JP2013192081A - Moving image decoding device, moving image decoding method and moving image decoding program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To compress encoding information and reduce a code amount in an image encoding system using motion compensation prediction.SOLUTION: A decoding section 202 decodes information showing a prediction motion vector to be selected. A prediction motion vector candidate generation section 221 predicts a prediction motion vector from a motion vector of one of decoded adjacent blocks which are spatially or temporally adjacent to a decoding object block, and generates candidates of a plurality of prediction motion vectors. A prediction motion vector selecting section 225 selects a prediction motion vector from the candidates of the plurality of prediction motion vectors on the basis of the information showing the prediction motion vector to be selected. A prediction motion vector candidate generation section 221 does not derive the candidate of the prediction motion vector using a scaling operation from the motion vectors of the adjacent blocks at least on a group whose priority of a list registration order of the prediction motion vector is the highest among a plurality of groups formed of the decoded adjacent blocks which are spatially adjacent.

Description

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

  As a typical moving image compression encoding method, MPEG-4 AVC / H. There is a standard of H.264 (hereinafter referred to as AVC). In AVC, motion compensation is used in which a picture is divided into a plurality of rectangular blocks, an already encoded / decoded picture is used as a reference picture, and motion from the reference picture is predicted. A technique for predicting motion by this motion compensation is called inter prediction. In inter prediction in AVC, a plurality of pictures can be used as reference pictures, and the most suitable reference picture is selected from the plurality of reference pictures for each block to perform motion compensation. Therefore, a reference index is assigned to each reference picture, and the reference picture is specified by this reference index. Note that, for B pictures, a maximum of two reference pictures that have been encoded and decoded can be selected and used for inter prediction. The predictions from these two reference pictures are distinguished as L0 prediction (list 0 prediction) mainly used as forward prediction and L1 prediction (list 1 prediction) mainly used as backward prediction.

  Furthermore, bi-prediction using two inter predictions of L0 prediction and L1 prediction at the same time is also defined. In the case of bi-prediction, bi-directional prediction is performed, the inter-predicted signals of L0 prediction and L1 prediction are multiplied by a weighting coefficient, an offset value is added and superimposed, and a final inter-predicted image signal is obtained. Generate. As weighting coefficients and offset values used for weighted prediction, representative values are set for each reference picture in each list and encoded. The encoding information related to inter prediction includes, for each block, a prediction mode for distinguishing between L0 prediction and L1 prediction and bi-prediction, a reference index for specifying a reference picture for each reference list for each block, and a moving direction and a moving amount of the block. There is a motion vector that expresses and encodes and decodes the encoded information.

  In a moving picture coding system that performs motion compensation, a prediction process is performed on a motion vector in order to reduce the amount of code of the motion vector generated in each block. In AVC, by using the fact that the motion vector to be encoded has a strong correlation with the motion vector of the neighboring adjacent block, the prediction motion vector is calculated by performing prediction from the neighboring neighboring block, and the encoding target motion vector is calculated. A difference motion vector that is a difference between the motion vector and the predicted motion vector is calculated, and the difference motion vector is encoded to reduce the amount of codes.

  Specifically, as shown in FIG. 34A, a median value is calculated from the motion vectors of neighboring blocks A, B, and C in the vicinity to obtain a predicted motion vector, and the difference between the motion vector and the predicted motion vector By taking this, the amount of code of the motion vector is reduced. However, as shown in FIG. 34 (b), when the size and shape of the encoding target block and the adjacent block are different, when there are a plurality of adjacent blocks on the left, the top block among them is When there are a plurality of adjacent blocks, the leftmost block among them is set as a prediction block, and prediction is performed from the motion vector of the determined prediction block.

  In addition, the technique described in Patent Literature 1 improves the accuracy of a prediction vector by calculating a prediction vector from motion vectors and reference image information of a plurality of adjacent blocks of the processing target block, and increases the amount of code of the encoded vector. Is suppressed.

JP 2011-147172 A

  However, in the method described in Patent Document 1, only one prediction vector can be obtained. Therefore, depending on the image, the prediction accuracy of the prediction motion vector may be lowered, and the encoding efficiency may not be improved.

  Under such circumstances, the present inventors have come to recognize the necessity of further compressing the encoded information and reducing the overall code amount in the moving image encoding scheme using motion compensation prediction.

