WO2012147306A1

WO2012147306A1 - Moving picture encoding device, moving picture encoding method, moving picture encoding program, moving picture decoding device, moving picture decoding method, and moving picture decoding program

Info

Publication number: WO2012147306A1
Application number: PCT/JP2012/002649
Authority: WO
Inventors: 福島　茂; 西谷　勝義; 上田　基晴; 徹熊倉
Original assignee: 株式会社Ｊｖｃケンウッド
Priority date: 2011-04-28
Filing date: 2012-04-17
Publication date: 2012-11-01

Abstract

A reference destination specifying unit (113) specifies, from among a plurality of already encoded adjacent blocks adjacent to a block to be encoded, any one of the adjacent blocks determined based on the adjacent condition between the block to be encoded and the adjacent blocks as a reference destination block used for referring to motion information. When, from among an intra-image prediction, an inter-image prediction using the motion vector of the block to be encoded, and another inter-image prediction using the motion vector of an adjacent block, the another inter-image prediction is selected as a prediction method, a prediction selecting unit (112) generates, in place of the motion information, reference destination valid information indicating whether the adjacent block referred to by the selected prediction method and the reference destination block specified by the reference destination specifying unit (113) match each other or not and information indicating the selected prediction method.

Description

Moving picture encoding apparatus, moving picture encoding method, moving picture encoding program, moving picture decoding apparatus, moving picture decoding method, and moving picture decoding program

The present invention relates to a moving picture encoding and decoding technique, and more particularly to a moving picture encoding and decoding technique using motion compensation prediction.

In the moving picture coding method that divides a picture into rectangular blocks, represented by MPEG (Moving Picture Experts 、 Group), and performs motion estimation and compensation between blocks, the coding amount of the motion vector generated in each block In order to reduce the motion vector, a prediction process is performed on the motion vector. In MPEG-2, the motion vector detected in units of macroblocks is taken as a difference from the motion vector of the macroblock encoded immediately before, and the amount of code is reduced by encoding the difference vector. Yes.

MPEG-4 AVC / H. H.264 uses the fact that motion vectors have a strong correlation with the motion vectors of neighboring blocks, and performs prediction from neighboring blocks and encodes the difference vector to reduce the amount of code. . Specifically, the median is calculated from the motion vectors of adjacent blocks on the left, top and top right of the block to be processed, and motion vector prediction is realized by taking the difference from the median.

Although the amount of code of the motion vector is reduced by this prediction, other motion information is encoded for each block to be processed, so even if it has the same motion information as the adjacent neighboring blocks, it overlaps. Therefore, there is a problem that efficient encoding has not been achieved. In response to this problem, the motion information of the block to be processed and the neighboring already-encoded neighboring blocks are the same in the standard work of moving picture coding in Patent Document 1 and recent ISO / IEC and ITU-T. If so, the block to be processed does not encode its own motion information, uses the motion information of the adjacent block for encoding, and encodes additional information specifying the adjacent block having the motion information to be referenced. Thus, attempts have been made to reduce the amount of code of motion information. Such a method is called “merge” and has attracted attention as a method for reducing the code amount of motion information.

JP-A-10-276439

As an example of the merging method described above, there is a model that refers to motion information of a block adjacent to the left that has already been encoded or a block immediately adjacent to the block to be processed. In this case, a 1-bit flag (whether to use the motion information of the block to be processed itself detected by motion vector detection or to use the motion information of an adjacent block by the merge method as additional information for each block) In the following, a 1-bit flag (hereinafter referred to as merge_lu_flag) that defines whether to refer to the left or right of neighboring blocks when the merge method is applied is defined. Entropy coding using arithmetic codes is performed on these flags. In general, in entropy coding, a short code is assigned to data having a high appearance rate, and a long code is assigned to data having a low appearance rate, and data compression is performed to reduce the code amount. However, although the flag merge_lu_flag depends on the texture in the screen, since a code is assigned in advance to select the left or top as the reference destination adjacent block, even if the occurrence frequency of the reference destination selection is biased Therefore, there is a problem that the code amount cannot be sufficiently reduced because the encoding is performed with equal probability.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a technique for improving the coding efficiency by reducing the code amount of a flag representing a reference destination of motion information.

In order to solve the above-described problem, a moving image encoding device according to an aspect of the present invention is a moving image encoding device that encodes the moving image using a motion vector in units of blocks obtained by dividing each picture of the moving image. Any one of the adjacent blocks determined based on an adjacency condition between the encoding target block and the adjacent block is moved from among a plurality of encoded adjacent blocks adjacent to the encoding target block. A reference destination designating unit (113) that designates a reference block for referring to information, intra-picture prediction, inter-picture prediction using a motion vector of an encoding target block, and inter-picture using a motion vector of an adjacent block When the inter-picture prediction using the motion vector of the adjacent block is selected as the prediction method from the prediction, it is referred to by the selected prediction method. A prediction selection unit that generates, instead of motion information, reference destination valid information indicating whether or not a contact block matches the reference destination block specified by the reference destination specifying unit, and information indicating the selected prediction method (112).

Another aspect of the present invention is a video encoding method. This method is a moving image encoding method for encoding the moving image using a motion vector in units of blocks obtained by dividing each picture of the moving image, and a plurality of encoded adjacent regions adjacent to the encoding target block. A reference destination designating step of designating any one adjacent block determined based on an adjacency condition between the encoding target block and the adjacent block as a reference destination block for referring to motion information from among the blocks; Intra-picture prediction, inter-picture prediction using the motion vector of the encoding target block, and inter-picture prediction using the motion vector of the adjacent block, inter-picture prediction using the motion vector of the adjacent block as a prediction method Is selected, the neighboring block referred to by the selected prediction method and the reference designated by the reference destination designation step. And a prediction selection step of the previous block is produced on behalf of the referenced enable information and motion information indicating the selected prediction method information indicating whether match.

A moving picture decoding apparatus according to an aspect of the present invention is a moving picture decoding apparatus that decodes a bitstream in which the moving picture is encoded using a motion vector in units of blocks obtained by dividing each picture of the moving picture. A variable for decoding the information indicating the prediction method and the reference destination valid information from the bitstream including information indicating the method and reference destination valid information indicating the validity of the adjacent block referred to by the selected prediction method Using the long decoding unit (201), intra-picture prediction, inter-picture prediction using a motion vector of a decoding target block, and one motion vector of a plurality of decoded adjacent blocks adjacent to the decoding target block A prediction selection unit (209) that selects one of prediction methods based on information indicating the prediction method from among the inter-image predictions; and the prediction selection Therefore, when inter-picture prediction using a motion vector of an adjacent block is selected as a prediction method, any one adjacent block determined based on an adjacent condition between the decoding target block and the adjacent block is moved. A motion information selection unit (210) that determines whether it can be designated as a reference block for referring to information based on the reference destination valid information.

Another aspect of the present invention is a video decoding method. This method is a moving image decoding method for decoding a bitstream in which the moving image is encoded using a motion vector in units of blocks obtained by dividing each picture of the moving image, and is selected as information indicating a prediction method A variable length decoding step for decoding the information indicating the prediction method and the reference destination valid information from the bitstream including the reference destination valid information indicating the validity of the adjacent block referred to by the prediction method, and intra prediction The prediction method includes: inter-picture prediction using a motion vector of a decoding target block; and inter-picture prediction using any one of a plurality of decoded adjacent blocks adjacent to the decoding target block. A prediction selection step of selecting one of the prediction methods based on the information indicated, and the prediction selection step, When inter-picture prediction using a motion vector of a block is selected, a reference for referring to motion information for any one adjacent block determined based on an adjacent condition between the decoding target block and the adjacent block A motion information selection step of determining whether or not it can be designated as a destination block based on the reference destination valid information.

It should be noted that an arbitrary combination of the above-described components and a conversion of the expression of the present invention between a method, an apparatus, a system, a recording medium, a computer program, and the like are also effective as an aspect of the present invention.

According to the present invention, it is possible to improve the coding efficiency by reducing the code amount of the flag representing the reference destination of the motion information.

It is a block diagram which shows the structure of the moving image encoder which comprised the reference destination prediction method of the motion information which concerns on embodiment. It is a block diagram which shows the structure of the moving image decoding apparatus which comprised the reference destination prediction method of the motion information which concerns on embodiment. It is a figure for demonstrating the encoding block in embodiment. It is a figure for demonstrating the kind of shape of the prediction block in embodiment. It is a figure for demonstrating the position of the partition used as the process target of the reference destination prediction method of the motion information in embodiment. It is a figure for demonstrating the periphery of the partition used as the process target of the reference destination prediction method of the motion information in embodiment. It is a figure which shows the syntax pattern of the bit stream which determines whether the reference destination prediction method of the motion information in embodiment is performed at a sequence level. It is a figure which shows the syntax pattern of the bit stream which determines whether the reference destination prediction method of the motion information in embodiment is performed at a prediction block level. It is a flowchart explaining operation | movement of the reference designation | designated part of FIG. FIG. 10A is a flowchart showing details of the calculation processing of the number of candidates NumMergeCandidates in FIG. 9, and FIG. 10B is a diagram showing an example of a reference destination partition adjacent to the processing target partition. It is a flowchart explaining operation | movement of the prediction selection part of FIG. It is a flowchart explaining operation | movement of the motion information selection part of FIG. FIG. 13A and FIG. 13B are flowcharts for explaining processing executed in accordance with a flag value indicating a predicted reference destination. It is a flowchart explaining the operation | movement of the 1st Example of the reference destination prediction method of the motion information in this Embodiment. It is a flowchart explaining the detailed operation | movement of a boundary determination with the adjacent partition of FIG. It is a figure explaining arrangement | positioning and a definition of the partition adjacent to the partition of a process target. It is a figure explaining the definition of a motion boundary when two or more adjacent partitions are arrange | positioned. It is a flowchart explaining the detailed operation | movement of adjacent edge length comparison with the adjacent partition of FIG. It is a figure explaining the reference destination partition determined based on the adjacency condition of the partition adjacent to a process target partition. It is a figure which shows an example of the adjacency conditions of the partition adjacent to a process target partition. It is a figure which shows another example of the adjacency conditions of the partition adjacent to a process target partition. It is a flowchart explaining operation | movement of the 2nd Example of the reference destination prediction method of motion information. It is a figure which shows the pixel area | region which calculates the boundary residual of an adjacent partition. It is a flowchart explaining the operation | movement of the determination of the reference destination partition based on the boundary residual of an adjacent partition. It is a figure which shows the syntax pattern of the bit stream which determines whether it performs in the prediction block level in the 3rd Example of the reference destination prediction method of motion information. It is a flowchart explaining operation | movement of the 3rd Example of the reference destination prediction method of motion information. It is a flowchart explaining operation | movement of the motion information selection part of a 3rd Example.

