WO2011149292A2 - Method and apparatus for processing video signals - Google Patents

Abstract

The present invention relates to a method and an apparatus for encoding or decoding video signals. To increase accuracy of image prediction, the present invention can efficiently and adaptively determine a motion vector prediction method according to whether a present image is an image having more motion changes in a horizontal direction or an image having more motion changes in a vertical direction, or an image to be rotated or enlarged/reduced. Particularly, the present invention provides a method for adaptively determining a first component to be processed at first between x and y components of a motion vector of a present partition and a method for processing a second component as the remaining component of the motion vector using the first component of the motion vector processed in advance.

Description

Method and apparatus for processing video signal

The present invention relates to a method and apparatus for processing a video signal, and more particularly, to a video signal processing method and apparatus for encoding or decoding a video signal.

Compression coding refers to a series of signal processing techniques for transmitting digitized information through a communication line or for storing in a form suitable for a storage medium. The object of compression encoding includes objects such as voice, video, text, and the like. In particular, a technique of performing compression encoding on an image is called video image compression. Compression coding on a video signal is performed by removing redundant information in consideration of spatial correlation, temporal correlation, and stochastic correlation. However, with the recent development of various media and data transmission media, there is a need for a method and apparatus for processing video signals with higher efficiency.

In the process of processing a video signal, inter-screen prediction obtains a motion vector prediction value of a current partition using motion information of neighboring partitions, obtains a prediction value of the current partition through a motion vector obtained by adding the motion vector difference value, and then obtains a current partition. You can restore the pixel value of.

In this case, the higher the accuracy of the motion vector predicted value, the less information may be transmitted for decoding. According to the characteristics of the image, the image having a lot of horizontal motion change, the image having a lot of vertical motion change, rotated or enlarged / reduced There may be various kinds of images such as images.

In order to efficiently predict the motion vector according to various image characteristics, a method of adaptively obtaining a motion vector prediction value according to the characteristics of each image needs to be changed. In this case, there is a need for an efficient video signal processing method capable of increasing the accuracy of motion vector prediction while minimizing the amount of transmitted information.

The present invention has been devised to solve the above requirements, and in the motion vector prediction of the current partition, first processing any one component of the motion vector x component and y component as a first component, and first processing the first It is an object of the present invention to provide a method for treating the remaining second component by using the component.

In this case, an object of the present invention is to provide a method for obtaining a motion vector first component prediction value of a current partition using a motion vector first component of neighboring partitions.

Another object of the present invention is to provide a method for determining a first component to be processed first among the motion vector x component and y component of the current partition in order to adaptively increase the accuracy of image prediction.

In addition, the present invention is to improve the accuracy of motion vector prediction by obtaining the motion vector prediction value of the current partition in consideration of the partition shape of the current partition and the neighbor partition in the motion vector prediction of the partition on which the geometric partition is performed. Have.

In addition, the present invention provides an affine transformation using motion vectors of neighboring partitions to predict the motion vector of the current partition, thereby improving the accuracy of motion vector prediction for a particular motion situation.

According to the video signal processing method according to the present invention, by varying the method of adaptively obtaining the motion vector prediction value according to various image characteristics, it is possible to increase the accuracy of motion vector prediction and to reduce the amount of information transmitted for the reconstruction of the image signal. Will be.

That is, the motion vector prediction method can be determined efficiently and adaptively depending on whether the current video is a video with a lot of motion change in the horizontal direction, a video with a lot of motion change in the vertical direction, or a video that is rotated or enlarged / reduced. Can be.

In addition, according to an embodiment of the present invention, even in a neighboring partition that does not have the same reference picture as the current partition, motion vector prediction can be performed using the motion vector scaled value, thereby extending the candidate group for motion vector prediction. More accurate predictions are possible.

In addition, according to an embodiment of the present invention, even if the accuracy of motion vector prediction is increased by transmitting a decision of a first component to be processed first among the motion vector x component and y component through a flag value, signaling of the flag is not required. In a special situation, the decoder directly derives the flag without transmitting the flag value, thereby increasing the accuracy of motion vector prediction and minimizing the amount of transmission information.

1 is a schematic block diagram of a video signal encoder apparatus according to an embodiment of the present invention.

2 is a schematic block diagram of a video signal decoder device according to an embodiment of the present invention.

3 is a diagram illustrating an example of splitting a coding unit according to an embodiment of the present invention.

FIG. 4 is a diagram illustrating an embodiment of a method for hierarchically representing the partition structure of FIG. 3.

5 is a diagram illustrating a partitioning form into various sub-prediction units within a prediction unit according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating a form obtained by dividing a prediction unit geometrically.

7 illustrates a spatial neighbor partition of the current partition.

8 is a diagram illustrating a temporal neighboring partition of the current partition.

9 is a diagram illustrating a method of scaling a motion vector according to a temporal distance between reference pictures.

10 is a flowchart illustrating a process of obtaining a motion vector first component and a motion vector second component of a current partition according to an embodiment of the present invention.

11 is a flowchart illustrating a specific embodiment of obtaining a motion vector first component prediction value of a current partition according to an embodiment of the present invention.

12 is a flowchart illustrating a specific embodiment of obtaining a motion vector second component prediction value of a current partition according to an embodiment of the present invention.

13 and 14 show examples of geometrically divided partitions and their neighboring partitions.

15 shows a motion vector of a current partition and neighboring partitions with reduced motion.

FIG. 16 illustrates center coordinates of current and neighboring partitions for obtaining an affine motion vector and center coordinates moved through the motion vector.

According to an aspect of the present invention, there is provided a method of processing a video signal, the method including: obtaining a motion vector first component prediction value of a current partition using motion vector first components of neighboring partitions; Obtaining a motion vector first component of the current partition using the motion vector first component prediction value of the current partition; Obtaining a motion vector second component prediction value of the current partition based on the motion vector first component of the current partition; Obtaining a motion vector second component of the current partition using the motion vector second component prediction value of the current partition; Obtaining a prediction value of a current partition using a motion vector first component and a second component of the current partition; And reconstructing pixel values of the current partition by using the predicted value of the current partition, wherein the first component is a component to be first decoded among the x component and the y component of the motion vector. It is characterized in that the remaining one component after the decoding is performed after the first component.