  The present invention has been made in view of such a situation, and an object of the present invention is to provide a moving picture decoding technique for improving the coding efficiency by calculating the prediction motion vector candidates and thereby reducing the code amount of the difference motion vector. Is to provide.

  In order to solve the above problems, a moving picture decoding apparatus according to an aspect of the present invention decodes a coded bit string obtained by coding the moving picture using motion compensation in units of blocks obtained by dividing each picture of the moving picture. A decoding unit (202) that is a moving image decoding apparatus that decodes information indicating a predicted motion vector to be selected together with a difference motion vector, and motion of a decoded adjacent block that is spatially or temporally adjacent to a decoding target block A plurality of prediction motion vector candidate generation units (221, 222) for predicting from a vector and generating a plurality of prediction motion vector candidates, and the plurality of predictions based on the decoded information indicating the prediction motion vector to be selected A predicted motion vector selection unit (225) for selecting a predicted motion vector from motion vector candidates; and the selected predicted motion vector and the difference motion And a motion vector calculation unit for calculating a motion vector used for motion compensation prediction (226) by adding the vector. The motion vector predictor candidate generation unit (221, 222) is a group having the highest priority in at least the motion vector list registration order among a plurality of groups formed by spatially adjacent decoded adjacent blocks. Does not derive prediction motion vector candidates from the motion vector of the adjacent block, and at least for the group with the lowest priority of the motion vector list registration order, the scaling operation is performed from the motion vector of the adjacent block. Derivation of motion vector predictor candidates using is allowed.

  Another aspect of the present invention is a video decoding method. This method is a moving picture decoding method for decoding a coded bit sequence in which the moving picture is coded using motion compensation in units of blocks obtained by dividing each picture of the moving picture, and indicates a prediction motion vector to be selected. A decoding step for decoding information together with a difference motion vector, and a prediction motion vector for generating a plurality of motion vector predictor candidates by predicting from a motion vector of a decoded adjacent block spatially or temporally adjacent to a decoding target block A candidate generation step; a prediction motion vector selection step of selecting a prediction motion vector from the plurality of prediction motion vector candidates based on the decoded information indicating the prediction motion vector to be selected; and the selected prediction motion A motion vector used for motion compensation prediction is calculated by adding the vector and the difference motion vector. And a motion vector calculation step. In the prediction motion vector candidate generation step, at least a group having a highest priority in the list registration order of prediction motion vectors among a plurality of groups formed by spatially adjacent decoded adjacent blocks Predictive motion vector candidates that do not derive scaling motion vectors from motion vectors and that have the lowest priority in the motion vector list registration order at least for predicted motion vectors that use scaling operations from adjacent block motion vectors Allow derivation of vector candidates.

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

  According to the present invention, a plurality of prediction motion vectors are calculated, an optimal prediction motion vector is selected from the plurality of prediction motion vectors, and the amount of generated code of the difference motion vector is reduced, thereby improving the encoding efficiency. Can be made.