Embodiments of the present invention relate to moving picture coding technology, and in particular, improve the efficiency of motion information coding in moving picture coding in which a picture is divided into rectangular blocks and motion estimation and compensation are performed in units of blocks between pictures. Therefore, the motion information of the block to be processed is obtained by using the motion information such as the motion vector of the encoded adjacent block around the block to be processed and the reference picture number as the motion information of the block to be processed. The present invention relates to a technique for reducing the amount of codes by encoding additional information indicating adjacent blocks to be referred to without encoding.

In the embodiment of the present invention, instead of the above-described flag merge_lu_flag that occurs at the same frequency, the reference destination of motion information is predicted using information such as the size of neighboring blocks and the prediction mode, and is determined by the prediction. In addition, a flag merge_proble_flag that represents a match between the adjacent block of the reference destination and the adjacent block actually selected by encoding is defined. If the determination rate by prediction is high, the probability of occurrence of the bit represented by the flag merge_proble_flag increases, so the number of bits representing the flag decreases, and the amount of generated code can be reduced.

Furthermore, in another embodiment of the present invention, the flag merge_proble_flag itself representing the match with the above-mentioned reference destination prediction is defined on the assumption that the coincidence accuracy of the determination of the adjacent block that becomes the reference destination by the reference destination prediction is high. In addition, by indicating the adjacent block of the reference destination in the determination result obtained by the prediction of the reference destination, and indicating the adjacent block of the reference destination with the flag merge_direc_flag having the same function as the merge_lu_flag only when the prediction of the reference destination is indefinite. It is also possible to offset the increase in the code amount due to the flag merge_direc_flag by reducing the code amount required for the flag merge_proble_flag, and to relatively improve the encoding efficiency.

Hereinafter, a moving picture coding apparatus and a moving picture decoding apparatus according to embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a moving picture encoding apparatus according to an embodiment. The moving image encoding apparatus according to the embodiment includes a motion vector detection unit 101, a subtraction unit 102, an orthogonal transform / quantization unit 103, a variable length encoding unit 104, an inverse quantization / inverse orthogonal transform unit 105, and a motion compensation unit 106. , A weighted prediction unit 107, an addition unit 108, a deblocking filter unit 109, a memory 110, an in-screen prediction unit 111, a prediction selection unit 112, and a reference destination designation unit 113.

The intra prediction unit 111 performs intra prediction from the input encoding target block and the encoded decoded block image adjacent to the encoding target block stored in the memory 110, and determines the intra prediction block. Output.

The reference destination specifying unit 113 specifies any one of a plurality of adjacent blocks adjacent to the encoding target block stored in the memory 110 as a reference destination adjacent block when performing the merge processing, and the reference destination adjacent The motion vector of the block is given to the motion compensation unit 106.

The motion vector detection unit 101 refers to the input image signal and the reference image signal stored in the memory 110 to perform matching between blocks in units of blocks to detect a motion vector, and to detect the detected motion vector. This is given to the motion compensation unit 106.

The motion compensation unit 106 generates a prediction image using the motion vector of the block to be encoded detected by the motion vector detection unit 101 or the motion vector of the adjacent block specified by the reference destination specifying unit 113. The weighted prediction unit 107 adaptively multiplies the prediction image generated by the motion compensation unit 106 by a weight coefficient to generate a final prediction image, and provides the prediction selection unit 112 with it.

The prediction selection unit 112 performs intra-screen prediction by the intra-screen prediction unit 111, inter-image prediction using the motion vector of the block to be encoded by the motion compensation unit 106, and the reference destination adjacent block specified by the reference destination specifying unit 113. One prediction method with the smallest code amount is selected from the inter-image prediction using the motion vector, and a prediction image by the selected prediction method is given to the subtraction unit 102 and the addition unit 108.

As a prediction method that minimizes the code amount, when the inter-picture prediction using the motion vector of the reference destination adjacent block is selected by the prediction selection unit 112, the reference destination specifying unit 113 selects the adjacent block that is the reference destination of the merge process. The prediction process is executed, and the prediction result of the merge reference destination is given to the prediction selection unit 112. The prediction selecting unit 112 determines whether or not the reference destination adjacent block predicted by the reference destination specifying unit 113 matches the reference destination adjacent block used in the actually selected prediction method, and the reference destination adjacent block The reference destination valid information indicating the validity of the prediction result is supplied to the variable length coding unit 104 together with the information indicating the selected prediction method.

The subtraction unit 102 generates a residual signal by subtracting the image to be encoded and the predicted image, and supplies the residual signal to the orthogonal transform / quantization unit 103. The orthogonal transform / quantization unit 103 performs orthogonal transform and quantization on the residual signal, generates a transform signal, and supplies the transform signal to the variable length coding unit 104 and the inverse quantization / inverse orthogonal transform unit 105. The variable length coding unit 104 entropy codes the residual signal that has been orthogonally transformed and quantized. The variable length encoding unit 104 also encodes reference destination valid information generated by the prediction selection unit 112 and information indicating the selected prediction method, and outputs a bit stream including an encoded image.

The inverse quantization / inverse orthogonal transform unit 105 performs inverse quantization and inverse orthogonal transform on the transform signal received from the orthogonal transform / quantization unit 103 to return to the original residual signal. The adding unit 108 adds the predicted image and the residual signal, generates a decoded image, and supplies the decoded image to the deblocking filter unit 109. The deblocking filter unit 109 performs a process of reducing block distortion due to encoding on the decoded image and stores it in the memory 110. The memory 110 stores information on decoded images and already encoded images.

FIG. 2 is a block diagram showing a configuration of a moving picture decoding apparatus according to an embodiment corresponding to the moving picture encoding apparatus of FIG. The moving picture decoding apparatus according to the embodiment includes a variable length decoding unit 201, an inverse quantization / inverse orthogonal transform unit 202, a motion compensation unit 203, a weighted prediction unit 204, an addition unit 205, a deblocking filter unit 206, and a memory 207. , An intra-screen prediction unit 208, an intra-screen / inter-screen prediction selection unit 209, and a motion information selection unit 210.

The decoding process of the moving picture decoding apparatus in FIG. 2 corresponds to the decoding process provided in the moving picture encoding apparatus in FIG. 1, so that the inverse quantization / inverse orthogonal transform unit 202 in FIG. Each configuration of the motion compensation unit 203, the weighted prediction unit 204, the addition unit 205, the deblocking filter unit 206, the memory 207, and the in-screen prediction unit 208 is the inverse quantization / inverse of the moving picture coding apparatus in FIG. The orthogonal transform unit 105, the motion compensation unit 106, the weighted prediction unit 107, the addition unit 108, the deblocking filter unit 109, the memory 110, and the in-screen prediction unit 111 have functions corresponding to the respective configurations.

The variable length decoding unit 201 decodes the bitstream and outputs information on the prediction residual signal and the motion vector, provides the prediction residual signal to the inverse quantization / inverse orthogonal transform unit 202, and provides information on the motion vector as the motion information. Information about the coding mode is given to the selection unit 210 to the intra-screen / inter-screen prediction selection unit 209.

The inverse quantization / inverse orthogonal transform unit 202 performs inverse orthogonal transform and inverse quantization on the prediction residual signal decoded by the variable length decoding unit 201. The addition unit 205 decodes the image signal by adding the prediction residual component inversely transformed by the inverse quantization / inverse orthogonal transformation unit 202 and the prediction image calculated by the weighted prediction unit 204, and decodes the image signal. This is applied to the blocking filter unit 206. The deblocking filter unit 206 performs a process of reducing block distortion due to encoding on the decoded image and stores it in the memory 207.

The intra-screen / inter-screen prediction selection unit 209 determines whether the encoding mode is intra-screen prediction or inter-screen prediction, and if it is intra-screen prediction, instructs the intra-screen prediction unit 208 to execute the intra-screen prediction process. In the case of inter-screen prediction, the motion information selection unit 210 is instructed to execute inter-screen prediction processing.

The motion information selection unit 210 selects whether or not motion information such as a motion vector is due to merge processing. If it is due to merge processing, the motion information selection unit 210 is based on predetermined adjacent conditions based on reference destination valid information. It is determined whether the adjacent block to be determined can be designated as a reference block.

The intra prediction unit 208 performs intra prediction from the input decoding target block and the decoded block image adjacent to the decoding target block stored in the memory 207, and passes the intra prediction block to the adding unit 205.

The motion compensation unit 203 generates a prediction image using the decoded motion vector of the decoding target block when the inter-screen prediction is not merge processing, that is, is based on normal motion vector detection processing. When the inter-frame prediction is based on the merge process and the reference destination valid information is affirmative, the motion compensation unit 203 performs prediction using the motion vector of the reference destination adjacent block predicted by the reference destination prediction of the motion information. Generate an image.

The weighted prediction unit 204 adaptively multiplies the prediction image generated by the motion compensation unit 203 by a weighting factor to generate a final prediction image, which is given to the addition unit 205.

The prediction method of an adjacent block of a reference destination according to the embodiment is performed in the prediction selection unit 112 of the video encoding device in FIG. 1 and the motion information selection unit 210 of the video decoding device in FIG.

