generated using only a value predicted from a previously coded region, and a separate motion information or residual signal is generated. It is not sent from the encoder to the decoder. Image data can be efficiently compressed by these image compression techniques.

An object of the present invention is to provide an image encoding method and apparatus capable of improving image encoding / decoding efficiency while minimizing the amount of transmitted information.

Another object of the present invention is to provide an image decoding method and apparatus capable of improving image encoding / decoding efficiency while minimizing the amount of transmitted information.

Another technical problem of the present invention is to provide a skipping mode prediction method and apparatus which can improve image encoding / decoding efficiency while minimizing the amount of transmitted information.

One embodiment of the present invention is a video decoding method. The method includes deriving a pixel value of a prediction block for a current block, deriving an error compensation value for the current block, and using the pixel value and the error compensation value of the prediction block. Deriving a pixel value of a final prediction block, wherein the error compensation value is a sample value of an error compensation block for compensating an error between the current block and a reference block, wherein the reference block is in a reference picture. As a block, the block includes predictor related information of the pixels in the current block, and the prediction for the current block is performed in an intra skip mode (SKIP mode).

Deriving a pixel value of the prediction block for the current block includes deriving motion information for the current block using a block that has already been decoded and using the derived motion information to obtain a pixel value of the prediction block. It may include the step of deriving.

The already decoded block may include neighboring blocks of the current block.

The already decoded block may include a neighboring block of the current block and a neighboring block of a collocated block in the reference image.

In the deriving pixel value of the prediction block by using the derived motion information, the pixel value of the derived prediction block may be a pixel value of a reference block indicated by the derived motion information.

In the case of deriving the pixel value of the prediction block using the derived motion information when there are two or more reference blocks, the pixel value of the prediction block is derived by a weighted sum of the pixel values of the reference block, The reference block may be a block indicated by the derived motion information.

Deriving an error compensation value for the current block includes deriving an error parameter for an error model of the current block and using the error model and the derived error parameter for the current block. Deriving an error compensation value may be included.

The error model may be a zero-order error model or a first-order error model.

In the deriving of an error parameter for the error model of the current block, the error parameter may be derived using information included in the neighboring block of the current block and the block in the reference image.

When the reference block has two or more, in the step of deriving an error compensation value for the current block using the error model and the derived error parameter, the error compensation value is derived by a weighted sum of the error block values. The error block value may be an error compensation value of the current block derived for each reference block.

In the deriving the pixel value of the final prediction block using the pixel value of the prediction block and the error compensation value, the error compensation value is selectively used according to information indicating whether to apply the error compensation, and the error compensation is applied. Information indicating whether the information is included in a slice header, a picture parameter set, or a sequence parameter set may be information transmitted from an encoder to a decoder.

Another embodiment of the present invention is an inter prediction mode prediction method. The method includes deriving a pixel value of the prediction block for the current block, deriving an error compensation value for the current block, and using the pixel value of the prediction block and the error compensation value. Wherein the error compensation value is a sample value of an error compensation block for compensating an error between the current block and a reference block, wherein the reference block is a block in a reference image and related to a predicted value of a pixel in the current block. It is a block that contains information.

Deriving a pixel value of the prediction block for the current block includes deriving motion information for the current block using a block that has already been decoded and using the derived motion information to obtain a pixel value of the prediction block. It may include the step of deriving.

Deriving an error compensation value for the current block includes deriving an error parameter for an error model of the current block and using the error model and the derived error parameter for the current block. Deriving an error compensation value may be included.

Another embodiment of the present invention is an image decoding device. The apparatus entropy-decodes a bitstream received from the decoder according to a probability distribution to generate a residual block related information, a pixel value of a prediction block for the current block, and the current block. A prediction unit for deriving an error compensation value for the prediction block and a pixel value of the final prediction block using the pixel value of the prediction block and the error compensation value, and generating a reconstruction block using the residual block and the final prediction block. And an adder, wherein the error compensation value is a sample value of an error compensation block for compensating an error between the current block and a reference block, and the reference block is a block in a reference image and includes information related to prediction values of pixels in the current block. The prediction unit may predict the current block in the inter-screen skip mode. The.

The prediction unit may derive motion information about the current block by using a block that has already been decoded, and may derive pixel values of the prediction block by using the derived motion information.

The prediction unit may derive an error parameter for the error model of the current block and derive an error compensation value for the current block by using the error model and the derived error parameter.

According to the video encoding method of the present invention, the video encoding / decoding efficiency is improved while minimizing the amount of transmitted information.

According to the image decoding method according to the present invention, the amount of transmitted information is minimized and the image encoding / decoding efficiency is improved.

According to the skipped mode prediction method according to the present invention, image encoding / decoding efficiency is improved while minimizing the amount of transmitted information.

1 is a block diagram illustrating a configuration of an image encoding apparatus according to an embodiment.

2 is a block diagram illustrating a configuration of an image decoding apparatus according to an embodiment.

3 is a flowchart schematically illustrating a skip mode prediction method using error compensation according to an exemplary embodiment of the present invention.

4 is a flowchart schematically illustrating a method of deriving pixel values of a prediction block according to an embodiment of the present invention.

FIG. 5 is a conceptual diagram schematically illustrating an embodiment of a neighboring block with respect to a current block, which is used when deriving motion information in the embodiment of FIG. 4.

FIG. 6 is a conceptual diagram schematically illustrating an embodiment of a neighboring block of a current block and a neighboring block of the same position block in a reference image, which are used when deriving motion information in the embodiment of FIG. 4.

7 is a flowchart schematically illustrating a method for deriving an error compensation value according to an embodiment of the present invention.

8 is a conceptual diagram schematically showing an embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention.

9 is a conceptual diagram schematically showing another embodiment of an error parameter derivation method for a zero-order error model according to the present invention.

10 is a conceptual diagram schematically showing still another embodiment of an error parameter derivation method for a zero-order error model according to the present invention.

11 is a conceptual diagram schematically showing still another embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention.

12 is a conceptual diagram schematically showing still another embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention.

FIG. 13 is a conceptual diagram schematically illustrating an embodiment of a motion vector used for deriving an error compensation value using weights in the embodiment of FIG. 7.

14 is a conceptual diagram illustrating an embodiment of a method of deriving a pixel value of a final prediction block by using information about a position in a current block of a prediction target pixel.

EMBODIMENT OF THE INVENTION Hereinafter, embodiment of this invention is described concretely with reference to drawings. In describing the embodiments of the present specification, when it is determined that a detailed description of a related well-known configuration or function may obscure the gist of the present specification, the detailed description thereof will be omitted.

When a component is said to be “connected” or “connected” to another component, it may be directly connected to or connected to that other component, but it may be understood that another component may exist in between. Should be. In addition, the description "include" a specific configuration in the present invention does not exclude a configuration other than the configuration, it means that additional configuration may be included in the scope of the technical spirit of the present invention or the present invention.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component.

In addition, the components shown in the embodiments of the present invention are shown independently to represent different characteristic functions, and do not mean that each component is made of separate hardware or one software component unit. In other words, each component is included in each component for convenience of description, and at least two of the components may be combined into one component, or one component may be divided into a plurality of components to perform a function. Integrated and separate embodiments of the components are also included within the scope of the present invention without departing from the spirit of the invention.

In addition, some of the components may not be essential components for performing essential functions in the present invention, but may be optional components for improving performance. The present invention can be implemented including only the components essential for implementing the essentials of the present invention except for the components used for improving performance, and the structure including only the essential components except for the optional components used for improving performance. Also included in the scope of the present invention.

1 is a block diagram illustrating a configuration of an image encoding apparatus according to an embodiment. Referring to FIG. 1, the image encoding apparatus 100 may include a motion predictor 111, a motion compensator 112, an intra predictor 120, a switch 115, a subtractor 125, and a converter 130. And a quantization unit 140, an entropy encoding unit 150, an inverse quantization unit 160, an inverse transform unit 170, an adder 175, a filter unit 180, and a reference image buffer 190.