In this case, the obtaining of the motion vector first component prediction value of the current partition may be obtained by using motion vector first components of neighboring partitions having the same reference picture as the current partition, but there is no neighboring partition having the same reference picture. The median value of the motion vector first components of the neighbor partitions may be obtained as the motion vector first component prediction value of the current partition.

In addition, when there is no neighbor partition having the same reference picture, the motion vector first component prediction value of the current partition may be obtained through the median of the scaled motion vector first component of the neighbor partitions.

The obtaining of the motion vector second component prediction value of the current partition may include: obtaining a neighbor partition having a motion vector first component whose minimum value of the difference with the motion vector first component of the current partition is minimum; A motion vector second component prediction value of the current partition may be obtained using the motion vector second component of the neighbor partition.

According to an embodiment of the present invention, before the obtaining of the motion vector first component prediction value of the current partition, the method may further include determining a first component of the motion vector x component and y component of the current partition to be decoded first. can do.

In this case, the determining of the first component may be obtained through the transmitted flag value.

In the determining of the first component, an average value of absolute values of the motion vector x component and y component of the previous frame may be obtained, and a component having a smaller size may be determined as the first component.

In the determining of the first component, an average value of absolute values of the motion vector x and y components of neighboring partitions of the current partition may be obtained, and a component having a smaller size may be determined as the first component.

In the determining of the first component, when there is a component having one valid value among the motion vector x component and y component of neighboring partitions of the current partition, the corresponding component may be determined as the first component.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, terms or words used in the present specification and claims should not be construed as having a conventional or dictionary meaning, and the inventors should properly explain the concept of terms in order to explain their invention in the best way. Based on the principle that can be defined, it should be interpreted as meaning and concept corresponding to the technical idea of the present invention. Therefore, the embodiments described in the specification and the drawings shown in the drawings are only the most preferred embodiment of the present invention and do not represent all of the technical idea of the present invention, various modifications that can be replaced at the time of the present application It should be understood that there may be equivalents and variations.

In the present invention, the following terms may be interpreted based on the following criteria, and terms not described may be interpreted according to the following meanings. Coding can be interpreted as encoding or decoding in some cases, and information is a term that includes values, parameters, coefficients, elements, and the like. May be interpreted otherwise, the present invention is not limited thereto. 'Unit' is used to mean a basic unit of image processing or a specific position of an image, and in some cases, may be used interchangeably with terms such as 'block', 'partition' or 'region'. Also, in the present specification, a unit may be used as a concept including a coding unit, a prediction unit, and a transform unit.

1 is a schematic block diagram of a video signal encoding apparatus according to an embodiment of the present invention. Referring to FIG. 1, the encoding apparatus 100 of the present invention is largely composed of a transformer 110, a quantizer 115, an inverse quantizer 120, an inverse transformer 125, a filter 130, and a predictor ( 150 and the entropy coding unit 160.

The converter 110 obtains a transform coefficient value by converting a pixel value of the input video signal. For example, Discrete Cosine Transform (DCT) or Wavelet Transform (DCT) may be used. In particular, the discrete cosine transform divides the input image signal into blocks having a predetermined size and performs the conversion. In the transform, the coding efficiency may vary depending on the distribution and the characteristics of the values in the transform domain.

The quantization unit 115 quantizes the transform coefficient value output from the transform unit 110. The inverse quantization unit 120 inverse quantizes the transform coefficient value, and the inverse transform unit 125 restores the original pixel value by using the inverse quantized transform coefficient value.

The filtering unit 130 performs a filtering operation for improving the quality of the reconstructed image. For example, a deblocking filter and an adaptive loop filter may be included. The filtered image is stored in the storage unit 156 for output or use as a reference image.

In order to improve coding efficiency, a method of predicting an image by using an already coded region and adding a residual value between the original image and the predicted image to a reconstructed image is used instead of coding the image signal as it is. . The intra predictor 152 performs intra prediction within the current image, and the inter predictor 154 predicts the current image using the reference image stored in the storage 156. The intra prediction unit 152 performs intra prediction from the reconstructed regions in the current image, and transmits the intra encoding information to the entropy coding unit 160. The inter predictor 154 may further include a motion compensator 162 and a motion estimator 164. The motion estimator 164 obtains a motion vector value of the current region by referring to the restored specific region. The motion estimator 164 transmits the position information (reference frame, motion vector, etc.) of the reference region to the entropy coding unit 160 so that the motion estimation unit 164 may be included in the bitstream. The motion compensator 162 performs inter-screen motion compensation by using the motion vector value transmitted from the motion estimator 164.

The entropy coding unit 160 entropy codes the quantized transform coefficients, inter picture encoding information, intra picture encoding information, and reference region information input from the inter prediction unit 154 to generate a video signal bitstream. The entropy coding unit 160 may use a variable length coding (VLC) method, arithmetic coding, or the like. The variable length coding (VLC) scheme converts input symbols into consecutive codewords, which may have a variable length. For example, frequently occurring symbols are represented by short codewords and infrequently occurring symbols by long codewords. As a variable length coding method, a context-based adaptive variable length coding (CAVLC) method may be used. Arithmetic coding converts consecutive data symbols into a single prime number, which can obtain the optimal fractional bits needed to represent each symbol. Context-based Adaptive Binary Arithmetic Code (CABAC) may be used as arithmetic coding.

2 is a schematic block diagram of a video signal decoding apparatus 200 according to an embodiment of the present invention. Referring to FIG. 2, the decoding apparatus 200 of the present invention largely includes an entropy decoding unit 210, an inverse quantization unit 220, an inverse transform unit 225, a filtering unit 230, and a prediction unit 250.

The entropy decoding unit 210 entropy decodes the video signal bitstream and extracts transform coefficients, motion vectors, and the like for each region. The inverse quantization unit 220 inverse quantizes the entropy decoded transform coefficient, and the inverse transform unit 225 restores the original pixel value by using the inverse quantized transform coefficient.