In the prediction selection unit 112 of the video encoding device, from among the motion information detected by the normal motion vector detection process or the inter-screen prediction based on the motion information of the adjacent block selected by the merge process, and the intra-screen prediction, One encoding mode is selected that has the least amount of generated code, the best image quality, or both. Further, if the selected encoding mode is inter-screen prediction, it is determined whether the motion information is detected by a normal motion vector detection process or is selected by a merge process, and the merge process is performed. In this case, the reference destination prediction of the motion information is performed, the identity between the predicted reference destination adjacent block and the adjacent block selected by the merging process is determined, and the result is encoded as reference destination valid information. Is transmitted. When the merge process is not performed, normal motion vector prediction is performed, and a difference motion vector between the motion vector and the predicted motion vector and other motion information are encoded and transmitted.

On the other hand, in the motion information selection unit 210 of the video decoding device, the intra-screen / inter-screen prediction selection unit 209 determines the inter-screen based on the encoding mode and motion information decoded from the bit stream by the variable length decoding unit 201. If it is determined to be prediction, it is further determined whether or not the inter-screen prediction is due to merge processing. If it is determined to be due to merge processing, motion information reference destination prediction is performed. The motion information is acquired by determining the identity between the predicted reference adjacent block and the adjacent block specified by the decoded additional information. In the following embodiments, details of motion information reference destination prediction will be described.

[Example 1]
Prior to describing an example of a motion information reference destination prediction method, terms used in this example will be described.

(About coding block)
In the embodiment, as shown in FIG. 3, the screen is equally divided into square rectangular blocks of the same size. This block is called a coding block and is the basis of processing when performing coding and decoding. According to the texture in the screen, the coding block can be divided into four blocks with a small block size in order to optimize the coding process. A coded block that is divided into equal screen sizes in the screen shown in FIG. 3 is called a maximum coded block, and the inside of which is divided into four according to the coding conditions is collectively called a coded block. An encoded block having a minimum size that cannot be further divided into four is referred to as a minimum encoded block.

(About prediction block)
When motion compensation is performed by dividing the screen into blocks, the smaller the motion compensation block size, the more detailed prediction can be made, so select the optimal block size from several block sizes. Thus, a mechanism for performing motion compensation by dividing the inside of the block is adopted. A block that performs this motion compensation is called a prediction block. The prediction block is represented by the same size as the coding block, and according to the motion compensation, the prediction block is regarded as one block without being divided, and is divided into two in the horizontal or vertical direction. Divided into four parts by equal division. The mode corresponding to the division type is defined according to the size after the division, and is shown in FIG.

(About partitions)
Each area obtained by dividing the prediction block is called a partition. In order to manage the partitions in the prediction block, numbers starting from 0 are assigned to the partitions existing in the prediction block in the zigzag scan order. This number is called a partition number and is represented by puPartIdx. The number described in the partition of the prediction block in FIG. 4 represents the partition number of the partition.

A motion information reference destination prediction method according to an embodiment will be described with reference to the drawings. The motion information reference destination prediction method is performed in any of encoding and decoding processes for each partition constituting a prediction block.

As shown in FIG. 5, the motion information is referred from the surrounding partition adjacent to the partition of the prediction block (partition to be processed in FIG. 5) defined for motion compensation within the coding block in the same picture. Select the destination partition.

Since the partition to be referred to is also used in decoding, a partition that has already been encoded before encoding the partition to be processed is a candidate. Since encoding is performed in block units in the raster scan order from the upper left to the lower right of the screen, here, in principle, the left or upper partition adjacent to the periphery of the partition to be processed is considered as a reference destination candidate. The following explanation will be based on this assumption unless otherwise noted.

FIG. 6 is an enlarged view of the area represented by the thick dotted circle in FIG.

FIG. 6 shows a partition to be processed and a partition adjacent to it. Predicted motion vector candidates include a partition group A composed of a partition Ak (k = 0,..., NA−1) adjacent to the left side of the partition to be processed, and a partition Bk (k = 0, .., NB-1) are selected from the two partition groups of partition group B, each of which is a reference destination of motion information. Here, nA represents the total number of partition groups adjacent to the left, and nB represents the total number of partition groups adjacent immediately above.

Since the size of the partition changes depending on motion compensation, as shown in FIG. 6, when the size of the partition to be processed and the neighboring partitions are different, the reference destination on the left or directly above is determined based on the following rule. Determine candidate partitions.

When there are a plurality of adjacent partitions on the left with respect to the partition to be processed, the top partition A0 among them is set as a reference destination candidate.
When there are a plurality of adjacent partitions directly above the processing target partition, the leftmost partition B0 among them is set as a reference destination candidate.

(About syntax)
First, the syntax of a bit stream of a moving image that is encoded by a moving image encoding device that includes the motion information reference destination prediction method according to the present embodiment will be described.

FIG. 7 shows a first syntax pattern described in a sequence parameter set (SPS) of a moving image bit stream. The sequence parameter set is a header in which information related to coding of the entire sequence is described. When performing inter-screen prediction over the entire sequence, the motion information reference destination prediction method according to the present embodiment is applied. A first flag inter_merge_flag indicating whether or not to perform is set.

FIG. 8 shows a second syntax pattern described for each prediction block in the slice. When the prediction mode of the prediction block is inter-screen prediction (MODE_INTER), the number of effective neighboring partitions NumMergeCandidates is obtained for each partition in the prediction block, and when NumMergeCandidates exceeds 0, merging is performed in this partition. A second flag merge_flag [i] indicating whether or not to apply is set. Here, i represents the partition number in the prediction block. The second flag merge_flag [i] is not coded and is not described in the bitstream when the motion information reference prediction method according to the present embodiment is not applied, that is, when the inter_merge_flag is false (0).

Next, when the second flag merge_flag [i] is true (1) and NumMergeCandidates exceeds 1, the reference destination adjacent partition selected by the motion information reference destination prediction method according to the present embodiment is correct. A third flag merge_proble_flag [i] indicating whether or not is set. When the third flag merge_proble_flag [i] does not apply merging, that is, when merge_flag [i] is false (0), the motion information detected by the normal motion vector detection is used, so the third flag merge_proble_flag [i] is used. i] does not need to be determined. Also, even when NumMergeCandidates is 1, it is not encoded. This is because, if the number of effective partitions adjacent to each other is one, one of the partitions becomes a reference destination partition, so that the motion information of the reference destination partition is determined without transmitting merge_proble_flag [i].

In addition, calculation of NumMergeCandidates will be described later.

In the following, in order to simplify the explanation, a variable obtained by removing the array part of the second flag and the third flag will be used.

(Reference prediction of motion information in encoding)
Based on the above-described syntax, the operation of the motion information reference destination prediction method according to the embodiment in the video encoding device that encodes a video bitstream will be described. When the motion information reference prediction method is applied to inter-screen prediction over the entire sequence, the first flag inter_merge_flag described in the sequence parameter set (SPS) is set to true (1). Next, application / non-application of the motion information reference destination prediction method according to the present embodiment is switched for each partition of a prediction block whose prediction mode in the slice is inter-frame prediction (MODE_INTER).

The reference destination specifying unit 113 determines whether or not to use a partition adjacent to the left and directly above the processing target partition as a reference destination of the merge processing. The operation of the reference destination specifying unit 113 will be described with reference to the flowchart of FIG. First, the number NumMergeCandidates of peripheral partitions adjacent to the partition to be processed is calculated (S101).

FIG. 10A is a flowchart showing details of the calculation processing of the number of candidates NumMergeCandidates. First, NumMergeCandidates is set to 0 (S201). Next, a partition located to the left of the partition to be processed is set as a reference destination partition (S202). It is determined from the position information in the screen of the partition to be processed whether or not the reference destination partition exists in the screen (S203).

FIG. 10 (b) shows an example of a reference destination partition (rectangle represented by diagonal lines in the figure) adjacent to the partition to be processed (gray rectangle in the figure). As shown in the figure, when the partition to be processed is located at the upper left, there is no reference destination partition, only when it is located at the top of the screen, only at the left, when located at the left edge of the screen, only at the top, other positions Then it is on the left and top. When the reference destination partition does not exist in the screen (No in S203), the subsequent processing is skipped and the process proceeds to Step S207.

When the reference destination partition exists in the screen (Yes in S203), the reference destination specifying unit 113 acquires the motion information of the reference destination partition (S204). The reference destination designating unit 113 reads the encoded motion information corresponding to the reference destination partition from the memory 110. Based on the read motion information of the reference destination partition, the reference destination specifying unit 113 determines whether or not the encoding mode of the reference destination partition is intra (S205). If it is intra, the process proceeds to S207. If it is not intra, NumMergeCandidates is incremented by 1 (S206).

It is determined whether the setting of the reference destination partition is on the partition to be processed (S207). If the reference destination partition is set on the processing target partition, the process ends and outputs NumMergeCandidates. When the reference destination partition is not set on the processing target partition, the reference destination partition is set again on the processing target partition, and the determination processing after step S203 is continued (S208).

The number of candidates NumMergeCandidates is calculated as described above.

Next, returning to FIG. 9, it is determined whether or not the calculated number of candidates NumMergeCandidates is greater than 0 (S102). When the number of candidates NumMergeCandidates is 0, there is no adjacent partition as a reference destination of the merge process, and therefore motion compensation based on the motion information selected by the merge process is not performed. If the number of candidates NumMergeCandidates is greater than 0, the process proceeds to the next step S103.

It is determined whether the candidate number NumMergeCandidates is greater than 1 (S103). If the number of candidates NumMergeCandidates is 1 (No in S103), one of them becomes a reference partition for the merge process, and the reference destination specifying unit 113 acquires the motion information of the reference destination partition from the memory 110 (S105). ). The reference destination specifying unit 113 supplies the acquired motion information of the reference destination partition to the motion compensation unit 106, and the motion compensation unit 106 performs motion compensation of the processing target partition based on the motion information of the specified reference destination partition. .

If NumMergeCandidates is greater than 1, that is, 2 (Yes in S103), the reference destination designating unit 113 reads out and acquires the motion information of the partitions adjacent to the left and above the processing target partition from the memory 110 (S104).