An image may also be called a picture. Hereinafter, a picture may have the same meaning as an image according to context and need. The image may include both a frame picture used in the progressive scan method and a field picture used in the interlace scan method. In this case, the field picture may be composed of two fields including a top field and a bottom field.

A block means a basic unit of image encoding and decoding. When encoding or decoding an image, a coding or decoding unit refers to a divided unit when an image is divided and encoded or decoded. Therefore, a macroblock, a coding unit (CU), a prediction unit (PU), and a PU are used. It may be called a partition (PU partition), a transform unit (TU), a transform block, or the like. Hereinafter, a block may have one of the above-described block types according to a unit in which current encoding is performed.

The coding unit may be hierarchically divided based on a quad tree structure. In this case, whether to split the coding unit may be expressed using depth information and a split flag. The largest coding unit is called a large coding unit (LCU), and the smallest coding unit is called a smallest coding unit (SCU). The coding unit may have a size of 8x8, 16x16, 32x32, 64x64, 128x128, and the like. Here, the split flag indicates whether the current coding unit is split. In addition, when the division depth is n, it may indicate that n divisions have been performed in the LCU. For example, the partition depth 0 may indicate that the partition is not divided in the LCU, and the partition depth 1 may indicate that the partition is divided once in the LCU. The structure of the coding unit as described above may also be called a coding tree block (CTB).

For example, one coding unit in a CTB may be divided into a plurality of small coding units based on size information, depth information, split flag information, and the like of the coding unit. At this time, each small coding unit may also be divided into a plurality of smaller coding units based on size information, depth information, split flag information, and the like.

The image encoding apparatus 100 may perform encoding in an intra mode or an inter mode on an input image and output a bit stream. Intra prediction means intra prediction and inter prediction means inter prediction. In the intra mode, the switch 115 is switched to intra, and in the inter mode, the switch 115 is switched to inter. The image encoding apparatus 100 may generate a prediction block for an input block of an input image and then encode a difference between the input block and the prediction block.

In the intra mode, the intra predictor 120 may generate a prediction block by performing spatial prediction using pixel values of blocks that are already encoded around the current block.

In the inter mode, the motion predictor 111 may obtain motion information by finding a region that best matches an input block in the reference image stored in the reference image buffer 190 during the motion prediction process. have. Motion information including motion vector information and reference picture index information may be encoded by the encoder and transmitted to the decoder. The motion compensator 112 may generate a prediction block by performing motion compensation using the motion information and the reference image stored in the reference image buffer 190.

Here, the motion information refers to related information used to obtain the position of a reference block used for intra prediction or inter prediction. The motion information may include motion vector information indicating a relative position between a current block and a reference block, and reference picture index information indicating whether a reference block exists in a reference picture when a plurality of reference pictures are used. If only one reference picture exists, the reference picture index information may not be included in the motion information. The reference block is a block in the reference picture and is a block including related information corresponding to a predictor of pixels in the current block. In the case of inter prediction, the position of a reference picture may be indicated by motion information such as a reference picture index value and a motion vector value. The reference block in the picture means a reference block existing in the current image. When the motion vector is encoded, the position of the reference block in the picture may be indicated by explicit motion information. In addition, the position of the reference block in the screen may be indicated by motion information inferred using a template or the like.

In the case of an image with a significant change in illumination between screens, for example, an image in which brightness changes in time, deterioration of image quality may occur if the change in illumination is not taken into account during encoding. In this case, in order to compensate for an error due to a lighting change, the encoder may adaptively apply a weight coefficient to the reference image and then perform prediction using the reference image to which the weight coefficient is applied. This prediction method may be called weighted prediction. When weight prediction is used, parameters used for weight prediction, including weight coefficients, may be transmitted from the encoder to the decoder. In this case, the encoder may perform weight prediction using the same parameter as the reference image unit used in the current image.

In performing the error compensation, the encoder uses a difference value of a mean value of pixel values of an already encoded block adjacent to a current block and a mean value of neighboring block pixel values of a reference block as a DC offset value for the current block. It may be. In addition, the encoder may perform prediction by considering lighting changes even when predicting a motion vector. This error compensation method may be referred to as local lighting change compensation. When a local illumination change compensation method is used for an image having a large change in illumination between screens, since the inter prediction is performed using the derived offset value, image compression performance may be improved.

The encoder may perform prediction on the current block using various motion prediction methods, and each prediction method may be applied in different prediction modes. For example, a prediction mode used in inter prediction may include a skip mode, a merge mode, an advanced motion vector prediction (AMVP) mode, and the like. The mode in which the prediction is performed can be determined through a rate-distortion optimization process. The encoder can send information about which mode is used for prediction to the decoder.

The skip mode is an encoding mode in which transmission of a residual signal and motion information, which is a difference between a prediction block and a current block, is omitted. The skip mode may be applied to intra prediction, uni-directional prediction between screens, bi-prediction between screens, inter-screen or multi-hypothesis skip mode, and so on. The skip mode applied to the unidirectional prediction between screens may be referred to as a P skip mode, and the skip mode applied to the inter prediction between screens may be referred to as a B skip mode.

In the skip mode, the encoder may generate the prediction block by deriving the motion information of the current block using the motion information provided from the neighboring block. In addition, in the skipped mode, values of residual signals of the prediction block and the current block may be zero. Accordingly, the motion information and the residual signal may not be transmitted from the encoder to the decoder, and the encoder may generate the prediction block using only information provided from the previously encoded region.

Peripheral blocks that provide motion information to the current block may be selected in various ways. For example, the motion information may be derived from motion information of a predetermined number of neighboring blocks. The number of neighboring blocks may be one or two or more.

In addition, neighboring blocks that provide motion information to the current block in the skipped mode may be the same as candidate blocks used to obtain the motion information in the merge mode. The encoder may transmit a merge index indicating which neighboring block among the neighboring blocks is used to derive motion information to the decoder. In this case, the decoder may derive the motion information of the current block from the neighboring block indicated by the merge index. The skip mode in this case may also be called a merge skip mode.

The motion vector may be obtained through median calculation for each horizontal and vertical component. The reference picture index may be selected as the closest picture on the time axis with the current picture among the pictures in the reference picture list. The method of deriving the motion information is not limited to the above method, and the motion information for the current block may be derived in various ways.

The subtractor 125 may generate a residual block by the difference between the input block and the generated prediction block. The transform unit 130 may output a transform coefficient by performing transform on the residual block. The quantization unit 140 may output a quantized coefficient by quantizing the input transform coefficient according to the quantization parameter.

The entropy encoder 150 may output a bit stream by performing entropy encoding based on values calculated by the quantizer 140 or encoding parameter values calculated in the encoding process. For entropy coding, coding methods such as exponential golomb, context-adaptive variable length coding (CAVLC), and context-adaptive binary arithmetic coding (CABAC) may be used.

When entropy encoding is applied, a small number of bits are assigned to a symbol having a high probability of occurrence and a large number of bits are assigned to a symbol having a low probability of occurrence, thereby representing bits for encoding symbols. The size of the heat can be reduced. Therefore, compression performance of image encoding may be increased through entropy encoding.

The quantized coefficients may be inversely quantized by the inverse quantizer 160 and inversely transformed by the inverse transformer 170. The inverse quantized and inverse transformed coefficients are added to the prediction block through the adder 175 and a reconstruction block can be generated.

The reconstruction block passes through the filter unit 180, and the filter unit 180 applies at least one or more of a deblocking filter, a sample adaptive offset (SAO), and an adaptive loop filter (ALF) to the reconstruction block or the reconstruction picture. can do. The reconstructed block that has passed through the filter unit 180 may be stored in the reference image buffer 190.