Meanwhile, the filtering unit 230 performs filtering on the image to improve the image quality. This may include a deblocking filter to reduce block distortion and / or an adaptive loop filter to remove distortion of the entire image. The filtered image is output or stored in the frame storage unit 256 for use as a reference image for the next frame.

In addition, the predictor 250 of the present invention includes an intra predictor 252 and an inter predictor 254, and includes a coding type decoded by the entropy decoder 210 described above, a transform coefficient for each region, The prediction image is reconstructed by using information such as a motion vector.

In this regard, the intra prediction unit 252 performs the intra prediction from the decoded samples in the current image.

The inter predictor 254 estimates a motion vector by using the reference image stored in the frame storage 256 and generates a predicted image. The inter predictor 254 may again include a motion compensator 262 and a motion estimator 264. The motion estimator 264 obtains a motion vector indicating a relationship between a current block and a reference block of a reference frame used for coding and transfers the motion vector to the motion compensator 262.

The predicted value output from the intra predictor 252 or the inter predictor 254 and the pixel value output from the inverse transform unit 225 are added to generate a reconstructed video frame.

Hereinafter, a method of dividing a coding unit, a prediction unit, etc. with reference to FIGS. 3 to 5 in the operations of the encoder device and the decoder device will be described.

A coding unit is a process such as intra / inter prediction, transform, quantization and / or entropy coding in the processing of the video signal described above. Means the basic unit for processing the image. The size of the coding unit used to code one image may not be constant. The coding unit may have a rectangular shape, and one coding unit may be further divided into several coding units.

3 illustrates an example of splitting a coding unit according to an embodiment of the present invention. For example, one coding unit having a size of 2N × 2N may be divided into four coding units having a size of N × N. The splitting of such coding units can be done recursively, and not all coding units need to be split in the same form. However, there may be a limitation on the maximum size 310 or the minimum size 320 of the coding unit for convenience in coding and processing.

For one coding unit, information indicating whether the corresponding coding unit is split may be stored. For example, assume that one coding unit can be divided into four square coding units as shown in FIG. 3. FIG. 4 illustrates an embodiment of a method for hierarchically representing a division structure of a coding unit illustrated in FIG. 3 using 0 and 1. FIG. Information indicating whether a coding unit is divided may be allocated to a value of '1' when the corresponding unit is divided and '0' when it is not divided. As shown in FIG. 4, when a flag value indicating whether to split is 1, a coding unit corresponding to a corresponding node is divided into 4 coding units, and when 0, a processing process for the coding unit is performed without being divided further. Can be.

The coding unit may not necessarily be divided into four forward regions. In this case, a code for a predetermined partitioning scheme may be mapped to the partitioning information. For example, if the information value is 1, the coding unit is divided into two horizontal rectangular sub-coding units, if 2, the coding unit is divided into two vertical rectangular sub-coding units, and 3 is four square sub-coding units. It can also be set to divide by. Such a method shows some embodiments of a division method, and does not limit the present invention.

The structure of the coding unit described above may be represented using a recursive tree structure. That is, a coding unit split into another coding unit with one picture or maximum size coding unit as a root has as many child nodes as the number of split coding units. Thus, coding units that are no longer split become leaf nodes. Assuming that only square division is possible for one coding unit, one coding unit may be divided into up to four other coding units, so the tree representing the coding unit may be in the form of a quad tree.

In the encoder, an optimal coding unit size may be selected according to the characteristics (eg, resolution) of the video image or in consideration of coding efficiency, and information about the coding unit may be included in the bitstream. For example, the size of the largest coding unit and the maximum depth of the tree can be defined. In the case of square division, since the height and width of the coding unit are half the height and width of the coding unit of the parent node, the minimum coding unit size can be obtained using the above information. Alternatively, the minimum coding unit size and the maximum depth of the tree may be defined in advance, and the maximum coding unit size may be derived and used if necessary. Since the size of the unit changes in a multiple of 2 in square division, the size of the actual coding unit is represented by a logarithm of the base of 2, thereby improving transmission efficiency.

The decoder may obtain information indicating whether the current coding unit is split. This information can be obtained only if the information is obtained (transmitted) under certain conditions, thereby increasing efficiency. For example, the condition that the current coding unit can be divided is information indicating whether the current coding unit size is divided in this case because the current coding unit size is smaller than the picture size and the current unit size is larger than the preset minimum coding unit size. Can be obtained.

If the information indicates that the coding unit is split, the size of the coding unit to be split is half of the current coding unit, and is split into four square coding units based on the current processing position. The above process can be repeated for each divided coding units.

Picture prediction (motion compensation) for coding is directed to coding units (i.e. leaf nodes of the coding unit tree) that are no longer divided. The basic unit for performing such prediction is called a prediction unit, a prediction block, or a partition. The prediction unit may be divided into various forms, and may be divided into symmetrical forms such as squares, rectangles, sub-prediction units of asymmetrical forms, or geometrical forms. For example, one prediction unit may not be split (2Nx2N) or may be divided into sub-prediction units of size NxN, 2NxN, Nx2N, etc. as shown in FIG. In addition, the possible division forms of the prediction unit may be defined differently in the intra coding unit and the inter coding unit. For example, in the intra coding unit, only division of 2Nx2N or NxN type is possible, and in the inter coding unit, the division of 2Nx2N, 2NxN, Nx2N, or NxN type may be configured. In this case, the bitstream may include information about whether the prediction unit is divided or in what form. Or this information may be derived from other information.

Partitioning may not necessarily be symmetrical. That is, the prediction unit may be divided into asymmetrically divided into rectangles of different sizes, and as shown in FIG. 6, geometric partitions may be generated in various shapes other than a rectangular shape. In this case, the prediction unit is divided into two partitions 610 and 620 through an arbitrary dividing line 600. The division information may be expressed as a distance ρ from the center of the dividing line, an angle θ between the repair line from the center O to the dividing line and the reference axis (for example, the x-axis). Such various types of prediction unit partitioning have an advantage of enabling more accurate prediction of various types of objects included in an image.

Hereinafter, as used herein, the term partition may be used as a replacement term for the prediction unit or sub-prediction unit, which is a basic unit for performing prediction. However, the present invention is not limited thereto.