The reference destination specifying unit 113 compares the acquired motion information of the left and upper adjacent partitions (S106). Specifically, the prediction direction, the reference picture number, and the motion vector constituting the motion information are compared. When the motion information of the left and upper adjacent partitions is exactly the same (Yes in S106), the number of candidates NumMergeCandidates is changed to 1 (S107), and the reference destination designating unit 113 sets either the left or the upper adjacent The motion information of the partition to be supplied is supplied to the motion compensation unit 106 as the motion information of the reference destination partition. Here, the motion information of the partition adjacent to the left is selected. When the motion information of the left and upper adjacent partitions is different (Yes in S106), the reference destination specifying unit 113 supplies the two pieces of motion information to the motion compensation unit 106, respectively.

As described above, the reference destination specifying unit 113 supplies the motion information of the partition of the reference block selected by the merging process to the motion compensation unit 106, and the motion compensation unit 106 specifies the partition of the specified reference block. Motion compensation is performed based on the motion information. Also, the motion compensation unit 106 performs normal motion compensation based on the motion vector for each partition of the prediction block detected by the motion vector detection unit 101. The in-screen prediction unit 111 performs in-plane prediction using pixel correlation in the screen. A prediction image generated by these three prediction methods is input to the prediction selection unit 112.

The prediction selection unit 112 calculates a generated code amount of additional information such as a residual image and motion information generated by the difference between the image to be encoded and the three input prediction images, and makes a prediction that minimizes the code amount Select a method.

Here, a case where a prediction method based on motion compensation based on the motion information of the reference partition selected by the merge process is selected will be described. At this time, the prediction selection unit 112 sets a flag merge_dirc_flag that indicates which one of the left or the upper is used as the motion information of the reference destination partition selected by the merge process. In the following description, the motion information of the partition adjacent to the left is selected when merge_dirc_flag is 1, and the motion information of the upper adjacent partition is selected when 0.

The prediction selection unit 112 sets a flag for merge processing defined by the above-described syntax when the prediction method by merge processing is selected. FIG. 11 is a flowchart illustrating a procedure in which the prediction selection unit 112 sets a flag for merge processing.

First, it is determined whether or not the prediction method selected by the prediction selection unit 112 is prediction based on motion information selected in the merge process (S301). In the case of inter-screen prediction using intra-picture prediction or normal motion vector detection (No in S301), merge_flag is set to 0 and the process ends (S303).

If the prediction is based on the motion information selected in the merge process (Yes in S301), merge_flag is set to 1 (S302). Next, the number of motion information selected by the merge process is checked (S304). Specifically, it is determined whether or not the number of adjacent partition candidates NumMergeCandidates to be specified by the reference destination specifying unit 113 is greater than one.

If the number of candidates NumMergeCandidates is 1 or less (No in S304), it is assumed that merge processing is selected, and there is always one adjacent partition of a valid reference destination, either on the left or above Since it can be specified, only merge_flag is set and the process ends.

If the number of candidates NumMergeCandidates is larger than 1, that is, if there is a candidate for motion information of two adjacent partitions (Yes in S304), the process proceeds to prediction of a reference destination partition (S305). Here, in prediction of the reference destination partition, a reference destination prediction method for motion information described later is used. Based on information such as the size of the partition adjacent to the left and top of the target partition and the length of the side where the target partition and the adjacent partition are in contact with each other, the reference partition is predicted and predicted A flag pred_direc_flag indicating the reference destination partition is output.

The flag pred_dirc_flag indicating the predicted reference destination partition indicates the selection destination on the left (1) or above (0) as the adjacent partition of the reference destination, as in the case of merge_direct_flag, and the reference destination partition cannot be selected by prediction. Indefinite (2) is set.

Next, the flag pred_direc_flag indicating the predicted reference destination partition set in this way is determined (S306). The flag pred_dirc_flag indicating the predicted reference destination partition is 2 when prediction is impossible in the prediction of the reference destination partition (Yes in S306), and at this time, it is selected as the prediction method that actually minimizes the generated code amount. The merge_direc_flag indicating the reference destination partition of the motion information used in the prediction performed by the merge processing is set in the flag merge_proble_flag indicating the reference destination valid information, and the process ends (S307). In this case, the flag merge_proble_flag indicating the reference destination valid information directly indicates the reference destination partition in the merge process as in the conventional case.

When the flag pred_dirc_flag indicating the predicted reference destination partition is not 2 (No in S306), the flags merge_dirc_flag and pred_dirc_flag are compared (S308). merge_direc_flag represents a reference destination partition of motion information used in prediction by merge processing selected as a prediction method that actually minimizes the amount of generated code, and pred_direc_flag has the same value if the prediction of the reference destination partition is correct. In this case, the flag merge_proble_flag is set to 1 (S309). On the other hand, if the prediction of the reference destination partition is not correct, pred_dirc_flag has a different value. In this case, the flag merge_proble_flag is set to 0 (S310).

When merge_proble_flag indicating success / failure of the merge destination partition is used, merge_proble_flag becomes 1 when the merge destination partition is predicted, and merge_proble_flag becomes 0 only when the prediction is off. In other words, merge_proble_flag can be transmitted with a smaller amount of information by using arithmetic coding using the fact that merge_proble_flag has a higher probability of 1 and merge_proble_flag has a higher probability of 1 as the merge destination partition is predicted. it can.

As described above, when motion-compensated prediction based on the motion information of adjacent partitions selected by the merge process for each partition is selected, the first flag inter_merge_flag described in the header of the SPS is described in the prediction block. The second flag merge_flag and the third flag merge_proble_flag are set and encoded.

(Prediction of motion information reference in decoding)
Based on the syntax described above, an operation of the motion information reference destination prediction method according to the embodiment in the video decoding device that decodes the bit stream of the video encoded by the video encoding device will be described. .

First, from the flag inter_merge_flag described in the SPS of the bitstream decoded by the variable length decoding unit 201, the motion information reference destination prediction method according to the present embodiment is used for the entire bitstream sequence. It is determined whether inter-screen prediction by merge processing is applied. When inter_merge_flag is true (1), merge processing is applied. When inter_merge_flag is false (0), merge processing is ignored and inter-screen prediction based on motion information decoded from a conventional bitstream is performed. . In the following description, it is assumed that inter_merge_flag is true (1), that is, the merge processing using the motion information reference destination prediction method according to the present embodiment is applied.

Next, the intra-screen / inter-screen prediction selection unit 209 determines whether to select intra-screen or inter-screen prediction with reference to the prediction mode for each prediction block in the slice. When the intra prediction (MODE_INTRA) is selected, the process proceeds to the intra prediction unit 208, and when the inter prediction (MODE_INTER) is selected, the process proceeds to the motion information selection unit 210.

Here, the motion information selection unit 210 uses, as information decoded from the bit stream for each partition of the prediction block, the motion information detected by the motion vector detection of encoding or the reference destination partition selected by the merge process. Either one of the motion information is selected, and the selected motion information is output to the motion compensation unit 203.

FIG. 12 is a flowchart showing a selection procedure of the motion information selection unit 210, which will be described with reference to this figure. First, the number NumMergeCandidates of peripheral partitions adjacent to the partition to be processed is calculated (S401). Since the calculation procedure is the same as S101 in encoding, it is omitted.

It is determined whether or not the calculated number of candidates NumMergeCandidates is greater than 0 (S402). In the case of 0, since there is no adjacent partition that becomes the reference destination of the merge process, the motion information selection unit 210 outputs the motion information decoded from the bitstream as in the conventional case. When the number is larger than 0, merge_flag decoded by the variable length decoding unit 201 is read (S403), and selection of motion information is determined based on merge_flag (S404). When merge_flag is false (0), the motion information selection unit 210 outputs motion information decoded from the bitstream as in the conventional case. When merge_flag is true (1), the process proceeds to processing for selecting motion information of the reference partition in the merge processing.

Again, it is determined whether or not the number of candidates NumMergeCandidates is larger than one (S405). When the number of candidates NumMergeCandidates is one, one of them becomes a reference destination partition for the merge process, so the motion information of the reference destination partition is acquired from the memory 207 in which the decoded information is recorded (S407). The acquired motion information is supplied to the motion compensation unit 203, and motion compensation is performed based on the motion information.

When the number of candidates NumMergeCandidates is larger than 1, that is, two, the motion information of the partitions adjacent to the left and above the partition to be processed is read from the memory 207 and acquired (S406). The obtained motion information of the left and upper adjacent partitions is compared (S408). Specifically, the prediction direction, the reference picture number, and the motion vector constituting the motion information are compared. If the motion information is exactly the same, the motion information of either the left or upper adjacent partition is acquired and supplied to the motion compensation unit 203 (S410). Here, the motion information of the partition adjacent to the left is selected as in the encoding.

If the motion information is different, the process proceeds to prediction of the reference destination partition (S409). For prediction of the reference destination partition, a motion information reference destination prediction method, which will be described later, is used as in the case of encoding. The target partition is predicted based on information such as the size of the partition adjacent to the processing target partition and the left and top of the periphery, and the length of the side where the processing target partition and the adjacent partition are in contact with each other. The flag pred_dirc_flag indicating the referred reference partition is output.

The flag pred_dirc_flag indicating the predicted reference destination partition is represented by left (1) or upper (0) when the adjacent reference destination partition is selected as in the encoding, and the reference destination partition can be selected by prediction. If not, undefined (2) is set.

Next, the pred_direc_flag output is determined (S411). The case where pred_dirc_flag is 2 is a case where prediction is impossible due to prediction of the reference destination partition, and at this time, the process proceeds to case 1 (S412). On the other hand, if pred_dirc_flag is not 2, the process proceeds to case 2 (S413).

FIG. 13A shows a flowchart of the process of case 1 in step S412. In case 1, the prediction result is undefined in the prediction of the reference destination partition, and the prediction result of the reference destination is not shown. Therefore, the partition adjacent to the left or above indicated by merge_proble_flag is selected as the reference destination partition as in the conventional case. This is because, in the encoding side, merge_direct_flag indicating the actual reference destination partition is set in merge_proble_flag.