2 is a block diagram illustrating a configuration of an image decoding apparatus according to an embodiment. 2, the image decoding apparatus 200 may include an entropy decoder 210, an inverse quantizer 220, an inverse transformer 230, an intra predictor 240, a motion compensator 250, and a filter. 260 and a reference picture buffer 270.

The image decoding apparatus 200 may receive a bitstream output from the encoder and perform decoding in an intra mode or an inter mode, and output a reconstructed image, that is, a reconstructed image. In the intra mode, the switch may be switched to intra, and in the inter mode, the switch may be switched to inter. The image decoding apparatus 200 may generate a reconstructed block, that is, a reconstructed block by obtaining a residual block reconstructed from the received bitstream, generating a prediction block, and adding the reconstructed residual block and the prediction block.

The entropy decoder 210 may entropy decode the input bitstream according to a probability distribution to generate symbols including symbols in the form of quantized coefficients. The entropy decoding method is similar to the entropy coding method described above.

When the entropy decoding method is applied, a small number of bits are allocated to a symbol having a high probability of occurrence and a large number of bits are allocated to a symbol having a low probability of occurrence, whereby the size of the bit string for each symbol is increased. Can be reduced. Therefore, the compression performance of image decoding can be improved through an entropy decoding method.

The quantized coefficients are inversely quantized by the inverse quantizer 220 and inversely transformed by the inverse transformer 230, and as a result of the inverse quantization / inverse transformation of the quantized coefficients, a reconstructed residual block may be generated.

In the intra mode, the intra predictor 240 may generate a predictive block by performing spatial prediction using pixel values of an already encoded block around the current block. In the inter mode, the motion compensator 250 may generate a prediction block by performing motion compensation using the motion information received from the encoder and the reference image stored in the reference image buffer 270.

The decoder may use an error compensation technique such as a weight prediction technique or a regional illumination change technique to prevent image degradation due to the change of illumination between screens. A method for compensating for an error caused by a change in illumination has been described above with reference to the embodiment of FIG. 1.

As in the case of the above-described encoder, the decoder may perform prediction on the current block using various prediction methods, and each prediction method may be applied in different prediction modes. For example, a prediction mode used in inter prediction may include a skip mode, a merge mode, an advanced motion vector prediction (AMVP) mode, and the like. The decoder may receive information from the encoder about which mode is used for prediction.

In particular, in the skip mode, the decoder may generate the prediction block by deriving the motion information of the current block by using the motion information provided from the neighboring block. In addition, in the skipped mode, values of residual signals of the prediction block and the current block may be zero. Therefore, the decoder may not receive the separate motion information and the residual signal, and may generate the prediction block using only information provided from the previously encoded region. The details of the omission mode are similar to those described above in the encoder.

The reconstructed residual block and the prediction block are added through the adder 255, and the added block passes through the filter unit 260. The filter unit 260 may apply at least one or more of the deblocking filter, SAO, and ALF to the reconstructed block or the reconstructed picture. The filter unit 260 outputs a reconstructed image, that is, a reconstructed image. The reconstructed picture may be stored in the reference picture buffer 270 to be used for inter prediction.

In the skipping mode described above in the embodiments of FIGS. 1 and 2, since the motion information and the residual signal are not transmitted to the decoder, the amount of information transmitted may be smaller than in the case where other prediction modes are used. Therefore, if the frequency of selecting the skip mode is high by the rate-distortion optimization process, the image compression efficiency may be improved.

In the skipped mode, the encoder and the decoder perform prediction using only the information of the previous encoded / decoded region. In the skipped mode, prediction is performed using only motion information of neighboring blocks and pixel values of a reference block. Therefore, when compared with other modes in terms of rate-distortion optimization, the rate required for encoding is minimal or distortion ( distortion can be large. Since the distortion between the screens is very large in the case of an image with a large lighting change between screens, the possibility that the skip mode is selected as the optimal prediction mode of the current block may be lowered. In addition, when encoding / decoding is performed at a low bit rate, that is, even when quantization / dequantization is performed using a large quantization step, the inter-screen distortion increases, so the skip mode is selected as the optimal prediction mode of the current block. Can be less likely.

When local illumination change compensation is used in another inter prediction mode except the omission mode, error compensation prediction considering the illumination change is performed on blocks to which the prediction mode other than the omission mode is applied. In addition, error compensation may be reflected in deriving the motion vector of the blocks. In this case, assuming that encoding is performed in a skipped mode for the current block, motion vector values obtained from neighboring blocks may be values reflecting error compensation, and the neighboring blocks may have a DC offset value other than zero. However, in the skipped mode, since the pixel value of the reference block is used as the prediction value of the current block pixel without additional motion compensation, error compensation cannot be reflected in the prediction value. In this case, there is a high possibility that the distortion between the screens increases in the skipped mode. Therefore, if error compensation such as Mean-Removed Sum of Absolute Difference (MRSAD) is used to reflect the lighting change in other inter prediction modes except skip mode, the skip mode is likely to be selected as the optimal prediction mode of the current block. Can be lowered.

Therefore, in order to increase the rate at which the skip mode is selected as the optimal prediction mode, a skip mode prediction method using error compensation may be provided.

3 is a flowchart schematically illustrating a skip mode prediction method using error compensation according to an exemplary embodiment of the present invention. In the embodiment of FIG. 3, the encoder and the decoder perform prediction on the current block in the skipped mode. As an example, prediction according to the embodiment of FIG. 3 may be performed by an intra predictor, a motion predictor, and / or a motion compensator of the encoder and the decoder.

Hereinafter, embodiments of the present invention can be equally applied to an encoder and a decoder, and embodiments are described with reference to a decoder. In the encoder, previously encoded blocks, rather than previously decoded blocks, described below, may be used for prediction of the current block. The previously encoded block may mean a block encoded before performing prediction and / or encoding on the current block, and the previously decoded block may mean a block decoded before performing prediction and / or decoding on the current block.

Referring to FIG. 3, the decoder derives pixel values of a prediction block for a current block from previously decoded blocks (S310). The prediction block refers to a block generated as a result of performing prediction on the current block.

The decoder may derive motion information for the current block by using motion information of previously decoded blocks. Once the motion information is derived, a pixel value of the reference block in the reference image may be derived using the motion information. In the embodiment of FIG. 3, since the decoder performs prediction in the skipped mode, the pixel value of the reference block may be the pixel value of the prediction block. There may be one reference block or two or more reference blocks. When there are two or more reference blocks, the decoder may generate two or more prediction blocks by using each reference block separately, and use the weighted sum of the pixel values of the two or more reference blocks to determine the prediction block for the current block. It is also possible to derive the pixel value. In this case, the prediction block may also be called a motion prediction block.

The decoder derives a sample value of the error compensation block for the current block (S320). The error compensation block may have the same size as the prediction block. Hereinafter, the error compensation value has the same meaning as the sample value of the error compensation block.

The decoder may derive an error parameter for an error model of the current block by using information included in a neighboring block of the current block and / or an error parameter included in a neighboring block of the current block. When the error parameter is derived, the decoder may derive an error compensation value for the current block by using the error model information and the derived error parameter information. In this case, if there are two or more reference blocks, the decoder may derive an error compensation value using motion information and the like together with the error parameter information and the error model information.

The decoder derives the pixel value of the last prediction block for the current block by using the pixel value and the error compensation value of the prediction block (S330). The pixel value derivation method of the final prediction block may vary depending on the number of reference blocks or the encoding / decoding method of the current image.

Details of each step of the embodiment of FIG. 3 described above will be described later.

4 is a flowchart schematically illustrating a method of deriving pixel values of a prediction block according to an embodiment of the present invention.

Referring to FIG. 4, the decoder derives motion information for the current block from previously decoded blocks (S410). The motion information may be used to derive the pixel value of the prediction block for the current block.

The decoder may derive the motion information from the motion information of the neighboring block for the current block. In this case, the neighboring block is a block existing in the current image, and is a block previously decoded. The decoder may derive one piece of motion information or may derive two or more pieces of motion information. If there is two or more pieces of motion information, two or more reference blocks may be used for prediction of the current block according to the number of motion information.