The decoded portion of the current picture or other pictures containing the current partition may be used to reconstruct the current partition on which decoding is performed. Intra picture or I picture (slice) using only the current picture for reconstruction, i.e., performing only intra-picture prediction, predicting a picture (slice) using at most one motion vector and reference index to predict each partition A picture using a predictive picture or a P picture (slice), up to two motion vectors, and a reference index (slice) is called a bi-predictive picture or a B picture (slice).

The intra prediction unit performs intra prediction for predicting pixel values of the target partition from the reconstructed regions in the current picture. For example, the pixel value of the current partition can be predicted from the encoded pixels of the partitions located at the top, left, top left and / or top right with respect to the current partition.

The intra mode may be divided into vertical, horizontal, DC, and angular modes according to the direction and prediction method of the reference region where the reference pixels used for pixel value prediction are located. The vertical mode is used as a prediction value of the current partition with reference to the value of the vertically adjacent area of the target partition, and the horizontal mode is used as a prediction value with reference to the value of the horizontally adjacent area. In the DC mode, the average value of the reference regions is used as a prediction value. Meanwhile, in the angular mode, the reference region is located in an arbitrary direction and may indicate the corresponding direction by an angle between the current pixel and the reference pixel. For convenience, a predetermined angle and prediction mode number may be used, and the number of angles used may vary depending on the size of the target partition.

Several specific modes can be defined and used for these various prediction methods. The prediction mode may be transmitted as a value indicating the mode itself, but in order to increase transmission efficiency, a method of predicting a prediction mode value of the current partition may be used. In this case, the decoder may obtain the prediction mode of the current partition using information using whether the prediction value for the prediction mode is used as it is or how much the difference with the actual value is.

Meanwhile, the inter prediction unit performs inter prediction for predicting a pixel value of a target partition by using information of other reconstructed pictures other than the current picture. In this case, a picture used for prediction is referred to as a reference picture. Which reference region is used to predict the current partition in the inter prediction process may be indicated by using an index indicating a reference picture including the reference region, motion vector information, and the like.

The inter prediction may include forward direction prediction, backward direction prediction, and bi-prediction. Forward prediction is prediction using one reference picture displayed (or output) before the current picture in time, and backward prediction means prediction using one reference picture displayed (or output) after the current picture in time. For this purpose, one motion information (eg, a motion vector and a reference picture index) may be required. In the bi-prediction method, up to two reference regions may be used, and these two reference regions may exist in the same reference picture or may exist in different pictures, respectively. That is, in the pair prediction method, up to two motion information (eg, a motion vector and a reference picture index) may be used. The two motion vectors may have the same reference picture index or may have different reference picture indexes. . In this case, the reference pictures may be displayed (or output) before or after the current picture in time.

The motion information of the current partition may include motion vector information and a reference picture index. The motion vector information may include a motion vector, a motion vector prediction (mvp), a motion vector difference (mvd), or mean index information for specifying the motion vector prediction value. . The differential motion vector means a difference value between the motion vector and the motion vector prediction value.

The reference partition of the current partition can be obtained using the motion vector and the reference picture index. The reference partition exists in a reference picture having the reference picture index. In addition, a pixel value or an interpolated value of a partition specified by the motion vector may be used as a predictor of the current partition. That is, using motion information, motion compensation is performed to predict an image of a current partition from a previously decoded picture.

Meanwhile, a reference picture list may be configured of pictures used for inter prediction for the current picture. In the case of a B picture, two reference picture lists are required, and in the following, each of them is referred to as reference picture list 0 (or L0) and reference picture list 1 (or L1).

In order to reduce the amount of transmission associated with the motion vector, a method of obtaining a motion vector prediction value mvp by using motion information of previously coded partitions and transmitting only a difference value mvd thereof. The decoder obtains a motion vector prediction value of the current partition using motion information of other decoded partitions, and obtains a motion vector value for the current partition using the transmitted difference value. In obtaining the motion vector prediction value, a motion vector competition method may be used in which various motion vector candidate values are obtained by using motion information of already coded partitions, and one of them is obtained as the motion vector prediction value.

For the motion vector prediction, the motion vector prediction value of the current partition may be obtained from the motion vectors of neighboring partitions of the current partition. Such neighbor partitions may include spatial neighbor partitions and temporal neighbor partitions. A spatial neighbor partition refers to partitions adjacent to the current partition within the same picture as the current partition (see FIG. 7), and a temporal neighbor partition is a partition existing at a position co-located with the current partition in a picture different from the current partition. (See Fig. 8).

For example, referring to FIG. 7, the spatial neighbor partitions include partitions A ₀ , A ₁ ,..., A _nA neighboring the left side of the current partition, and neighboring partitions B ₀ at the top of the current partition. , B ₁ ,..., B _nB ) and / or partitions C, D, E adjacent to the diagonal of the current partition may be included. In this case, the size of each neighboring partition may be different.

As such, the motion vector prediction value of the current partition is obtained from the motion vectors of the spatial neighboring partition and / or the temporal neighboring partition of the current partition, and the motion vector value for the current partition can be obtained by adding this to the transmitted difference value. In this case, when the current picture (or slice) is a P picture (slice), a motion vector prediction value and a motion vector for a prediction direction (reference picture list) of either L0 or L1 may be obtained, and the current picture (or slice) may be obtained. If) is a B picture (slice), the motion vector prediction value and the motion vector for the prediction directions of both L0 and L1 can be obtained.

In this case, even when the reference picture of the neighbor partition used to obtain the motion vector prediction value of the current partition is different from the reference picture of the current partition, if the motion vector of the neighbor partition is used as it is, there is a limit to accurate motion vector prediction. Therefore, in order to perform more accurate motion vector prediction, a motion vector scaling technique can be used.

Motion vector scaling refers to a technique of scaling up or scaling down a motion vector according to a temporal distance between reference pictures and a current picture. 9 illustrates a method of performing motion vector scaling using a temporal distance between pictures. When the motion vector in the L0 direction of a neighboring partition is referred to as mvL0 and the reference picture of mvL0 is referred to as ref1, mvL0N_scaled which scales mvL0 with respect to ref0 which is a reference picture of the current partition can be expressed as follows.

mvL0N_scaled = mvL0 * (tb / td)

Here, td denotes a temporal distance between the current picture and ref1, and tb denotes a temporal distance between the current picture and ref0. The temporal distance between each picture may be obtained through a difference of a picture order count (POC) value between each picture.