The merge_proble_flag decoded by the variable length decoding unit 201 is read (S421), and a reference destination partition is selected based on the merge_proble_flag (S422). When pred_dirc_flag is 1, the left adjacent partition (S423) is selected, and when merge_proble_flag is 0, the upper adjacent partition is selected as the reference destination partition (S424).

FIG. 13B shows a flowchart of the process of case 2 in step S413. In the case 2 process, the probability of the flag pred_direc_flag indicating the predicted reference destination partition is represented by the flag merge_proble_flag indicating the reference destination validity. That is, merge_proble_flag is represented by 1 (true) if the reference destination partition indicated by pred_dirc_flag is correct, and merge_proble_flag is represented by 0 (false) if not correct.

The merge_proble_flag decoded by the variable length decoding unit 201 is read (S431), and the selection of the reference destination partition is determined based on the merge_proble_flag (S432). When merge_proble_flag is 1, a partition having a value indicated by pred_dirc_flag determined by prediction of the reference destination partition is selected as the reference destination partition (S433). When merge_proble_flag is 0, a partition opposite to the value indicated by pred_direc_flag is selected as a reference destination partition. For example, if pred_dirc_flag is 0 (left), the upper adjacent partition is set as the reference destination partition (S434).

Returning to FIG. 12, the motion information of the reference destination partition selected in the processing of case 1 and case 2 is acquired (S415), supplied to the motion compensation unit 203, and the process ends.

As described above, when inter-screen prediction by merge processing is selected according to each flag decoded by the variable-length decoding unit 201 and the calculation result in the decoding process, the reference is made by the motion information reference destination prediction method described later. Select the destination partition. A decoded image is generated by adding the predicted image generated by motion compensation from the motion information of the selected reference destination partition and the residual signal decoded from the bitstream.

(Motion information reference destination prediction method)
FIG. 14 is a flowchart for explaining a motion information reference destination prediction method according to the present embodiment. The process shown in FIG. 14 is performed in units of partitions. In each process, a partition that refers to motion information is determined for the partition to be processed, and if a partition is determined as a reference destination, the process does not proceed to the next process and ends. Hereinafter, each process will be described step by step.

First, a boundary determination with a neighboring partition adjacent to the left or top of the partition to be processed is performed (S500). FIG. 15 is a flowchart showing a boundary determination process with adjacent surrounding partitions in step S500. Details of the processing will be described with reference to FIG.

First, motion boundary detection is performed between a partition to be processed and neighboring partitions around it (S501).

FIG. 16 shows an arrangement example in which the processing target partition is X and the left and upper adjacent partitions are A and B, respectively. The rectangular size of the partition to be processed is represented by width wx and height hx, the height of the adjacent left partition is represented by pa, and the width of the adjacent upper partition is represented by pb.

The boundary between the processing target partition and the left and upper adjacent partitions is determined by comparing the height hx of the processing target partition with the height pa of the adjacent left partition and the width wx of the processing target partition. This is done by comparing the width pb of the upper partition.

An edge where an adjacent partition is adjacent to the target partition is called a “motion boundary”. Here, the motion boundary is not only simply evaluating the height or width of one adjacent left or upper partition, but as shown in FIG. 17, for example, a partition adjacent to the partition to be processed. If the motion information of the leftmost partition B0, which is a candidate for the reference destination, and the partition information B1 adjacent to the right are determined to be exactly the same, the motion boundary is extended to the position of the partition B1. Further, the same determination is performed for the partition Bk (k = 2, 3,..., NB-1) on the right side, and the width to the position of the previous partition of the partition determined to have the same motion information is calculated. The added value is defined as the length of the motion boundary.

In the example shown in FIG. 17, it is determined that B0 and B1 have the same motion information. As a result, the length of the motion boundary between the upper adjacent partitions is the width of each partition of B0 and B1. , I.e., pb = pb0 + pb1. It is determined whether the length of the motion boundary thus obtained matches the height or width of the partition to be processed.

First, it is determined whether or not the length of the motion boundary between the adjacent left partition A and the upper partition B matches the height or width of the processing target partition X (S502). If both match, the reference destination partition cannot be determined, and the process proceeds to the next determination step S510. If both do not match, it is determined whether either A or B matches (S503). If A and B do not match, the determination cannot be made, and the process proceeds to the next determination step S510. If the length of the motion boundary of either A or B matches the height or width of the processing target partition X, the matching partition is selected as a reference destination (S504), and the process proceeds to the next determination step S510. Finish first. This is because, when the length of the motion boundary matches the height or width of the processing target partition X, the reference destination using the fact that the adjacent partition and the processing target partition are likely to have the same motion information. Partition prediction.

Next, returning to FIG. 14, a determination is made based on the length of the common part of adjacent sides of the left or upper partition adjacent to the processing target partition (hereinafter referred to as “adjacent side length”) (S510). The partition to be processed is determined based on the estimation that the correlation with the motion of the partition having a long adjacent side length is high. FIG. 18 is a flowchart showing the adjacent side length comparison process with the adjacent partition in step S510. Details of the processing will be described with reference to FIG.

First, the length of the adjacent side in contact with the left or upper partition adjacent to the processing target partition is calculated (S511). As shown in FIG. 16, the adjacent side length to the left partition A is represented by L (A), and the adjacent side length to the upper partition B is represented by L (B) (the hatched portion in FIG. 16 is adjacent). Part). The adjacent side length L (A) with the left partition A is the smaller of the height hx of the partition X to be processed and the height pa of the left partition, and is expressed by the following equation.
L (A) = min (pa, hx)
Here, the function min (a, b) is a function for selecting the smaller one of a and b. On the other hand, the adjacent side length L (B) with the upper partition B is the smaller one of the width wx of the partition X to be processed and the height pb of the upper partition, and is expressed by the following equation.
L (B) = min (pb, wx)
Here, pa and pb are the lengths of the motion boundaries obtained in step S500.

The obtained adjacent side lengths L (A) and L (B) are compared, and a reference destination partition is selected (S512). If L (A) and L (B) are equal, the reference destination partition cannot be determined, so the reference destination partition is not selected and is undefined (S513), and the process ends. If L (A) and L (B) are not equal, L (A) and L (B) are compared in size (S514). If L (A) is greater than L (B), A is selected (S515) and the process ends. If L (A) is smaller than L (B), B is selected (S516) and the process ends.

Based on the boundary determination with the adjacent partition described above / (S500) and the adjacent side length comparison (S510), the determination results classified according to the motion boundary and side length conditions of the left and upper partitions adjacent to the partition to be processed are as follows. The summary is shown in the table of FIG. The selection column in the table of FIG. 19 indicates the partition to be referred to, and the symbol “-” represents the case where the reference partition is indefinite.

Here, some examples of determination based on the conditions in the table of FIG. 19 will be described. For example, the fifth boundary comparison condition “pb <wx and pa <hx” in FIG. 19 is arranged as shown in FIG. 20, and the lengths of the motion boundaries of adjacent partitions A and B are both subject to processing. This is the case when the width and height of the partition do not match. In this case, since the determination based on the motion boundary matching cannot be performed, the reference destination is determined by comparing the lengths of adjacent side lengths in contact with the partition to be processed. The adjacent side lengths L (A) and L (B) of the adjacent partitions A and B are the sides pa and pb that are in contact with the partition X to be processed, so that the reference partition is determined by comparing pa and pb. Determined.

In addition, in the ninth boundary comparison condition “pb> wx and pa> hx” in FIG. 19, the length of the motion boundary between the adjacent partitions A and B does not match the width and height of the partition to be processed. Is the case. Also in this case, since the determination based on the motion boundary matching cannot be performed, the reference destination is determined by comparing the lengths of the adjacent sides in contact with the processing target partition. FIG. 21 shows an example of the processing target and the arrangement of adjacent partitions under the boundary comparison condition “pb> wx and pa> hx”. As shown in FIG. 4, since the size and shape of the partition inside the prediction block are defined, the size and shape of the partition meeting the boundary comparison condition “pb> wx and pa> hx” shown in FIG. Limited. In this case, the partition to be processed is limited to N × N regardless of the shape of adjacent partitions. Therefore, the adjacent side lengths L (A) and L (B) of adjacent partitions A and B are both N. L (A) = L (B) = N, and it cannot be determined from the comparison by the adjacent side length, so it is determined as indefinite.

Thus, the prediction result of the determined reference destination partition is output as a flag pred_dirc_flag indicating the reference destination partition predicted in the encoding and decoding processes. pred_direc_flag is the same as merge_direc_flag, when it is selected as a reference destination adjacent partition, “1 (left)” or “0 (upper)” indicates the reference destination, and the reference destination partition cannot be selected by prediction. Is set to “2 (undefined)”.

As described above, according to the first embodiment, the occurrence of the bit represented by the flag merge_proble_flag used for designating the reference destination is biased and generated by predicting the reference block in the merge process. The amount of codes can be reduced, and the encoding efficiency can be improved. Conventionally, when the left or upper adjacent block is designated as the reference destination, the reference destination flag has a frequency of ½ and the arithmetic coding cannot be used. However, according to the first embodiment, the predicted reference is By using the flag merge_proble_flag indicating the probability of the previous, if the prediction is correct, the bit frequency of the flag can be biased, and the amount of generated codes can be reduced in arithmetic coding. In addition, the adjacent block that matches the size of the encoding target block, or the adjacent block having a longer side length is highly correlated with the encoding target block, so that the reference block It is possible to improve the accuracy of prediction by predicting.

[Example 2]
A second example of the motion information reference destination prediction method according to the embodiment of the present invention will be described. The difference from the first embodiment is that a determination process based on the residual signal of the left or upper partition adjacent to the processing target partition is added as a determination condition for the reference destination prediction.

FIG. 22 is a flowchart for explaining the operation of the second embodiment of the motion information reference destination prediction method. The processes of steps S600 and S610 shown in FIG. 22 are the same as the processes of steps S500 and S510 of the first embodiment shown in FIG. A boundary determination is made with the surrounding partition adjacent to the left or upper side of the partition to be processed (S600), and the length of the common part of adjacent sides of the left or upper partition adjacent to the processing target partition ("adjacent" The side length ") is determined (S610).