The decoder may derive the motion information using not only the motion information of the neighboring block with respect to the current block but also the motion information of the neighboring block with respect to the collocated block in the reference picture. In this case, the encoder may derive one or two or more pieces of motion information. If there is two or more pieces of motion information, two or more reference blocks may be used for prediction of the current block according to the number of motion information.

The decoder may derive the motion information using at least one of the above-described methods, and details of the motion information derivation method for the above cases will be described later.

The decoder derives pixel values of the prediction block by using the derived motion information (S420).

In the skipped mode, when the values of the residual signals of the reference block and the current block are 0 and there is one reference block, the pixel values of the reference block in the reference image may be the pixel values of the prediction block with respect to the current block.

The reference block may be a block in which a co-located block in the reference image with respect to the current block is moved by the value of the motion vector. In this case, the pixel value of the prediction block for the current block may be a block in which a block located at the same position as the prediction block in the reference image is moved by the value of the motion vector, that is, the pixel value of the reference block. The pixel value of the prediction block with respect to the current block may be expressed as in Equation 1 below.

[Equation 1]

P _cur (t0, X, Y) = P _ref (X + MV (x), Y + MV (y))

Here, P _cur denotes a pixel value of the prediction block with respect to the current block, and P _ref denotes a pixel value of the reference block. X represents coordinates in the x-axis direction of the pixel in the current block, and Y represents coordinates in the y-axis direction of the pixel in the current block. In addition, MV (x) represents the magnitude of the motion vector in the x-axis direction and MV (y) represents the magnitude of the motion vector in the y-axis direction.

When there are two or more reference blocks, the pixel value of the prediction block for the current block may be derived by using the weighted sum of the pixel values of the reference blocks. That is, when N reference blocks are used to generate the prediction block, the weighted sum of the pixel values of the N reference blocks may be the pixel value of the prediction block. The motion information for the current block derived when there are N reference blocks is {{ref_idx1, MV1}, {ref_idx2, MV2},... , {ref_idxn, MVn}}, the pixel value of the prediction block derived using only the reference block corresponding to the i-th motion information may be represented by Equation 2 below.

[Equation 2]

P _{cur_ref_i} (t0, X, Y) = P _{ref_i} (X + MV _i (x), Y + MV _i (y))

Here, P _{cur_ref_i} denotes a pixel value of the prediction block derived using only the reference block corresponding to the i-th motion information, and P _{ref_i} denotes a pixel value of the reference block corresponding to the i-th motion information. In addition, MV _i (x) represents the magnitude in the x-axis direction of the i-th motion vector and MV _i (y) represents the magnitude in the y-axis direction of the i-th motion vector.

In this case, the pixel value of the prediction block for the current block, which is derived by the weighted sum of the pixel values of the N reference blocks, may be expressed as in Equation 3 below.

[Equation 3]

Here, P _{cur_ref} represents a pixel value of the prediction block for the current block, and w _i represents a weight applied to the pixel value of the reference block corresponding to the i-th motion information. In this case, the sum of the weights may be 1. In one embodiment, in pair prediction where two reference blocks are used, w ₁ = 1/2 and w ₂ = 1/2 when N = 2.

In the above-described embodiment of Equation 3, the weights applied to the pixel values of each reference block may use a predetermined fixed value. In addition, each weight may be set differently according to the distance between the reference picture including the reference block to which the weight is applied and the current picture. In addition, the respective weights may be set differently according to the spatial position of the pixel in the prediction block. The spatial position of the pixel in the prediction block may be the same as the spatial position of the pixel currently predicted in the current block. When the weights are set differently according to the spatial position of the pixel in the prediction block, the pixel value of the prediction block with respect to the current block may be expressed as in Equation 4 below.

[Equation 4]

Here, the weight w _i (X, Y) is a weight applied to the pixel value of the reference block corresponding to the i-th motion information. Referring to Equation 4, the weight w _i (X, Y) may have different values according to the coordinates (X, Y) of the pixels in the prediction block.

The decoder may use at least one of the aforementioned methods in deriving a pixel value of the prediction block using the derived motion information.

FIG. 5 is a conceptual diagram schematically illustrating an embodiment of a neighboring block with respect to a current block, which is used when deriving motion information in the embodiment of FIG. 4. The neighboring block for the current block is a block in the current image, which is a block previously decoded previously.

The motion information may include information about the position of the reference block with respect to the current block. The decoder may derive one piece of motion information or may derive two or more pieces of motion information. If there is two or more pieces of motion information, two or more reference blocks may be used for prediction of the current block according to the number of motion information.

Referring to the embodiment of FIG. 5, the decoder may derive motion information for the current block using motion information of blocks A, B, and C. According to an embodiment, the decoder may select, as a reference image, an image that is closest in time axis to an image including a current block among the images included in the reference image list. In this case, the decoder may select only the motion information indicating the selected reference image among the motion information of the block A, the block B, and the block C, and use the same to derive the motion information of the current block. The decoder can obtain the median of the selected motion information for the horizontal and vertical components, in which case the median can be the motion vector of the current block.

The position and number of neighboring blocks used for deriving the motion information of the current block are not limited to the above-described embodiments, and various positions and numbers of neighboring blocks different from the embodiment of FIG. 5 may be used to derive the motion information according to the implementation.

The motion information on the neighboring block may be information encoded / decoded by a merge method. In the merge mode, the decoder may use the motion information included in the block indicated by the merge index (merge idx) of the current block, as the motion information of the current block. Therefore, when the neighboring block is encoded / decoded in the merge mode, the motion information deriving method according to the embodiment may further include deriving motion information of the neighboring block by the merge method.

FIG. 6 is a conceptual diagram schematically illustrating an embodiment of a neighboring block of a current block and a neighboring block of the same position block in a reference image, which are used when deriving motion information in the embodiment of FIG. 4.

As described above in the embodiment of FIG. 4, the decoder may derive the motion information using not only the motion information of the neighboring block for the current block but also the motion information of the neighboring block for the same position block in the reference image. .

Referring to FIG. 6, all blocks in a reference picture may be blocks that have already been decoded. Therefore, all blocks around the same location block in the reference image may be selected as a block used for deriving motion information of the current block. The neighboring block for the current block is a block in the current image, which is a block previously decoded.

In order to derive the motion information of the current block, in one embodiment, the median of the blocks may be used. In addition, a motion vector competition method may be used to derive motion information of the current block. When motion information prediction encoding is performed in the motion vector competition method, a plurality of prediction candidates may be used. The decoder may select an optimal prediction candidate in consideration of rate control cost among a plurality of prediction candidates, and perform motion information prediction encoding using the selected prediction candidate. In this case, information about which candidate is selected from among the plurality of prediction candidates may be further needed. The motion vector competition method may include, for example, an AVP (Adavanced Motion Vector Prediction).

7 is a flowchart schematically illustrating a method for deriving an error compensation value according to an embodiment of the present invention.

Referring to FIG. 7, the decoder derives an error parameter for an error model of a current block (S710).

In order to derive an error compensation value according to an embodiment of the present invention, an error model obtained by modeling an error between a current block and a reference block may be used. There may be various types of error models used for error compensation of the current block, for example, there may be a zero-order error model, a first-order error model, an N-order error model, and a nonlinear error model.

The error compensation value of the zero-order error model may be represented by Equation 5 according to an embodiment.

[Equation 5]

Error compensation value (x, y) = b

Here, (x, y) is the coordinate of the prediction target pixel in the current block. b is a DC offset and corresponds to an error parameter of the 0th order error model. When error compensation is performed using the zero-order error model, the decoder may derive the pixel value of the final prediction block by using the pixel value of the prediction block and the error compensation value. This can be represented by the following equation (6).

[Equation 6]

Pixel value (x, y) of the final prediction block = Pixel value (x, y) of the prediction block + b

The error compensation value of the first order error model may be represented by Equation 7 according to one embodiment.