It is a block diagram which shows the structure of the moving image encoder which performs the motion vector prediction method which concerns on embodiment. It is a block diagram which shows the structure of the moving image decoding apparatus which performs the motion vector prediction method which concerns on embodiment. It is a figure explaining a tree block and an encoding block. It is a figure explaining the division mode of a prediction block. It is a figure explaining a prediction block group. It is a figure explaining a prediction block group. It is a figure explaining a prediction block group. It is a figure explaining a prediction block group. It is a figure explaining a prediction block group. It is a figure explaining the syntax of the bit stream in the slice level regarding the prediction method of a motion vector. It is a figure explaining the syntax of the bit stream in the prediction block level regarding the prediction method of a motion vector. It is a block diagram which shows the detailed structure of the difference motion vector calculation part of FIG. FIG. 3 is a block diagram illustrating a detailed configuration of a motion vector calculation unit in FIG. 2. It is a flowchart which shows the difference motion vector calculation process procedure of the difference motion vector calculation part of FIG. It is a flowchart which shows the motion vector calculation process procedure of the motion vector calculation part of FIG. It is a flowchart which shows the process of derivation | leading-out of the motion vector predictor candidate, and a motion vector predictor list construction process. It is a flowchart which shows the candidate derivation | leading-out process procedure of the motion vector predictor of 1st embodiment. It is a flowchart which shows the candidate derivation | leading-out process procedure of a motion vector predictor. It is a flowchart which shows the candidate derivation | leading-out process procedure of the motion vector predictor of 1st embodiment. It is a flowchart which shows the scaling calculation process procedure of a motion vector. It is a flowchart which shows the scaling calculation process procedure of the motion vector by integer calculation. It is a flowchart which shows the prediction derivation | leading-out process procedure of the motion vector predictor of 2nd embodiment. It is a flowchart which shows the prediction derivation | leading-out process procedure of the motion vector predictor of 2nd embodiment. It is a flowchart which shows the candidate derivation | leading-out process procedure of a motion vector predictor. It is a flowchart which shows the derivation | leading-out process of the picture of a different time. It is a flowchart which shows the prediction block candidate derivation | leading-out process of the picture of a different time. It is a flowchart which shows the candidate derivation | leading-out process procedure of a motion vector predictor. It is a flowchart which shows the candidate derivation | leading-out process procedure of a motion vector predictor. It is a flowchart which shows the scaling calculation process procedure of a motion vector. It is a flowchart which shows the scaling calculation process procedure of the motion vector by integer calculation. It is a flowchart which shows the registration process procedure of the candidate of a motion vector predictor to a motion vector predictor candidate list. It is a flowchart which shows the deletion process procedure of the redundant motion vector predictor candidate from a motion vector predictor candidate list. It is a flowchart which shows the restriction | limiting process procedure of the number of prediction motion vector candidates. It is a figure explaining the calculation method of the conventional prediction motion vector.

  In the present embodiment, with regard to moving picture coding, in particular, to improve coding efficiency in moving picture coding in which a picture is divided into rectangular blocks of an arbitrary size and shape and motion compensation is performed in units of blocks between pictures. In addition, by calculating a plurality of predicted motion vectors from the motion vectors of the surrounding blocks that have been encoded, and calculating and encoding a difference vector between the motion vector of the block to be encoded and the selected predicted motion vector Reduce the amount of code. Alternatively, the amount of code is reduced by estimating the encoding information of the encoding target block by using the encoding information of the surrounding blocks that have already been encoded. Also, in the case of decoding a moving image, a plurality of predicted motion vectors are calculated from the motion vectors of the peripheral blocks that have been decoded, and a decoding target is calculated from the difference vector decoded from the encoded stream and the selected predicted motion vector. The motion vector of the block is calculated and decoded. Alternatively, the encoding information of the decoding target block is estimated by using the encoding information of the peripheral blocks that have been decoded.

  First, techniques used in the present embodiment and technical terms are defined.

(About tree blocks and coding blocks)
In the embodiment, as shown in FIG. 3, the picture is equally divided into square units of any same size. This unit is defined as a tree block, and is an encoding / decoding target block in a picture (an encoding target block in the encoding process and a decoding target block in the decoding process. Hereinafter, unless otherwise specified, It is used as a basic unit of address management for specifying. Except for monochrome, the tree block is composed of one luminance signal and two color difference signals. The size of the tree block can be freely set to a power of 2 depending on the picture size and the texture in the picture. In order to optimize the encoding process according to the texture in the picture, the tree block divides the luminance signal and chrominance signal in the tree block hierarchically into four parts (two parts vertically and horizontally) as necessary, The block can be made smaller in block size. Each block is defined as a coding block, and is a basic unit of processing when performing coding and decoding. Except for monochrome, the coding block is also composed of one luminance signal and two color difference signals. The maximum size of the coding block is the same as the size of the tree block. An encoded block having the minimum size of the encoded block is called a minimum encoded block, and can be freely set to a power of 2.

  In FIG. 3, the encoding block A is a single encoding block without dividing the tree block. The encoding block B is an encoding block formed by dividing a tree block into four. The coding block C is a coding block obtained by further dividing the block obtained by dividing the tree block into four. The coding block D is a coding block obtained by further dividing the block obtained by dividing the tree block into four parts and further dividing the block into four hierarchically, and is a coding block of the minimum size.