According to an embodiment of the present invention, in order to increase the efficiency of the motion vector prediction, one component (first component) of the motion vector x component and y component of the current partition is first obtained, and the remaining other components (first component) are obtained using the first component. Can be used to predict the value of two components). 10 is a diagram illustrating such a motion vector prediction value and a motion vector acquisition process according to an embodiment of the present invention.

Referring to FIG. 10, when the first decoding component of the motion vector x component and y component of the current partition is performed as the first component, and the other component performing decoding after the first component is referred to as the second component First, a motion vector first component prediction value of the current partition may be obtained (S1020). At this time, before the step S1020, the step of determining the first component of the motion vector x component and y component of the current partition (S1010), that is, the step of determining the component to be decoded first of the x component and the y component to be added This will be described later. In operation S1020, the motion vector first component prediction value of the current partition may be obtained using the motion vector first components of neighboring partitions of the current partition.

When the motion vector first component prediction value of the current partition is obtained through the operation S1020, the motion vector first component of the current partition may be obtained by adding the first motion vector difference value (first component) to the received motion vector (S1030). Next, in the present invention, based on the motion vector first component of the current partition, a motion vector second component prediction value of the current partition can be obtained (S1040). In this case, the present invention finds a neighbor partition having a motion vector first component value most similar to the motion vector first component of the current partition among the neighbor partitions, and uses the motion vector second component value of the partition to determine the motion of the current partition. A vector second component prediction value may be obtained.

When the motion vector second component prediction value of the current partition is obtained through operation S1040, the motion vector second component of the current partition may be obtained by adding the motion vector difference value (second component) to the transmitted motion vector (S1050). A prediction value of the current partition may be obtained using the motion vector first component and the second component of the current partition obtained through the above process (S1060), and the pixel value of the current partition may be restored using this.

11 and 12 illustrate a process of obtaining a motion vector first component prediction value and a second component prediction value of a current partition in detail according to an embodiment of the present invention. Hereinafter, a case in which a motion vector y component is a first component and an x component is a second component will be described with reference to FIGS. 11 and 12. According to an exemplary embodiment of the present invention, the first component may be substituted with the motion vector x component and the second component with the motion vector y component.

First, steps S1110 to S1160 of FIG. 11 specify an example of a process of obtaining a motion vector y component prediction value of the current partition (S1020). As shown, a motion vector y component (mvX [1]) of neighboring partitions having the same reference picture as the current partition may be added to the y component list listMvCompY (S1110). For this purpose, it is necessary to first decode the reference picture index value of the current partition. Next, the number of candidate motion vector y component predicted values numCompY of the current partition is obtained through the number of elements included in the y component list listMvCompY (S1112).

According to the present invention, the motion vector y component predicted value mvPred [1] of the current partition can be obtained according to the value of the y component predicted value candidate number numCompY in the following manner.

i) if numCompY = 0. If there is no neighbor partition having the same reference picture as the current partition (S1120), the median value of the motion vector y components mvX [1] of the neighbor partitions may be set as the motion vector y component prediction value mvPred [1] of the current partition. There is (S1122). In this case, for more accurate motion vector prediction, the median value of the scaled motion vector y component of the neighbor partitions may be set to mvPred [1]. That is, the motion vector y components of neighboring partitions are scaled with respect to the reference picture of the current partition, and the median value of the scaled motion vector y components is assigned to mvPred [1]. Meanwhile, according to another embodiment of the present invention, when numCompY = 0, the mvPred [1] may be set to 0.

ii) if numCompY = 1. When there is one neighbor partition having the same reference picture as the current partition (S1130), the motion vector y component of the neighbor partition may be set as the motion vector y component prediction value mvPred [1] of the current partition (S1132). That is, listMvCompY [0], the only element added to the y component list, may be assigned as the value of mvPred [1].

iii) when numCompY = 2. If there are two neighbor partitions having the same reference picture as the current partition (S1140), the average of the motion vector y components (listMvCompY [0], listMvCompY [1]) of the two neighboring partitions is the motion vector y component of the current partition. The prediction value mvPred [1] can be set (S1142). Meanwhile, according to another embodiment of the present invention, when numCompY = 2, the motion vector y component of the partition having the larger size among two neighboring partitions included in the y component list may be set to mvPred [1]. If the sizes of the two neighboring partitions are the same, the average value of the motion vector y component may be set to mvPred [1] as in step S1142.

iv) in other cases. When there are three or more neighbor partitions having the same reference picture as the current partition, the median value of the motion vector y component of the corresponding neighbor partitions may be set as the motion vector y component prediction value mvPred [1] of the current partition. At this time, the elements included in the y component list listMvCompY are sorted in ascending (or descending) order (S1150), and the motion vectors y of neighboring partitions are obtained by obtaining the elements of the center index of the sorted y component list listMvCompY. The median value of the components can be obtained (S1160).

The motion vector y component prediction value mvPred [1] of the current partition obtained through the processes of S1110 to S1160 is summed with the transmitted motion vector difference value y component mvDiff [1] to add the motion vector y component of the current partition ( currMv [1]) can be obtained (S1030).

Meanwhile, in operations S1210 to S1270 of FIG. 12, a process of obtaining a motion vector x component prediction value mvPred [0] of the current partition based on the motion vector y component (currMv [1]) of the current partition (S1040). An example is given. First, in order to obtain the motion vector x component prediction value mvPred [0] of the current partition, the motion vector x component (mvX [0]) of neighboring partitions having the same reference picture as the current partition is added to the motion vector list listMv. Can be added (S1210).

Next, the number numMv of elements included in the motion vector list listMv is obtained (S1212). In this case, when the numMv is 0 (S1220), that is, when there is no neighbor partition having the same reference picture as the current partition, the median value of the motion vector x components (mvX [0]) of the neighbor partitions is determined. The predicted value mvPred [0] can be set (S1222). As described with reference to FIG. 11, the median value of the scaled motion vector x component of neighboring partitions may be set to mvPred [0] for more accurate motion vector prediction. That is, the motion vector x components of neighboring partitions are scaled with respect to the reference picture of the current partition, and the median value of the scaled motion vector x components is assigned to mvPred [0]. Meanwhile, according to another embodiment of the present invention, when numMV = 0, the mvPred [0] may be set to 0.