[Equation 7]

Error compensation value (x, y) = (a-1) * pixel value of prediction block (x, y) + b

Here, a and b correspond to error parameters of the first order error model. When the first order error model is used, the decoder calculates the error parameters a and b and then performs compensation. The pixel value of the final prediction block derived when error compensation is performed using the first order error model may be represented by Equation (8).

[Equation 8]

Pixel value of the last prediction block (x, y)

= Pixel value (x, y) of the prediction block + error compensation value (x, y)

= a * pixel value of the prediction block (x, y) + b

The error compensation value of the N-th order error model and the pixel value of the final prediction block may be represented by Equation 9 according to an embodiment.

[Equation 9]

Error compensation value (x, y) = a _n * P (x, y) ⁿ + a _n-1 * P (x, y) ^n-1 +... + (a-1) P (x, y) + b

Pixel value of the last prediction block (x, y)

= Pixel value (x, y) of the prediction block + error compensation value (x, y)

= a _n * P (x, y) ⁿ + a _n-1 * P (x, y) ^n-1 +... + a * P (x, y) + b

Here, P means the pixel value of the prediction block for the current block.

The decoder may use other nonlinear error models. In this case, the error compensation value may be represented by the following Equation 10 in one embodiment.

[Equation 10]

Error Compensation Value (x, y) = f (P (x, y))

Where f may be any function that is not linear.

The error parameters used in the current block may all be the same value as described above in the above embodiments. However, different error parameters may be used according to the position of the prediction target pixel in the current block. The decoder may derive the final error compensation value by applying a weight to an error compensation value derived using the same error parameter according to the spatial position of the pixel to be predicted in the current block. In this case, the final error compensation value may vary according to the position in the current block of the prediction target pixel. The final error compensation value can be represented by the following equation (11).

[Equation 11]

Final error compensation value (x, y) = error compensation value derived using the same error parameter (x, y) * w (x, y)

Where w means weight.

The decoder may derive the error parameter according to the error model of the current block. Since the decoder cannot obtain the actual pixel value of the current block in the error parameter deriving step, an error parameter of the error model for the current block may be derived using information included in a neighboring block of the current block. Here, the neighboring block of the current block is a neighboring block adjacent to the current block and means a previously encoded block.

In deriving the error parameter, the decoder may use only pixel values of the neighboring block for the current block, pixel values of the reference block, and / or pixel values of the neighboring pixel for the reference block. Further, in deriving the error parameter, the decoder may use not only the pixel values but also motion information and / or encoding mode information included in a neighboring block for the current block, and motion information and / or included in the current block. Alternatively, encoding mode information may be used together.

When the zero-order error model is used for error compensation, and only pixel values of the neighboring block for the current block, pixel values of the reference block, and / or pixel values of the neighboring pixel for the reference block are used for error parameter derivation, an embodiment The error parameter b can be represented by the following equation (12).

 [Equation 12]

b = Mean (Current Block ’)-Mean (Reference Block’)

Here, Mean (Current Block ') may be an average of pixel values of a block that is already encoded as a neighboring block of the current block. In the error parameter deriving step, since the decoder cannot obtain an average of the current block pixel values, the decoder can obtain an average of the neighboring block pixel values of the current block and use the same to derive the error parameter. Mean (Reference Block ') may be an average of reference block pixel values or an average of neighboring block pixel values of the reference block. In the error parameter deriving step, since the decoder can obtain the reference block pixel value, not only the neighboring block pixel value of the reference block but also the reference block pixel value can be used for the error parameter derivation.

Specific embodiments of the error parameter derivation method for the 0 th order error model will be described later.

When the first order error model is used for error compensation, the error parameter a may be obtained by Equation 13 below.

[Equation 13]

a = Mean (Current Block ’) / Mean (Reference Block’)

Here, Mean (Current Block ') and Mean (Reference Block') have the same meaning as Mean (Current Block ') and Mean (Reference Block') in Equation 12. In this case, the error parameter b may be obtained in the same manner as in the zero order error model.

When the first error model is used for error compensation for error compensation, in another embodiment, the decoder may derive an error parameter using a method used in weighted prediction (WP). At this time, the error parameter a can be obtained by the following equation (14).

[Equation 14]

Here, w ^Y [n] is a weighting value and represents an error parameter a. In addition, H represents the height of the current block and the reference block, and W represents the width of the current block and the reference block. c ^Y _ij denotes a luma pixel value in the current frame, and r ^Y [n] _ij denotes a luma pixel value in the reference frame. Also, the value of iclip3 (a, b, c) is a when c is less than a, b when c is greater than b, and c otherwise.

In addition, when the method used in the weight prediction is used for error parameter derivation, the error parameter b can be obtained by the following equation (15).

[Equation 15]

Here, offset ^Y [n] is an offset value and represents the error parameter b.

If the error parameter information for the neighboring block of the current block exists, the decoder may derive the error parameter of the current block using the error parameter of the neighboring block for the current block.

In an embodiment, when there is one neighboring block having error parameter information as a neighboring block having the same motion vector as the current block, the decoder may use the error parameter as an error parameter of the current block. In another embodiment, when there are two or more neighboring blocks having error parameter information as neighboring blocks having the same motion vector as the current block, the decoder obtains a weighted sum of the error parameters of the neighboring blocks and returns the value to the error parameter of the current block. Can also be used as In addition, the decoder may use the error parameter of the neighboring block with respect to the current block as the initial prediction value in deriving the error parameter.

The decoder may derive the error parameter using additional information received from the encoder. In an embodiment, the encoder may further transmit difference information between the actual error parameter and the prediction error parameter to the decoder. In this case, the decoder may derive the actual error parameter by adding the difference information received to the prediction error parameter information. Here, the prediction error parameter means an error parameter predicted by the encoder and the decoder. The parameter information additionally transmitted from the encoder is not limited to the difference information and may have various types.

Referring back to FIG. 7, the decoder derives an error compensation value for the current block by using the error model and the derived error parameter (S720).

When one reference block is used, the decoder may use the error compensation value derived by the error model and the error parameter as the error compensation value of the current block.

If there are two or more reference blocks used, the decoder may generate two or more prediction blocks by using each reference block separately. In this case, the decoder may derive an error parameter for each prediction block and derive an error compensation value for each prediction block. Hereinafter, the error compensation value for each prediction block may be referred to as an error block value. The decoder may derive an error compensation value for the current block by the weighted sum of the error block values.

In one embodiment, the weight may be determined using distance information between a current block and a prediction block corresponding to each reference block. In addition, when there are two or more reference blocks, there may be two or more pieces of motion information indicating each reference block, and the weight may be determined using directional information of each piece of motion information. In addition, the weight may be determined by using not only the directional information between the respective motion information but also the size information of each motion information. When two or more reference blocks are used, specific embodiments of the weights used to derive the error compensation value will be described later.

The weight may be obtained using at least one or more of the above-described embodiments. In addition, the method of determining the weight is not limited to the above-described embodiments, and the weight may be determined in various ways depending on the implementation.

8 is a conceptual diagram schematically showing an embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention. In the embodiment of FIG. 8, N represents the height and width of the current block and the reference block. D represents the number of lines of neighboring pixels adjacent to the current block and the reference block. For example, D is 1, 2, 3, 4,... , N, and the like.

Referring to FIG. 8, the decoder may use pixel values of pixels around the current block and pixels of pixels around the reference block to derive an error parameter. At this time, the error parameter of the 0th order model can be represented by the following equation (16).

[Equation 16]

offset = Mean (Current Neighbor)-Mean (Ref.Neighbor)

Here, offset represents an error parameter of the 0th order model. Mean (Current Neighbor) represents an average of pixel values of neighboring pixels adjacent to the current block. In the error parameter deriving step, since the decoder cannot obtain an average of the current block pixel values, the decoder can obtain an average of pixel values of pixels around the current block and use the same to derive the error parameter. Mean (Ref. Neighbor) represents an average of pixel values of neighboring pixels of the reference block.