After the determination based on the adjacent side length between the left or upper partition adjacent to the partition to be processed (S610), if the partition is not determined as the reference destination, the process proceeds to the determination based on the residual signal (S620).

Here, parameters when performing the determination process based on the residual signal will be described. FIG. 23 is a diagram illustrating a pixel region for calculating a boundary residual between adjacent partitions. As shown in FIG. 23, let A be the left partition adjacent to partition X to be processed, and B be the upper partition. Each partition A, B is assumed to be composed of a pixel residual signal after motion compensation prediction in the encoding process or a pixel residual signal after inverse quantization and inverse orthogonal transformation in the decoding process.

The pixel area of the partition A that is in contact with the partition X is represented by an area having a height of min (ha, hx) and a width of 1 pixel located at the left boundary of the partition X, and a residual signal of the pixel. Δ (A) is expressed by the following equation, where δ (A) is the sum of absolute values of.

Here, ai represents the residual signal of the pixel in partition A adjacent to partition X. The number of pixels in the partition A adjacent to the partition X is the smaller of the number of pixels ha representing the height of the partition A and the height hx of the partition X, and is represented by a function min (ha, hx).

Similarly, the pixel area of the partition B that is in contact with the partition X is represented by an area having a width of min (wb, wx) and a height of 1 pixel located at the upper boundary of the partition X. If the sum of absolute values of residual signals is δ (B), δ (B) is expressed by the following equation.

Here, bi represents the residual signal of the pixel in partition B adjacent to partition X. The number of pixels in the partition B adjacent to the partition X is the smaller of the number of pixels wb representing the width of the partition B and the width wx of the partition X, and is represented by a function min (wb, wx).

FIG. 24 is a flowchart for explaining the operation of determining the reference destination partition based on the boundary residual between adjacent partitions. First, as a boundary residual with the partition X of the left or upper partition adjacent to the partition X to be processed, the sum of the absolute value of the residual signal of the pixel is calculated (S621).

Next, normalization of the calculated boundary residual is performed (S622). As shown in FIG. 23, when the adjacent side lengths of the partitions A and B adjacent to the partition X are different, it is not appropriate to directly compare the boundary residual of the partition A and the boundary residual of the partition B. As shown in the following equation, normalization is performed by dividing the boundary residuals of the partitions A and B by the adjacent side lengths of the partitions A and B, respectively.

Compare the normalized boundary residuals of partitions A and B. First, it is determined whether or not the boundary residual δ (A) of the partition A is equal to the boundary residual δ (B) of the partition B (S623). If δ (A) and δ (B) are not equal, δ (A) and δ (B) have a magnitude relationship, so a magnitude comparison is performed. Here, it is determined whether or not δ (A) is smaller than δ (B) (S624). It is presumed that the prediction error of partition X can be reduced by selecting motion information in which the residual calculated by motion compensated prediction becomes smaller in the pixel region adjacent to the processing target partition X and applying it to partition X. Is done. Therefore, if δ (A) is smaller than δ (B), partition A is selected as a reference destination (S625), and if not, partition B is selected as a reference destination (S626).

If the boundary residual δ (A) of the partition A is equal to the boundary residual δ (B) of the partition B, the reference destination cannot be determined by the boundary residual, and the process proceeds to calculation of the residual ratio (S627). Here, the residual ratios Δ (A) and Δ (B) of the partitions A and B are the boundary residuals δ (A) and δ (B) before normalization of the partitions A and B calculated in step S621. ) To the total residual of partitions A and B is expressed by the following equation.

Here, Ai and Bi are residual signals of pixels in partitions A and B adjacent to partition X, and nA and nB are the total number of pixels in partitions A and B. However, an area that is not adjacent to the partition X to be processed is not a target. For example, in the partition A of FIG. 23, when ha> hx, the adjacent side length is min (ha, hx) = hx, and therefore the pixels in the upper rectangular area up to the height hx of the partition A are the objects of calculation. Yes, pixels in the rectangular area below the partition A having a height (ha-hx) exceeding the above are excluded. Further, the number of pixels wa representing the width of the partition A is compared with the number of pixels hb representing the height of the partition B, and the smaller one is defined as the width of the target region. In the example shown in FIG. 23, the width of the target region is min (wa, hb) = hb. Therefore, the target area is an area indicated by oblique lines having a width hb and a height hx, and the number of pixels nA of the target area of partition A is given by the following equation.

Similarly, the width of the target area of the partition B is the adjacent side length min (wb, wx) = wb of the partition B, and the height is the number of pixels wa indicating the width of the partition A and the number of pixels indicating the height of the partition B. Since hb is smaller, min (wa, hb) = hb. The number of pixels nB in the target area of partition B is given by the following equation.

nA is a rectangular area indicated by light gray in the thick dotted line in the partition A of FIG. 23, and the denominator of the calculation formula of Δ (A) is the absolute value of the residual signal of the pixel in this rectangular area. Expressed as the sum. On the other hand, nB covers the entire area of partition B in FIG. 23, and the denominator part of the calculation formula of Δ (B) is expressed as the sum of absolute values of residual signals of pixels in partition B. The boundary residuals δ (A) and δ (B) before normalization are sums of absolute values of residual signals of pixels adjacent to the partition X, respectively, and the number of pixels adjacent to the partition X is min (ha, hx). ) = Hx, min (wb, wx) = wb, the residual ratios Δ (A) and Δ (B) are calculated with the same number of pixel ratios, and Δ (A) and Δ (B) are the partition X Represents the ratio of the residual signals of adjacent pixels to the residual signals of the pixels in partitions A and B. If this ratio is high, the motion compensation prediction near the pixel adjacent to the partition X is not applied, that is, the motion information used for the motion compensation prediction is not appropriate for the prediction near the pixel adjacent to the partition X. Means that. That is, it is determined that the motion information of the adjacent partition having the higher residual ratio is not referred to.

The residual ratio calculated in this way is compared. It is determined whether or not Δ (A) and Δ (B) are equal (S628). If Δ (A) and Δ (B) are not equal, Δ (A) and Δ (B) have a magnitude relationship, so a magnitude comparison is performed. Here, it is determined whether or not Δ (A) is smaller than Δ (B) (S629). If Δ (A) is smaller than Δ (B), A is selected (S630). Otherwise, B is selected (S631). If Δ (A) and Δ (B) are equal, there is no difference in the residual that occurs regardless of which one is selected, so A is selected (S632), and the process ends.

As described above, the prediction result of the determined reference destination partition is output as the flag pred_direc_flag to the encoding and decoding processes. Similarly to merge_direc_flag, pred_direc_flag indicates a reference destination with “1 (left)” or “0 (upper)” when it is selected as an adjacent partition to be referred to. The reason for selecting A in step S632 is that A (left) is assigned to bit "1" in this embodiment. This is to reduce the amount of generated code in encoding by biasing the occurrence frequency of pred_direc_flag to “1” as prediction of a reference destination partition of motion information.

In Example 2, in step S623 shown in FIG. 24, strict determination as to whether or not the boundary residuals δ (A) and δ (B) are equal is performed using δ (A) and δ (B The determination may be changed based on whether or not the absolute value difference | δ (A) −δ (B) | Even if the absolute value difference when the values of δ (A) and δ (B) are large and the absolute value difference when the values are small, the meanings are different. For example, in the case of δ (A) = 1000 and δ (B) = 1001 and in the case of δ (A) = 8 and δ (B) = 7, the absolute value difference is 1, but the former individual In terms of the boundary residual value, this means a value equal to the error. Therefore, in the comparison of the values of the boundary residuals δ (A) and δ (B), the absolute value difference | δ (A) −δ (B) | between δ (A) and δ (B) and the threshold value ε By performing the comparison, it is possible to extend the determination by including the case where the conventional boundary residuals δ (A) and δ (B) are equal (equivalent to ε = 0). . Here, ε may be a fixed value or may be a variable value that is changed according to the value of the quantization parameter Qp, for example.

Furthermore, in the second embodiment, the difference determination of the boundary residual and the residual ratio is performed in order, but the boundary residual and the residual ratio are determined separately, and the final reference destination partition is selected from the results. You may do it. Further, the boundary residual and the residual ratio may be collected and a new evaluation criterion may be generated and determined. For example, the determination may be made based on the evaluation value generated by the combination of the boundary residual and the residual ratio as in the following equation.

Here, the variable x is A or B, and α is a weighting coefficient. α may be a fixed value, or may be set as a variable value according to a partition, a slice unit, a slice type, or the like.

As described above, according to the second embodiment, prediction of a reference destination is performed by predicting a reference block based on the magnitude of a residual signal of a block adjacent to the block to be encoded. Accuracy can be further increased.

[Example 3]
A third example of the motion information reference destination prediction method according to the embodiment of the present invention will be described. The difference from the first embodiment is that the partition represented by the flag pred_direc_flag determined and output by prediction of the reference destination partition is selected as the reference destination partition. When the flag pred_direc_flag based on prediction of the reference destination partition is 0 or 1, the left (pred_direc_flag is “1”) or the top (pred_dirc_flag is “0”) is selected as the reference destination adjacent partition.

When it is determined as indefinite in the prediction of the reference destination partition, that is, when pred_direc_flag is “2”, the reference destination of the motion information used in the prediction by the merge processing selected as the prediction method that actually minimizes the generated code amount A flag merge_direc_flag representing a partition is used to represent a partition to be referred to. When this condition is expressed as syntax, FIG. 25 is obtained. Subsequent to merge_flag for determining whether or not to merge in units of partitions, “the number of valid partitions adjacent to the execution of merge processing and NumMergeCandidates is greater than 1 and undefined in prediction of reference destination partition” Is set, a flag merge_direc_flag indicating a reference destination partition is set. When prediction by merge processing is selected, a flag for merge processing defined by the syntax described above is set. The operation of the third embodiment of the motion information reference destination prediction method in encoding will be described using the flowchart of FIG.

FIG. 26 is a flowchart for explaining a procedure for setting a flag for merge processing. First, it is determined whether or not the prediction method selected by the prediction selection unit 112 is prediction based on motion information selected in the merge process (S701). In the case of inter-screen prediction using intra-picture prediction or normal motion vector detection, merge_flag is set to 0 and the process ends (S703). If the prediction is based on the motion information selected in the merge process, merge_flag is set to 1 (S702).