On the other hand, when there is at least one neighbor partition having the same reference picture as the current partition (numMv is not 0), based on the motion vector y component (currMv [1]) of the current partition according to an embodiment of the present invention. A motion vector x component prediction value mvPred [0] of the current partition may be obtained. That is, the neighbor partition having the motion vector y component value most similar to the motion vector y component (currMv [1]) of the current partition is searched, and the motion vector x component of the neighbor partition is predicted by the motion vector x component prediction value (mvPred) of the current partition. [0]).

In this case, two or more neighbor partitions that may be selected as the motion vector x component prediction value (mvPred [0]) of the current partition may be obtained. For this, the motion vector x component prediction value of the current partition may be obtained by the following process. can do. First, among the motion vector y components of the neighbor partition added to the motion vector list listMv, the element listMv [i whose absolute value of the difference from the motion vector y component currMv [1] of the current partition is minimum ] [1]) is set to compY (S1230). Next, among the elements listMv [i] [1] included in the motion vector list listMv, the motion vector x component listMv [i] [0] of neighboring partitions having the same value as the compY is set. It adds to x component list listMvCompX (S1232). Through the number of elements included in the x component list listMvCompX, a motion vector x component prediction value candidate number numCompX of the current partition may be obtained (S1234), and according to the value of the x component prediction value candidate number numCompX. The motion vector x component prediction value (mvPred [0]) of the current partition can be obtained by the following method.

i) when numCompX = 1. When there is one neighbor partition having a motion vector y component of which the absolute value of the difference from the motion vector y component of the current partition is the minimum (S1240), the motion vector x component of the current partition is estimated as the motion vector x component of the current partition. (mvPred [0]) can be set (S1242). That is, listMvCompX [0], the only element added to the x component list, may be assigned as the value of mvPred [0].

ii) when numCompX = 2. If there are two neighbor partitions having the motion vector y component of which the absolute value of the difference with the motion vector y component of the current partition is the minimum (S1250), the motion vector x component of the corresponding two neighbor partitions (listMvCompX [0], An average of listMvCompX [1]) may be set as the motion vector x component predicted value mvPred [0] of the current partition (S1252). Meanwhile, according to another embodiment of the present invention, when numCompX = 2, the motion vector x component of a partition having a larger size among two neighboring partitions included in the x component list may be set to mvPred [0]. If the sizes of the two neighboring partitions are the same, the average value of the motion vector x component may be set to mvPred [0] in step S1252.

iii) in other cases. If there are three or more neighbor partitions whose motion vector y component is the absolute value of the difference from the motion vector y component of the current partition, the median value of the motion vector x component of the corresponding partitions is set to the motion vector x component of the current partition. The prediction value may be set to mvPred [0]. In this case, the elements included in the x component list listMvCompX are sorted in ascending (or descending order) (S1260), and the motion elements of the neighboring partitions are obtained by obtaining the center elements of the sorted list of the x component listMvCompX. The median value may be obtained (S1270).

The motion vector x component prediction value mvPred [0] of the current partition obtained through the processes of S1210 to S1270 is summed with the transmitted motion vector difference value x component (mvDiff [0]) to obtain the motion vector x component of the current partition ( currMv [0]) can be obtained (S1050).

In the present invention, in order to perform more accurate motion vector prediction according to the characteristics of the image, the order of the coded motion vector x component and y component may be adaptively determined. Depending on the characteristics of the image, the x and y components of the motion vector may be different from each other. When coding components that are difficult to accurately predict, more information to process the components has better coding efficiency. Therefore, processing the components that can be predicted more accurately first and processing the remaining components using the same may have a good effect on the coding performance.

To this end, in the present invention, as shown in FIG. 10, the method may further include a step (S1010) of determining a first component from which the first decoding of the motion vector x component and y component of the current partition is performed. The determining step S1010 may use a signaling method for acquiring the first component through a transmitted flag value, or may use a method of deriving the first component from a decoder side without signaling.

In case of using the signaling method, signaling overhead exists, but there is an advantage of accurately indicating a better component in terms of RD (Rate Distortion) -cost. The signaling comprises i) at the Sequence Parameter Set (SPS) step, ii) at the Picture Parameter Set (PPS) step, iii) at the slice header, iv) at the largest Coding Tree Block. , Largest CTB) step, v) prediction unit (PU) step, or vi) partition (Motion partition) step.

Meanwhile, in order to reduce signaling overhead, a method of deciding which component to determine as the first component may be used. According to an embodiment of the present invention, the derivation method may be performed in units of frames (or pictures) or in units of partitions.

First, when deriving the first component on a frame basis, statistical information of motion vectors of a previously coded frame may be used. For example, if the previously decoded image has more motion vector changes in the horizontal direction, the current frame first decodes the motion vector y component, and vice versa, the current frame decodes the motion vector x component first. can do. According to an embodiment of the present invention, the first component may be derived in frame units through the following method.

i) The average magnitude of the absolute value of each of the x component and the y component of the motion vectors of the previously coded frame is obtained, and the component having the smaller size is determined as the first component.

ii) Obtain the variance of each of the x and y components of the motion vectors of the previously coded frame, and determine the component having the smaller size as the first component.

iii) The average magnitude of the absolute value of each of the x component and the y component of the motion vector difference values of the previously coded frame is obtained, and the component having the smaller size is determined as the first component.

iv) the method of i) to iii) is combined.