Referring to Equation 16, the decoder may derive a difference value between a pixel value average of pixels around the current block and a pixel value average of pixels around the reference block as an error parameter value. In this case, the range of pixels used to derive the error parameter may be determined in various ways. In an embodiment, when D = 1, only pixel values included in a line immediately adjacent to the current block and / or a reference block may be used for deriving an error parameter.

9 is a conceptual diagram schematically showing another embodiment of an error parameter derivation method for a zero-order error model according to the present invention. N and D in the embodiment of FIG. 9 have the same meaning as N and D in the embodiment of FIG. 8.

Referring to FIG. 9, the decoder may use pixel values of pixels around a current block and pixel values of pixels in a reference block to derive an error parameter. At this time, the error parameter of the 0th order model can be represented by the following equation (17).

[Equation 17]

offset = Mean (Current Neighbor)-Mean (Ref. Block)

Here, Mean (Ref. Block) represents an average of pixel values of pixels in a reference block.

Referring to Equation 17, the decoder may derive the difference value between the pixel value average of the pixels around the current block and the pixel value average of the pixels in the reference block as an error parameter value. In this case, as in the embodiment of FIG. 8, the range of pixels used may be variously determined.

10 is a conceptual diagram schematically showing still another embodiment of an error parameter derivation method for a zero-order error model according to the present invention. N and D in the embodiment of FIG. 10 have the same meaning as N and D in the embodiment of FIG. 8.

Referring to FIG. 10, the decoder may use pixel values of pixels around the current block and pixel values of pixels around the reference block to derive an error parameter. In addition, as described above in the embodiment of FIG. 7, the decoder may derive an error parameter using not only the pixel values but also motion information derived for the current block. At this time, the error parameter of the zero-order model can be represented by the following equation (18).

Equation 18

offset = Mean (weight * Current Neighbor)-Mean (weight * Ref.Neighbor)

Mean (weight * Current Neighbor) represents a weighted average of pixel values of neighboring pixels adjacent to the current block. Mean (weight * Ref. Neighbor) represents a weighted average of pixel values of neighboring pixels of the reference block.

Referring to Equation 18, the decoder may derive a difference value between a weighted average of pixel values of pixels around a current block and a weighted average of pixel values of pixels around a reference block as an error parameter value.

The weight may be obtained using the directionality of the motion information derived for the current block and / or neighboring blocks of the current block. Referring to the embodiment of FIG. 10, when the motion vector has only a horizontal component, the decoder may use only pixel values of pixels included in a current block and a left block of a reference block to derive an error parameter. In this case, a weight value of 1 may be applied to the pixel values of the left block, and a weight value of 0 may be applied to the pixel values of the upper block.

In this case, as in the embodiment of FIG. 8, the range of pixels used may be variously determined.

11 is a conceptual diagram schematically showing still another embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention. N and D in the embodiment of FIG. 11 have the same meanings as N and D in the embodiment of FIG. 8.

Referring to FIG. 11, a decoder may use pixel values of pixels around a current block and pixel values of pixels in a reference block to derive an error parameter. At this time, the error parameter of the 0th order model can be represented by the following equation (19).

[Equation 19]

offset = Mean (weight * Current Neighbor)-Mean (Ref.Block)

Referring to Equation 19, the decoder may derive a difference value between a weighted average of pixel values of pixels around a current block and an average of pixel values of pixels in a reference block as an error parameter value.

The weight may be obtained using the directionality of the motion information derived for the current block and / or neighboring blocks of the current block. Referring to the embodiment of FIG. 11, when the motion vector has only a horizontal component, the decoder may use only pixel values of pixels included in a left block of the current block among pixel values of a pixel adjacent to the current block to derive an error parameter. Can be. In this case, a weight value of 1 may be applied to the pixel values of the left block, and a weight value of 0 may be applied to the pixel values of the upper block.

In this case, as in the embodiment of FIG. 8, the range of pixels used may be variously determined.

12 is a conceptual diagram schematically showing still another embodiment of a method for deriving an error parameter for a zero-order error model according to the present invention. N and D in the embodiment of FIG. 12 have the same meaning as N and D in the embodiment of FIG. 8.

Referring to 1210 of FIG. 12, the decoder may use pixel values of pixels around the current block and pixels of pixels around the reference block to derive an error parameter. At this time, the error parameter of the 0th order model can be represented by the following equation (20).

[Equation 20]

offset = Mean (weight * Current Neighbor)-Mean (weight * Ref.Neighbor)

Mean (weight * Current Neighbor) represents a weighted average of pixel values of pixels around the current block. Mean (weight * Ref. Neighbor) represents a weighted average of pixel values of pixels around the reference block. In the embodiment of FIG. 12, the neighboring pixels of the current block may be pixel values included in block C illustrated in 1220 of FIG. 12. In addition, the neighboring pixels of the reference block may also be pixel values included in the block in the reference image corresponding to the block C. FIG. Hereinafter, in the embodiment of FIG. 12, block C is referred to as an upper left block of the current block, and a block in a reference image corresponding to block C is referred to as an upper left block of the reference block.

Referring to Equation 20, the decoder may derive the difference between the weighted average of the pixel values of the pixels around the current block and the weighted average of the pixel values of the pixels around the reference block as an error parameter value.

The weight may be obtained using the directionality of the motion information derived for the current block and / or neighboring blocks of the current block. Referring to the embodiment of FIG. 12, when the motion vector derived for the current block is the same as the motion vector of block C illustrated in 1220 of FIG. 12, the decoder may determine the number of pixels included in the upper left block of the current block and the neighboring block. Only pixel values can be used to derive the error parameter. For example, a weight value of 1 may be applied to pixel values of an upper left block, and a weight value of 0 may be applied to pixel values of an upper block and a left block. At this time, referring to 1210 of FIG. 12, the upper block of the current block is block B, the left block of the current block is block A, and the upper left block of the current block is block C. The upper block, the left block, and the upper left block of the left block are blocks in the reference picture corresponding to the blocks B, Block A, and Block C, respectively.

In this case, as in the embodiment of FIG. 8, the range of pixels used may be variously determined.

FIG. 13 is a conceptual diagram schematically illustrating an embodiment of a motion vector used for deriving an error compensation value using weights in the embodiment of FIG. 7. 13 shows an embodiment in the case of two reference blocks. T-1, T, and T + 1 mean time for each image. In the embodiment of FIG. 13, an image of time T represents a current image, and a block in the current image represents a current block. In addition, the image of time T-1, T + 1 represents a reference image, and the block in a reference image represents a reference block. Also, in the embodiment of FIG. 13, the motion vector of the current block indicating the reference block in the reference picture at time T-1 is the motion vector of the reference picture list 0 and the current block indicating the reference block in the reference picture at time T + 1. The motion vector of is referred to as the motion vector of the reference picture list1.

As described above in the embodiment of FIG. 7, when there are two or more reference blocks, the decoder may derive an error compensation value for the current block by a weighted sum of error block values. In an embodiment, the weight may be determined using motion information derived for the current block.

Referring to 1310 of FIG. 13, the motion vector of the reference picture list 0 and the motion vector of the reference picture list 1 are symmetric with each other. In this case, the decoder may not apply an error compensation value when deriving the pixel value of the final prediction block. Therefore, when the motion vector of the reference picture list 0 and the motion vector of the reference picture list 1 are symmetric with each other, the weight may be set to zero.

Referring to 1320 of FIG. 13, the motion vector of the reference picture list 0 and the motion vector of the reference picture list 1 are not symmetric with each other. In this case, the decoder may apply an error compensation value when deriving the pixel value of the final prediction block. For example, the weights used in deriving the error compensation value may be set to 1/2 respectively.