Next, the number of motion information selected by the merge process is checked (S704). It is determined whether or not the number of adjacent partition candidates NumMergeCandidates to be specified by the reference destination specifying unit 113 is greater than one. When the number of candidates NumMergeCandidates is 1 or less, it is assumed that merge processing is selected, and there is always one adjacent partition of a valid reference destination, and either the left or the upper can be specified. Therefore, only merge_flag is set and the process ends.

If the number of candidates NumMergeCandidates is larger than 1, that is, if there are motion information candidates of two adjacent partitions, the process proceeds to prediction of a reference destination partition (S705). Here, in the prediction of the reference destination partition, the motion information reference destination prediction method described in the first embodiment is used. Based on information such as the size of the partition adjacent to the left and top of the target partition and the length of the side where the target partition and the adjacent partition are in contact with each other, the reference destination partition is predicted, and the flag pred_dirc_flag is set. Is output.

Like the merge_direc_flag, the flag pred_dirc_flag indicating the predicted reference destination partition represents the selection destination on the left (1) or above (0) as the adjacent partition of the reference destination, and the reference destination partition cannot be selected by prediction. Indeterminate (2) is set.

Next, the flag pred_direc_flag indicating the predicted reference destination partition set in this way is determined (S706). The flag pred_dirc_flag indicating the predicted reference destination partition is 2 when prediction is impossible in the prediction of the reference destination partition, and at this time, it depends on the merge process selected as the prediction method that actually minimizes the generated code amount. The process sets the merge_dir_flag indicating the reference partition of the motion information used in the motion compensation prediction, and the process ends (S707). On the other hand, when pred_direc_flag is not 2, pred_direc_flag directly indicates a reference destination partition, and when “0” is selected, a partition adjacent to the left is selected.

With the flag pred_dirc_flag indicating the success or failure of the prediction of the reference destination partition, it is unnecessary to transmit the flag merge_direc_flag indicating the reference destination partition as the probability that the adjacent partition is selected as the reference destination is higher. Therefore, the encoding efficiency is improved.

As described above, when motion-compensated prediction based on the motion information of adjacent partitions selected by the merge process for each partition is selected, the first flag inter_merge_flag described in the header of the SPS is described in the prediction block. The second flag merge_flag and the third flag merge_direc_flag are set and encoded.

Next, the operation of the reference destination prediction method in the case of decoding the encoded moving image bit stream based on the above-described syntax will be described.

FIG. 27 is a flowchart showing the selection procedure of the motion information selection unit 210, which will be described with reference to this figure. First, the number NumMergeCandidates of neighboring partitions adjacent to the partition to be processed is calculated (S801). Since the calculation procedure is the same as S701 in encoding, it is omitted.

It is determined whether or not the calculated number of candidates NumMergeCandidates is greater than 0 (S802). In the case of 0, since there is no adjacent partition that becomes the reference destination of the merge process, the motion information selection unit 210 outputs the motion information decoded from the bitstream as in the conventional case. When the number is larger than 0, merge_flag decoded by the variable length decoding unit 201 is read (S803), and the selection of motion information is determined based on the merge_flag (S804). When merge_flag is false (0), the motion information selection unit 210 outputs motion information decoded from the bitstream as in the conventional case. When merge_flag is true (1), the process proceeds to processing for selecting motion information of the reference partition in the merge processing.

Again, it is determined whether or not the number of candidates NumMergeCandidates is greater than one (S805). If the number of candidates NumMergeCandidates is one, one of them becomes a reference destination partition for the merge process, so the motion information of the reference destination partition is acquired from the memory 207 in which the decoded information is recorded (S807). The acquired motion information is supplied to the motion compensation unit 203, and motion compensation is performed based on the motion information.

If the number of candidates NumMergeCandidates is larger than 1, that is, two, the motion information of the partitions adjacent to the left and above the processing target partition is read from the memory 207 and acquired (S806). The obtained motion information of the left and upper adjacent partitions is compared (S808). Specifically, the prediction direction, the reference picture number, and the motion vector constituting the motion information are compared. If the motion information is exactly the same, the motion information of either the left or upper adjacent partition is acquired and output to the motion compensation unit 203 (S810). Here, the motion information of the partition adjacent to the left is selected as in the encoding.

If the motion information is different, the process proceeds to prediction of the reference destination partition (S809). For the prediction of the reference destination partition, the motion information reference destination prediction method described in the first embodiment is used, similarly to the encoding. The target partition is predicted based on information such as the size of the partition adjacent to the processing target partition and the left and top of the periphery, and the length of the side where the processing target partition and the adjacent partition are in contact with each other. The flag pred_dirc_flag indicating the referred reference partition is output.

Next, the pred_direc_flag output is determined (S811). The pred_dirc_flag is 2 when prediction is impossible due to prediction of the reference destination partition, and the reference destination partition cannot be determined. Therefore, merge_direc_flag decoded by the variable length decoding unit 201 is read (S812). Based on merge_direc_flag, the selection of the reference destination partition is determined (S813). If merge_dirc_flag is 1, the left adjacent partition (S814) is selected. If merge_dirc_flag is 0, the upper adjacent partition is selected as the reference destination partition (S815).

On the other hand, when pred_direc_flag is not 2, the value indicated by pred_direc_flag directly indicates the adjacent partition of the reference destination. That is, if pred_dirc_flag is 1, the left adjacent partition is selected, and if merge_proble_flag is 0, the upper adjacent partition is selected as a reference destination partition. The motion information of the reference destination partition thus selected is acquired, supplied to the motion compensation unit 203, and the process ends.

As described above, when inter-screen prediction by merge processing is selected according to each flag decoded by the variable-length decoding unit 201 and the calculation result in the decoding process, the reference is made by the above-described motion information reference destination prediction method. Select the destination partition. A decoded image is generated by adding the predicted image generated by motion compensation from the motion information of the selected reference destination partition and the residual signal decoded from the bitstream.

Furthermore, the determination processing based on the residual signal of the left or upper partition adjacent to the partition to be processed described in the second embodiment can be applied in the same manner.

As described above, according to the third embodiment, it is obtained by prediction of reference destination without defining merge_proble_flag itself, assuming that the matching accuracy of the determination of the adjacent block that is the reference destination by the prediction of the reference destination is high. The reference destination neighboring block may be indicated by the determination result, and the reference destination neighboring block may be indicated by the flag merge_dirc_flag only when the prediction of the reference destination is indefinite. In this case, the increase in the code amount due to merge_direct_flag can be offset by the reduction in the code amount required for merge_proble_flag, and the encoding efficiency can be relatively improved.

Furthermore, in the third embodiment described above, the left or upper adjacent partition is always designated as the reference destination partition in the prediction of the reference destination partition, so that it is not determined that the reference destination partition is indeterminate by the prediction of the reference destination partition. May be. When it is determined as indefinite in the prediction of the reference destination partition, it is predetermined in advance that the adjacent partition on the left or the upper side is uniquely specified. Since it is not necessary to encode and transmit the flag merge_dirc_flag indicating the reference destination partition when it is determined indefinitely by partition prediction, it is possible to improve the encoding efficiency. In the reference destination prediction method of motion information in decoding corresponding to encoding, the same reference destination partition as that on the encoding side is predicted, and the left or upper adjacent partition determined by prediction of the reference destination partition is used as the reference destination partition. select.

The moving image encoded stream output from the moving image encoding apparatus of the embodiment described above has a specific data format so that it can be decoded according to the encoding method used in the embodiment. Therefore, the moving picture decoding apparatus corresponding to the moving picture encoding apparatus can decode the encoded stream of this specific data format.

When a wired or wireless network is used to exchange an encoded stream between a moving image encoding device and a moving image decoding device, the encoded stream is converted into a data format suitable for the transmission form of the communication path. It may be transmitted. In that case, a video transmission apparatus that converts the encoded stream output from the video encoding apparatus into encoded data in a data format suitable for the transmission form of the communication channel and transmits the encoded data to the network, and receives the encoded data from the network Then, a moving image receiving apparatus that restores the encoded stream and supplies the encoded stream to the moving image decoding apparatus is provided.

The moving image transmitting apparatus is a memory that buffers the encoded stream output from the moving image encoding apparatus, a packet processing unit that packetizes the encoded stream, and transmission that transmits the packetized encoded data via the network. Part. The moving image receiving apparatus generates a coded stream by packetizing the received data, a receiving unit that receives the packetized coded data via a network, a memory that buffers the received coded data, and packet processing. And a packet processing unit provided to the video decoding device.

The above processing relating to encoding and decoding can be realized as a transmission, storage, and reception device using hardware, and is also stored in a ROM (Read Only Memory), a flash memory, or the like. It can also be realized by firmware or software such as a computer. The firmware program and software program can be provided by recording them on a computer-readable recording medium, provided from a server through a wired or wireless network, or provided as a data broadcast of terrestrial or satellite digital broadcasting. Is also possible.

The present invention has been described based on the embodiments. The embodiments are exemplifications, and it will be understood by those skilled in the art that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are within the scope of the present invention. .

101 motion vector detection unit, 102 subtraction unit, 103 orthogonal transform / quantization unit, 104 variable length coding unit, 105 inverse quantization / inverse orthogonal transform unit, 106 motion compensation unit, 107 weighted prediction unit, 108 addition unit, 109 deblocking filter unit, 110 memory, 111 intra prediction unit, 112 prediction selection unit, 113 reference destination designation unit, 201 variable length decoding unit, 202 inverse quantization / inverse orthogonal transform unit, 203 motion compensation unit, 204 weight Attached prediction unit, 205 addition unit, 206 deblocking filter unit, 207 memory, 208 intra-screen prediction unit, 209 intra-screen / inter-screen prediction selection unit, 210 motion information selection unit.

The present invention can be used for moving picture encoding and decoding techniques.