Next, when the first component is derived in a partition unit, statistical information of motion vectors of neighboring partitions may be used. For example, if the motion vectors of neighboring partitions have more changes in the horizontal direction, the current partition may first decode the motion vector y component, and vice versa, the current partition may decode the motion vector x component first. have. According to an embodiment of the present invention, the first component may be derived in a partition unit through the following method.

i) The average magnitude of the absolute value of each of the x and y components of the motion vectors of neighboring partitions is obtained, and a component having the smaller size is determined as the first component.

ii) obtain the variance of each of the x and y components of the motion vectors of the neighbor partitions, and determine a component having a smaller size as the first component.

iii) The average magnitude of the absolute value of each of the x component and the y component of the motion vector difference values of neighboring partitions is obtained, and the component having the smaller size is determined as the first component.

iv) Analyze the residual information of the current partitions to measure the magnitude of the horizontal frequency and the vertical frequency, and determine a component having the smaller size as the first component (or vice versa).

v) If the width is longer than the length according to the shape of the current partition, the x component is determined as the first component, otherwise the y component is determined as the first component (or vice versa).

vi) combine the methods of i) to v) above.

According to another embodiment of the present invention, the signaling method and the derivation method may be used in combination to determine the first component of the current partition. That is, a flag for determining a first component for each prediction direction L0 / L1 may be transmitted in partition units, but the first component may be directly derived without transmitting the flag for a partition having a specific condition. . According to an embodiment of the present invention, the flag may not be transmitted when any one of the following conditions is met.

i) One or zero motion vectors of a valid neighboring partition to be used for motion vector prediction of the current partition.

If there are one or zero motion vectors of a valid neighboring partition, the predicted value of the component to be processed later is fixed, whichever of the motion vector x component and y component predictive value of the current partition is processed first. In this case, therefore, it is possible to decide to code any component first without having to transmit a separate flag. Here, the neighbor partition may include both a spatial neighbor partition and a temporal neighbor partition. The motion vector of a valid neighbor partition to be used for motion vector prediction of the current partition may include only the motion vector of a neighbor partition having the same reference picture as the current partition, or the motion vector of a neighbor partition having another reference picture is included together. You can also If a motion vector of a neighbor partition having a reference picture different from the current partition is included, the scaled motion vector may be included in consideration of the temporal distance between the reference pictures as described above. In addition, even when using a motion vector of a temporal neighboring partition, a scaled motion vector may be used in consideration of a temporal distance between reference pictures.

ii) When there is a component having one valid value among the x and y components of the motion vector of the valid neighbor partition to be used for motion vector prediction of the current partition.

When later processing a component having one valid value, the efficiency of motion vector prediction is reduced because the motion vector prediction value of the component to be processed later is fixed regardless of the value of the component to be processed first. Therefore, in this case, the component having one valid value can be coded first without having to transmit a separate flag. For example, if the motion vector predictor candidate list of the current partition is {(5, -1), (5, 3), (5, 7)}, the x component of the motion vector predictor candidate is {5}, and the motion The y component of the vector predictor candidate is {-1, 3, 7}. At this time, by processing the x component having one valid value first, it can be used to increase the accuracy of the motion vector y component prediction of the current partition.

According to another embodiment of the present invention, a method of processing the y component of the motion vector predicted value first, but processing the x component first for a partition in a specific condition may be used. That is, when there is only one value of the motion vector x component of the valid neighbor partition used for motion vector prediction of the current partition, the x component may be coded first.

Meanwhile, in the present invention, motion vector prediction may be performed on geometrically divided partitions as shown in FIG. 6. In this case, a motion vector prediction value of the current partition may be obtained by using motion vectors of neighboring partitions adjacent to each partition.

For example, as shown in FIG. 13, the topmost partition A of the left neighboring partitions of the current coding unit, the leftmost partition B of the top neighboring partitions of the current coding unit, and the neighbors of the upper right corner of the current coding unit. When the partition C is adjacent to the current partition G1, the median of the motion vectors of the partitions A, B, and C may be used as the motion vector prediction value of the current partition G1. If this is not satisfied, neighbor partitions A, B, E (the rightmost partition among the top neighboring partitions of the current coding unit), F (the bottommost partition among the left neighboring partitions of the current coding unit), and C A partition adjacent to the current partition G1 may be searched to use the motion vector of the neighboring partition as the motion vector prediction value of the current partition G1. If there is no neighbor partition adjacent to the current partition, the motion vector prediction value of the current partition may be set to (0, 0).

Meanwhile, according to another embodiment of the present invention, if there is a neighbor partition having the geometric partition shape having the same shape as the current partition in consideration of the angle θ of the split line, the motion vector of the neighbor partition is used as the motion vector prediction value of the current partition. Can be used.

Further, according to another embodiment of the present invention, when there is no neighbor partition adjacent to the current partition on which the geometric division is performed, the motion vector of another partition in the same coding unit may be used as the motion vector prediction value of the current partition. That is, in FIG. 12, the partition G2 in which the geometric division is performed does not have adjacent neighbor partitions, but may obtain a motion vector of G1, which is another adjacent partition in the same coding unit, and use it as a motion vector prediction value.

In predicting the motion vector of the current partition, as described above, the motion vector prediction of the current partition is performed by using the median of the motion vectors of the neighboring partitions or the motion vector of the selected neighboring partition. As the motion vector difference is reduced, the amount of data to be transmitted is reduced. However, in the actual image sequence, various motions such as rotation and zooming are present, and in such an image, the possibility that the current partition has the same motion vector as the neighboring partition becomes small.

For example, FIG. 15 illustrates a situation in which peripheral partition areas including partition X are reduced around partition X. The motion vector MV of the current partition is one of the motion vectors MVa, MVb, and MVc of neighboring partitions. It may not have the same value as anything. In this case, when the motion vector MVb of partition B, which is the median motion vector of neighboring partitions, is used as the motion vector prediction value MVp of the current partition (or when the motion vector of any of partitions A, B, and C is used). MVd, which is a difference from the actual motion vector MV, should be transmitted. Therefore, a method for reducing the motion vector difference value MVd is needed.

According to an embodiment of the present invention, affine transformation may be performed using the motion vectors of neighboring partitions on the assumption that the current partition X has a motion vector similar to neighboring partitions A, B, and C. The affine transformation vector thus obtained can more accurately predict the operation of the current partition X, etc., thereby reducing the motion vector difference value.