As described above in the embodiment of FIG. 3, the decoder may derive the pixel value of the last prediction block for the current block by using the pixel value and the error compensation value of the prediction block. Specific embodiments of the method of deriving the pixel value and the error compensation value of the prediction block have been described above with reference to the embodiments of FIGS. 4 to 13.

When there is one reference block, the decoder may derive the pixel value of the final prediction block by adding an error compensation value to the pixel value of the prediction block.

In deriving the final prediction block pixel value when there is one reference block, the decoder may use the pixel value and the error compensation value of the prediction block derived in the previous step as it is. In this case, the error compensation values for the prediction target pixels in the current block may be values to which the same error parameters are applied.

In this case, the pixel value of the final prediction block for the zero-order error model may be represented by Equation 21 according to an embodiment.

[Equation 21]

Pixel value (x, y) of the final prediction block = Pixel value (x, y) of the prediction block + b

Here, (x, y) is the coordinate of the prediction target pixel in the current block. b is a DC offset and corresponds to an error parameter of the 0th order error model.

Also, in the above case, the pixel value of the final prediction block for the first order error model may be represented by Equation 22 as an embodiment.

[Equation 22]

Pixel value (x, y) of the last prediction block = a * pixel value (x, y) of the prediction block + b

Where a and b correspond to error parameters of the first order error model.

In deriving the final prediction block pixel value when there is one reference block, the decoder can use not only the pixel value and the error compensation value of the prediction block derived in the previous step, but also information about the position in the current block of the prediction target pixel. have.

In this case, the pixel value of the final prediction block for the zero-order error model may be represented by Equation 23 according to an embodiment.

[Equation 23]

Pixel value (x, y) of the last prediction block = pixel value (x, y) of the prediction block + b * w (x, y)

or

Pixel value (x, y) of the last prediction block = pixel value (x, y) of the prediction block + b (x, y)

Here, w (x, y) is a weight applied to the error parameter b. Referring to Equation 23, the weight w (x, y) and the error parameter b (x, y) may have different values according to positions in the current block of the prediction target pixel. Therefore, the pixel value of the final prediction block may vary according to the position in the current block of the prediction target pixel.

Also, in the above case, the pixel value of the final prediction block for the first order error model may be represented by Equation 24 according to an embodiment.

[Equation 24]

Pixel value (x, y) of the last prediction block = a * w1 (x, y) * pixel value (x, y) of the prediction block + b * w2 (x, y)

or

Pixel value (x, y) of the final prediction block = a (x, y) * pixel value (x, y) + b (x, y) of the prediction block

Here, w1 (x, y) is a weight applied to the error parameter a, and w2 (x, y) is a weight applied to the error parameter b. Referring to Equation 24, the weight w1 (x, y), the weight w2 (x, y), the error parameter a (x, y) and the error parameter b (x, y) are located at the position in the current block of the pixel to be predicted. It can have different values. Therefore, the pixel value of the final prediction block may vary according to the position in the current block of the prediction target pixel.

When a large coding unit (large CU) is used, when the same error parameter is applied to all the pixels in the large coding unit, encoding / decoding efficiency may be reduced. In this case, the encoder and the decoder may improve encoding / decoding performance by using information on the current block position of the prediction target pixel together with other information.

14 is a conceptual diagram illustrating an embodiment of a method of deriving a pixel value of a final prediction block by using information about a position in a current block of a prediction target pixel. In the embodiment of FIG. 14, it is assumed that a zero-order error model is used, and a method of deriving the final prediction block pixel value according to the second equation in the embodiment of Equation 23 is used.

In FIG. 14, reference numeral 1410 denotes a reference block, and reference numeral 1420 of FIG. 14 indicates a current block, and the size of the reference block and the current block is 4x4. In the embodiment of FIG. 14, N1, N2, N3, and N4 represent neighboring pixels adjacent to the top of the current block and the reference block, and NA, NB, NC, and ND represent neighboring pixels adjacent to the left of the current block and the reference block. The pixel values of the pixels around the reference block and the pixels around the current block may be different from each other.

Referring to FIG. 14, 16 error parameters b may be derived for 16 pixels in the current block. Each error parameter may have a different value depending on the position of the pixel corresponding to the error parameter.

In an embodiment, the error parameter b (2,3) may be derived using only information included in the N3 and NB pixels among the pixels around the current block and the pixels around the reference block.

In another embodiment, the decoder may derive only some b (i, j) of the error parameters and then derive the remaining error parameters by using the error parameter information and / or information included in the neighboring pixels. For example, the remaining error parameters may be derived by interpolation or extrapolation of already derived error parameters.

For example, referring to FIG. 14, the decoder may first derive three parameters of b (1,1), b (4,1), and b (1,4). At this time, the decoder can obtain b (1,2) and b (1,3) by interpolation of b (1,1) and b (1,4), and b (1,1) and b (4). By interpolation of, 1, b (2,1) and b (3,1) can be obtained. Similarly, the remaining b (i, j) can also be obtained by interpolation or extrapolation of the closest error parameter b values around.

When there are two or more reference blocks, the decoder may generate two or more prediction blocks by using each reference block separately, and use the weighted sum of the pixel values of the two or more reference blocks to determine the prediction block for the current block. It is also possible to derive the pixel value. In this case, the decoder may derive an error block value for each prediction block, and may also derive one error compensation value for the current block by using the weighted sum of the error block values.

When the decoder derives only one pixel value of the prediction block for the current block and one error compensation value for the current block, respectively, by the sum of weights, the pixel value of the final prediction block is derived in a similar manner to the case where there is one reference block. Can be.

If the decoder derives pixel values of two or more prediction blocks using each reference block separately, and derives only error block values for each prediction block instead of one error compensation value, the decoder Only error block values greater than or equal to a certain size may be used to derive the final prediction block pixel value, or only error block values less than or equal to a certain size may be used to derive the final prediction block pixel value. In one embodiment, when a zero-order error model is used, the method may be represented by the following equation (25).

[Equation 25]

Pixel value (x, y) of the final prediction block = 1 / N (pixel value 1 (x, y) of the prediction block + error block value 1 (x, y) * W _th ) + 1 / N (pixel of the prediction block) Value2 (x, y) + error block _value2 (x, y) * W _th ) +. + 1 / N (pixel value N (x, y) of prediction block + error block value N (x, y) * W _th )

Here, W _th is a value multiplied by the error block value to indicate whether or not the error block value is used. In addition, in the embodiment of Equation 25, pixel values of the prediction block are values derived by using each reference block separately.

In the embodiment of Equation 25, for example, only an error block value whose value is greater than or equal to a certain size may be used to derive the final prediction block pixel value. That is, for each prediction block value, if the error parameter b is greater than a predetermined threshold, the value of W _th may be 1, otherwise, the value of W _th may be zero.

In addition, in the embodiment of Equation 25, for example, only an error block value whose value is equal to or less than a specific size may be used to derive the final prediction block pixel value. That is, for each prediction block value, if the error parameter b is smaller than a predetermined threshold, the value of W _th may be 1, otherwise, the value of W _th may be zero.

When the decoder derives pixel values of two or more prediction blocks by using each reference block separately, and derives only error block values for each prediction block instead of one error compensation value, the decoder determines the prediction block. The pixel value of the final prediction block can be derived by the weighted sum of the values and the error block values. In one embodiment, the pixel value of the final prediction block derived by the method may be represented by the following equation (26).

[Equation 26]

Pixel value (x, y) of the final prediction block = W _P1 * Pixel value 1 (x, y) of the prediction block + W _E1 * Error block value 1 (x, y) + W _P2 * Pixel value 2 of the prediction block x, y) + W _E2 * error block value 2 (x, y) +… + W _PN * Pixel value N (x, y) of prediction block + W _EN * Error block value N (x, y)

Here, each of W _P and W _E represents a weight. In this case, the weight may be determined using distance information between a current block and a prediction block corresponding to each reference block. In addition, when there are two or more reference blocks, there may be two or more pieces of motion information indicating each reference block, and the weight may be determined using directional information of each piece of motion information. The weight may also be determined using symmetry information between the pieces of motion information.