Claims

A moving image encoding apparatus that encodes the moving image using a motion vector in units of blocks obtained by dividing each picture of the moving image,
Refer to the motion information for any one of the neighboring blocks determined based on the neighboring condition between the coding target block and the neighboring block from among the plurality of coded neighboring blocks adjacent to the coding target block. A reference destination designating part that is designated as a reference destination block for
Among intra prediction, inter prediction using motion vector of the encoding target block, and inter prediction using motion vector of adjacent block, inter prediction using motion vector of adjacent block is used as a prediction method. When selected, the reference destination valid information indicating whether or not the adjacent block referred to by the selected prediction method matches the reference destination block specified by the reference destination specifying unit and the selected prediction method And a prediction selection unit that generates information to be displayed instead of motion information.
The reference destination designating unit determines, as the adjacency condition, whether or not the length of the adjacent side of the adjacent block is the same as the height or the width of the encoding target block. 2. The moving picture coding according to claim 1, wherein an adjacent block whose length is the same as a height or a width of the encoding target block is selected, and the selected adjacent block is designated as a reference block. apparatus.
The reference destination designating unit determines, as the adjacency condition, the size of the common part of adjacent sides of the encoding target block and the adjacent block by determining the size relationship of the common parts of adjacent sides. The moving picture encoding apparatus according to claim 1, wherein an adjacent block having a larger value is selected, and the selected adjacent block is designated as a reference block.
The reference destination designating unit obtains a boundary residual represented by an absolute value of a residual signal in an area of the adjacent block that is in contact with the encoding target block based on the residual signal of the adjacent block; 2. The moving picture code according to claim 1, wherein an adjacent block having a smaller boundary residual is selected by determining a magnitude relation of the residual, and the selected adjacent block is designated as a reference block. Device.
The reference destination specifying unit, based on the residual signal of the adjacent block, a boundary residual represented by an absolute value of a residual signal of the area of the adjacent block that is in contact with the encoding target block, and the code The residual ratio of the boundary residual to the total residual is obtained from the total residual represented by the absolute value of the residual signal of the adjacent block in contact with the block to be converted, and the magnitude relationship of the residual ratio is determined. The moving picture encoding apparatus according to claim 1, wherein an adjacent block having a smaller residual ratio is selected, and the selected adjacent block is designated as a reference block.
When the plurality of adjacent blocks having the same motion information are continuously arranged on any side of the encoding target block, the reference destination designating unit determines the length of the adjacent sides The moving image encoding apparatus according to claim 2 or 3, wherein the adjacency condition is determined as an addition of lengths of sides of the adjacent blocks.
When the inter-picture prediction using a motion vector of an adjacent block is selected as a prediction method, the prediction selection unit is configured to select the reference destination block when the reference destination block designated by the reference destination designation unit is indefinite. 7. The moving picture encoding apparatus according to claim 1, wherein reference destination information for specifying an adjacent block referred to by the selected prediction method is generated instead of the valid information.
When the inter-picture prediction using the motion vector of the adjacent block is selected as the prediction method, the prediction selection unit is preset in advance when the reference block specified by the reference destination specifying unit is indefinite. 7. The moving picture coding apparatus according to claim 1, wherein the moving picture coding apparatus is designated to refer to an adjacent block set as.
A moving image encoding method for encoding the moving image using a motion vector in units of blocks obtained by dividing each picture of the moving image,
Refer to the motion information for any one of the neighboring blocks determined based on the neighboring condition between the coding target block and the neighboring block from among the plurality of coded neighboring blocks adjacent to the coding target block. A reference destination specifying step that is designated as a reference destination block for
Among intra prediction, inter prediction using motion vector of the encoding target block, and inter prediction using motion vector of adjacent block, inter prediction using motion vector of adjacent block is used as a prediction method. If selected, the reference destination valid information indicating whether or not the adjacent block referred to by the selected prediction method matches the reference destination block specified by the reference destination specifying step, and the selected prediction method And a prediction selection step of generating information indicating the motion information instead of the motion information.
A moving image encoding program for encoding the moving image using a motion vector in units of blocks obtained by dividing each picture of the moving image,
Refer to the motion information for any one of the neighboring blocks determined based on the neighboring condition between the coding target block and the neighboring block from among the plurality of coded neighboring blocks adjacent to the coding target block. A reference destination specifying step that is designated as a reference destination block for
Among intra prediction, inter prediction using motion vector of the encoding target block, and inter prediction using motion vector of adjacent block, inter prediction using motion vector of adjacent block is used as a prediction method. If selected, the reference destination valid information indicating whether or not the adjacent block referred to by the selected prediction method matches the reference destination block specified by the reference destination specifying step, and the selected prediction method A motion picture encoding program that causes a computer to execute a prediction selection step of generating information to be displayed instead of motion information.
A moving picture decoding apparatus for decoding a bitstream in which the moving picture is encoded using a motion vector in units of blocks obtained by dividing each picture of the moving picture,
The information indicating the prediction method and the reference destination valid information are decoded from the bitstream including the information indicating the prediction method and the reference destination valid information indicating the validity of the adjacent block referred to by the selected prediction method. A variable length decoding unit;
From among intra-picture prediction, inter-picture prediction using a motion vector of a decoding target block, and inter-picture prediction using any one of a plurality of decoded adjacent blocks adjacent to the decoding target block, A prediction selector that selects one of the prediction methods based on the information indicating the prediction method;
When inter prediction using motion vectors of adjacent blocks is selected as a prediction method by the prediction selection unit, any one of the adjacent determinations determined based on an adjacent condition between the decoding target block and the adjacent block A motion image decoding apparatus comprising: a motion information selection unit that determines whether a block can be designated as a reference block for referring to motion information based on the reference destination valid information.
The motion information selection unit, as the adjacency condition, determines whether the length of the adjacent side of the adjacent block is the same as the height or width of the decoding target block, thereby determining the length of the adjacent side. The adjacent block whose length is the same as the height or width of the decoding target block is selected, and it is determined whether or not the selected adjacent block can be designated as a reference block. Video decoding device.
The motion information selection unit determines, as the adjacency condition, the size relationship between the lengths of the common parts of the adjacent sides of the decoding target block and the adjacent block, thereby determining the length of the common part of the adjacent sides. 12. The moving picture decoding apparatus according to claim 11, wherein a larger adjacent block is selected, and it is determined whether or not the selected adjacent block can be designated as a reference block.
The motion information selection unit obtains a boundary residual represented by an absolute value of a residual signal in a region of the adjacent block that is in contact with the decoding target block based on the residual signal of the adjacent block, and obtains a boundary residual. 12. The method according to claim 11, further comprising: selecting an adjacent block having a smaller boundary residual by determining a magnitude relationship of the differences, and determining whether the selected adjacent block can be designated as a reference block. The moving picture decoding apparatus described.
The motion information selection unit, based on the residual signal of the adjacent block, a boundary residual represented by an absolute value of a residual signal in the area of the adjacent block in contact with the decoding target block, and the decoding target By calculating the residual ratio of the boundary residual with respect to the total residual from the total residual represented by the absolute value of the residual signal of the adjacent block in contact with the block, and determining the magnitude relationship of the residual ratio, The moving picture decoding apparatus according to claim 11, wherein an adjacent block with a smaller residual ratio is selected, and it is determined whether or not the selected adjacent block can be designated as a reference block.
When the plurality of adjacent blocks having the same motion information are continuously arranged on any side of the decoding target block, the motion information selection unit determines that the lengths of the adjacent sides are continuous. The moving image decoding apparatus according to claim 12 or 13, wherein the adjacency condition is determined by adding the lengths of the sides of the adjacent blocks.
When the inter-picture prediction using a motion vector of an adjacent block is selected as a prediction method, the motion information selection unit is determined based on an adjacent condition between the decoding target block and the adjacent block The reference block is specified based on reference destination information that specifies an adjacent block to be referred to by the selected prediction method, instead of the reference destination valid information. To 16. The moving picture decoding apparatus according to any of claims 16 to 16.
When the inter-picture prediction using a motion vector of an adjacent block is selected as a prediction method, the motion information selection unit is determined based on an adjacent condition between the decoding target block and the adjacent block The moving picture decoding apparatus according to any one of claims 11 to 16, wherein the reference destination block is designated based on reference destination information for designating an adjacent block set in advance as a default value when is undefined. .
A moving picture decoding method for decoding a bitstream in which the moving picture is encoded using a motion vector in units of blocks obtained by dividing each picture of the moving picture,
The information indicating the prediction method and the reference destination valid information are decoded from the bitstream including the information indicating the prediction method and the reference destination valid information indicating the validity of the adjacent block referred to by the selected prediction method. A variable length decoding step;
From among intra-picture prediction, inter-picture prediction using a motion vector of a decoding target block, and inter-picture prediction using any one of a plurality of decoded adjacent blocks adjacent to the decoding target block, A prediction selection step of selecting one of the prediction methods based on the information indicating the prediction method;
When inter prediction using motion vectors of adjacent blocks is selected as a prediction method by the prediction selection step, any one of the adjacent determinations determined based on an adjacent condition between the decoding target block and the adjacent block A motion information decoding method comprising: a motion information selection step for determining whether a block can be designated as a reference block for referring to motion information based on the reference destination valid information.
A moving picture decoding program for decoding a bitstream in which the moving picture is encoded using a motion vector in units of blocks obtained by dividing each picture of the moving picture,
The information indicating the prediction method and the reference destination valid information are decoded from the bitstream including the information indicating the prediction method and the reference destination valid information indicating the validity of the adjacent block referred to by the selected prediction method. A variable length decoding step;
From among intra-picture prediction, inter-picture prediction using a motion vector of a decoding target block, and inter-picture prediction using any one of a plurality of decoded adjacent blocks adjacent to the decoding target block, A prediction selection step of selecting one of the prediction methods based on the information indicating the prediction method;
When inter prediction using motion vectors of adjacent blocks is selected as a prediction method by the prediction selection step, any one of the adjacent determinations determined based on an adjacent condition between the decoding target block and the adjacent block A moving picture decoding program that causes a computer to execute a motion information selection step of determining whether a block can be designated as a reference block for referring to motion information based on the reference destination valid information.