A method of obtaining an affine transform vector from motion vectors of neighboring partitions according to an embodiment of the present invention will be described with reference to FIG. 16. The origin (0, 0) for coordinate calculation is the center of the current partition X, and the coordinates of the centers of neighboring partitions A, B, and C are (xa, ya), (xb, yb), and (xc, yc), respectively. do. In addition, the point where the center of each neighboring partition is moved by each motion vector MVa, MVb, MVc is called (xa ', ya'), (xb ', yb'), (xc ', yc'). .

In this case, the affine transformation vector A obtained from the motions of the three neighboring partitions A, B, and C is as follows.

Equation 1

The motion vector prediction value MVp of the current partition can be obtained using the affine transform vector A as follows.

MVp = (p3, p6)

Meanwhile, according to another embodiment of the present invention, a parametric transform vector is calculated from the decoded neighbor partitions and the motion vector of the current partition, and motion compensation of the current partition is performed using the same. Can be. In this case, the affine transformation model of the aforementioned method may be used as the model of the parametric transformation vector.

Referring to FIG. 16, the origin (0, 0) for the coordinate calculation is the center of the current partition X, and the coordinates of the centers of the neighbor partitions A, B, C and the current partition X are (xa, ya), (xb, yb), (xc, yc) and (xd, yd). Also, the centers of the neighboring partitions A, B, C and the current partition X are moved by the respective motion vectors MVa, MVb, MVc, and MV (xa ', ya'), (xb ', yb'). ), (xc ', yc') and (xd ', yd').

In this case, the affine transformation vector A 'obtained from the motions of the three neighboring partitions A, B, and C and the current partition X is as follows.

Equation 2

here,

Motion compensation for the current partition may be performed using the affine transform vector A '. That is, when the coordinate of any point in the current partition X is (x, y), the coordinate (x ', y') in the reference picture may be calculated as follows.

x '= p1 * x + p2 * y + p3

y '= p4 * x + p5 * y + p6

At this time, since the (x ', y') is calculated in real units, when the reference picture is interpolated in units of 1/4 pixels, the coordinate value (x '', y 'is rounded in units of 1/4) ') Can be obtained, and the pixel value of (x' ', y' ') of the corresponding reference picture can be allocated to the pixel value of the position (x, y) of the current partition.

According to another embodiment of the present invention, the motion vector prediction value MVp of the current partition obtained by using the affine transform vector A may be added as one of candidates used for existing motion vector competition.

The embodiments described above are the components and features of the present invention are combined in a predetermined form. Each component or feature should be considered optional unless stated otherwise. Each component or feature may be embodied in a form that is not combined with other components or features. It is also possible to combine some of the components and / or features to form an embodiment of the invention. The order of the operations described in the embodiments of the present invention may be changed. Some components or features of one embodiment may be included in another embodiment or may be replaced with corresponding components or features of another embodiment.

The decoding / encoding method to which the present invention is applied may be stored in a computer-readable recording medium that is produced as a program for execution on a computer, and the computer-readable recording medium may also have multimedia data having a data structure according to the present invention. Can be stored in. The computer readable recording medium includes all kinds of storage devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. In addition, the bitstream generated by the encoding method may be stored in a computer-readable recording medium or transmitted using a wired / wireless communication network.

Embodiments according to the present invention may be implemented by various means, for example, hardware, firmware, software, or a combination thereof. According to a hardware implementation, embodiments described herein include application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), and the like. It may be implemented using at least one of processors, controllers, micro-controllers, microprocessors, and electrical partitions for performing other functions. The described embodiments may be implemented by the controller itself.

According to the software implementation, embodiments such as the procedures and functions described herein may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein. Software code may be implemented in software applications written in a suitable programming language. The software code may be stored in a memory and executed by a controller.

As described above, although the present invention has been described by way of limited embodiments and drawings, the present invention is not limited thereto and is intended by those skilled in the art to which the present invention pertains. Of course, various modifications and variations are possible within the scope of the claims to be described.

The present invention can be applied to encoding or decoding a video signal.

Claims (9)

Obtaining a motion vector first component prediction value of the current partition using the motion vector first component of the neighbor partitions; Obtaining a motion vector first component of the current partition using the motion vector first component prediction value of the current partition; Obtaining a motion vector second component prediction value of the current partition based on the motion vector first component of the current partition; Obtaining a motion vector second component of the current partition using the motion vector second component prediction value of the current partition; Obtaining a prediction value of a current partition using a motion vector first component and a second component of the current partition; And Restoring pixel values of the current partition by using the predicted value of the current partition, The first component is a component in which decoding is first performed among the x component and the y component of the motion vector, and the second component is the other component in which decoding is performed after the first component. Way. The method of claim 1, Obtaining a motion vector first component prediction value of the current partition, Obtained using the motion vector first component of neighboring partitions having the same reference picture as the current partition, but if there are no neighbor partitions having the same reference picture, the median value of the motion vector first component of the neighboring partitions is determined as the motion vector of the current partition. A video signal processing method characterized in that it is obtained by a one-component prediction value. The method of claim 2, If there are no neighbor partitions with the same reference picture, And the motion vector first component prediction value of the current partition is obtained through a median of the scaled motion vector first component of the neighboring partitions. The method of claim 1, Obtaining a motion vector second component prediction value of the current partition, Obtain a neighbor partition having a motion vector first component of which the absolute value of the difference with the motion vector first component of the current partition is minimum, and use the motion vector second component of the neighbor partition to obtain a motion vector second of the current partition. A method of processing a video signal, characterized by obtaining a component prediction value. The method of claim 1, Before obtaining the motion vector first component prediction value of the current partition, And determining a first component of the motion vector x component and y component of the current partition to be decoded first. The method of claim 5, Determining the first component, A method of processing a video signal, characterized in that obtained through the transmitted flag value. The method of claim 5, Determining the first component, A method of processing a video signal, characterized in that the absolute value of the motion vector x component and y component of the previous frame is obtained, and the component having the smaller size is determined as the first component. The method of claim 5, Determining the first component, And obtaining an average of absolute values of motion vector x and y components of neighboring partitions of the current partition, and determining a component having a smaller size as a first component. The method of claim 5, Determining the first component, And when there is a component having one valid value among the motion vector x component and y component of neighboring partitions of the current partition, determining the corresponding component as the first component.

Images