The weight may be adaptively obtained using at least one or more of the above-described embodiments. In addition, the method of determining the weight is not limited to the above-described embodiments, and the weight may be determined in various ways depending on the implementation.

The above-described error compensation scheme applied to the prediction in the skipped mode may not be always applied but may be selectively applied according to a coding scheme, a block size, and the like of the current image.

In one embodiment, the encoder decodes a slice header, a picture parameter set, and / or a sequence parameter set by including information indicating whether error compensation is applied. Can be transmitted to the device. For example, when the information is included in a slice header, the decoder applies error compensation to the slice when the value of the information is a first logical value, and applies the error compensation to the slice when the information is a second logical value. Error compensation may not be applied.

In this case, when the information is included in the slice header, whether or not error compensation is applied to each slice may be changed. When the information is included in the picture parameter set, whether or not error compensation is applied to all slices using the picture parameter may be controlled. When the information is included in the sequence parameter set, whether or not error compensation is applied for each type of slice may be controlled. The slice may be an I slice, a P slice, a B slice, or the like.

The information may differently indicate whether error compensation is applied according to a block size of a coding unit, a prediction unit, or the like. For example, the information may include information that error compensation is applied only in a coding unit CU and / or a prediction unit PU of a specific size. Even in such a case, the information may be included in the slice header, the picture parameter set, and / or the sequence parameter set and transmitted.

For example, suppose that the maximum CU size is 128x128, the minimum CU size is 8x8, the depth of the coding tree block (CTB) is 5, and the error compensation is applied only to the 128x128 and 64x64 size CUs. In this case, when error compensation is applied for each CU size, since the size of the CU can be 5 and the depth of the CTB is 5, five flag information is required to indicate whether the error compensation is applied. In one embodiment, the flag information may be represented as 11000 and / or 00011.

When the depth of the CTB is large, it may be inefficient to transmit information on whether error compensation is applied for each CU size. In this case, a method of defining a table for some predefined cases and transmitting an index capable of indicating information on the defined table to the decoder may be used. The defined table may be equally stored in the encoder and the decoder.

The method of selectively applying the error compensation is not limited to the above embodiment, and various methods may be used depending on the implementation or the need.

Even in intra mode, when a prediction mode similar to that of the inter mode can be used, i.e., a prediction mode in which a prediction block is generated using information provided from neighboring blocks and a separate residual signal is not transmitted can be used. The present invention described above may be applied.

When the prediction method using the error compensation according to the embodiment of the present invention is used, when the lighting change between the screens is severe, when the quantization of the large quantization step is applied and / or in other general cases, the error occurring in the skipped mode Can be reduced. Therefore, when the prediction mode is selected by the rate-distortion optimization method, the rate at which the skip mode is selected as the optimal prediction mode may be increased, and thus the compression performance of image encoding may be improved.

Since the error compensation according to the embodiment of the present invention used in the skip mode may be performed using only information of a previously encoded block, the decoder may also perform the same error compensation as the encoder without additional information transmitted from the encoder. have. Accordingly, since the encoder does not need to transmit additional additional information to the decoder for error compensation, the amount of information transmitted from the encoder to the decoder can be minimized.

In the above-described embodiment, the methods are described based on a flowchart as a series of steps or blocks, but the present invention is not limited to the order of steps, and any steps may occur in a different order or at the same time than the other steps described above. have. Also, one of ordinary skill in the art appreciates that the steps shown in the flowcharts are not exclusive, that other steps may be included, or that one or more steps in the flowcharts may be deleted without affecting the scope of the present invention. I can understand.

The above-described embodiments include examples of various aspects. While not all possible combinations may be described to represent the various aspects, one of ordinary skill in the art will recognize that other combinations are possible. Accordingly, the invention is intended to embrace all other replacements, modifications and variations that fall within the scope of the following claims.

Claims (17)

Deriving a pixel value of the prediction block for the current block;
Deriving an error compensation value for the current block; And
Deriving a pixel value of a final prediction block by using the pixel value of the prediction block and the error compensation value,
The error compensation value is a sample value of an error compensation block for compensating for an error between the current block and the reference block, and the reference block is a block in a reference image and includes information related to predictor values of pixels in the current block. Is the containing block,
The prediction of the current block is performed in an intra skip mode (SKIP mode). The method of claim 1, wherein the deriving pixel value of the prediction block for the current block comprises:
Deriving motion information on the current block using an already decoded block; And
And deriving a pixel value of the prediction block by using the derived motion information. The image decoding method of claim 2, wherein the already decoded block includes a neighboring block of the current block. The image decoding method of claim 2, wherein the already decoded block includes a neighboring block of the current block and a neighboring block of a colocated block in the reference image. The method according to claim 2, wherein in the step of deriving the pixel value of the prediction block using the derived motion information, the pixel value of the derived prediction block,
And a pixel value of a reference block indicated by the derived motion information. The method of claim 2, wherein in the case of deriving the pixel value of the prediction block by using the derived motion information when the reference block is two or more,
A pixel value of the prediction block is derived by a weighted sum of pixel values of the reference block, and the reference block is a block indicated by the derived motion information. The method of claim 1, wherein the derivation of an error compensation value for the current block comprises:
Deriving an error parameter for an error model of the current block; And
Deriving an error compensation value for the current block by using the error model and the derived error parameter. The method of claim 7, wherein the error model is a zero-order error model or a first-order error model. The method of claim 7, wherein the deriving the error parameter for the error model of the current block,
And an error parameter is derived using information included in a neighboring block of the current block and a block in the reference image. The method of claim 7, wherein when the reference block is two or more, in the step of deriving an error compensation value for the current block by using the error model and the derived error parameter,
The error compensation value is derived by a weighted sum of error block values, and the error block value is an error compensation value of the current block derived for each reference block. The method of claim 1, wherein in the deriving the pixel value of the final prediction block by using the pixel value of the prediction block and the error compensation value,
The error compensation value may be selectively used according to the information indicating whether the error compensation is applied, and the information indicating whether the error compensation is applied may be a slice header, a picture parameter set, or a sequence parameter set. sequence parameter set), which is information transmitted from an encoder to a decoder. Deriving a pixel value of the prediction block for the current block;
Deriving an error compensation value for the current block; And
Deriving a pixel value of a final prediction block by using the pixel value of the prediction block and the error compensation value;
The error compensation value is a sample value of an error compensation block for compensating an error between the current block and a reference block, and the reference block is a block in a reference image and includes a block containing information related to prediction values of pixels in the current block.
How to predict skip mode between screens. The method of claim 12, wherein the step of deriving the pixel value of the prediction block for the current block,
Deriving motion information on the current block using an already decoded block; And
And deriving a pixel value of the prediction block by using the derived motion information. The method of claim 12, wherein the deriving the error compensation value for the current block,
Deriving an error parameter for an error model of the current block; And
And deriving an error compensation value for the current block by using the error model and the derived error parameter. An entropy decoder configured to entropy decode a bitstream received from the decoder according to a probability distribution to generate residual block related information;
A prediction unit for deriving a pixel value of the prediction block for the current block and an error compensation value for the current block, and deriving a pixel value of the final prediction block using the pixel value of the prediction block and the error compensation value; And
An adder for generating a reconstruction block using the residual block and the last prediction block;
The error compensation value is a sample value of an error compensation block for compensating for an error between the current block and a reference block, and the reference block is a block in a reference image and includes a block containing information related to prediction values of pixels in the current block.
And the predictor performs prediction on the current block in an inter-screen skip mode. The method of claim 15, wherein the prediction unit
And a pixel value of the predicted block is derived using the derived motion information. The method of claim 15, wherein the prediction unit
And an error parameter for the error model of the current block, and an error compensation value for the current block using the error model and the derived error parameter.

Images