(About prediction mode)
Intra-prediction (MODE_INTRA) in which prediction is performed from an encoded / decoded surrounding image signal in the same picture as the encoding target block, and an encoded / decoded image signal of a picture different from the encoding target block, in units of encoding block Switch the inter prediction (MODE_INTER) to perform prediction from. A mode for identifying the intra prediction (MODE_INTRA) and the inter prediction (MODE_INTER) is defined as a prediction mode (PredMode). The prediction mode (PredMode) has intra prediction (MODE_INTRA) or inter prediction (MODE_INTER) as a value, and can be selected and encoded.

(About split mode, prediction block, prediction unit)
When performing intra prediction (MODE_INTRA) and inter prediction (MODE_INTER) by dividing the picture into blocks, the coded block is divided as necessary to reduce the unit for switching the intra prediction and inter prediction methods. Make predictions. A mode for identifying the division method of the luminance signal and the color difference signal of the coding block is defined as a division mode (PartMode). Furthermore, this divided block is defined as a prediction block. As shown in FIG. 4, four types of partition modes (PartMode) are defined according to the method of dividing the luminance signal of the coding block. The division mode (PartMode) of what is regarded as one prediction block without dividing the luminance signal of the coding block (FIG. 4A) is 2N × 2N division (PART_2Nx2N), and the luminance signal of the coding block is horizontally The division mode (PartMode) of the two prediction blocks (FIG. 4B) is divided into 2N × N divisions (PART_2NxN), the luminance signal of the encoded block is divided in the vertical direction, and the encoded block is The division mode (PartMode) of the two prediction blocks (FIG. 4 (c)) is N × 2N division (PART_Nx2N), and the luminance signal of the encoded block is divided into four prediction blocks by horizontal and vertical equal divisions. The division mode (PartMode) of (FIG. 4D) is defined as N × N division (PART_NxN), respectively. Note that the color difference signal is also divided in the same manner as the vertical and horizontal division ratios of the luminance signal for each division mode (PartMode).

  In order to specify each prediction block within the coding block, a number starting from 0 is assigned to the prediction block existing inside the coding block in the coding order. This number is defined as a split index PartIdx. A number described in each prediction block of the encoded block in FIG. 4 represents a partition index PartIdx of the prediction block. In the 2N × N division (PART_2NxN) shown in FIG. 4B, the division index PartIdx of the upper prediction block is set to 0, and the division index PartIdx of the lower prediction block is set to 1. In the N × 2N division (PART_Nx2N) shown in FIG. 4C, the division index PartIdx of the left prediction block is set to 0, and the division index PartIdx of the right prediction block is set to 1. In the N × N partition (PART_NxN) shown in FIG. 4D, the partition index PartIdx of the upper left prediction block is 0, the partition index PartIdx of the upper right prediction block is 1, and the partition index PartIdx of the lower left prediction block is 2. And the division index PartIdx of the prediction block at the lower right is set to 3.

  When the prediction mode (PredMode) is inter prediction (MODE_INTER), except for the coding block D which is the smallest coding block, the partition mode (PartMode) is 2N × 2N partition (PART_2Nx2N), 2N × N partition (PART_2NxN), and N × 2N partition (PART_Nx2N) is defined, and only the coding block D which is the smallest coding block, the partition mode (PartMode) is 2N × 2N partition (PART_2Nx2N), 2N × N partition (PART_2NxN), and N × 2N In addition to the division (PART_Nx2N), N × N division (PART_NxN) is defined. The reason why N × N division (PART_NxN) is not defined other than the smallest coding block is that, except for the smallest coding block, the coding block can be divided into four to represent a small coding block.

(Position of tree block, coding block, prediction block, transform block)
The position of each block including the tree block, the encoding block, the prediction block, and the transform block according to the present embodiment has the position of the pixel of the luminance signal at the upper left of the luminance signal screen as the origin (0, 0). The pixel position of the upper left luminance signal included in each block area is represented by two-dimensional coordinates (x, y). The direction of the coordinate axis is a right direction in the horizontal direction and a downward direction in the vertical direction, respectively, and the unit is one pixel unit of the luminance signal. Of course, the luminance signal and the color difference signal have the same image size (number of pixels) and the color difference format is 4: 4: 4. Of course, the luminance signal and the color difference signal have a different color size format of 4: 4. Even in the case of 2: 0, 4: 2: 2, the position of each block of the color difference signal is represented by the coordinates of the pixel of the luminance signal included in the block area, and the unit is one pixel of the luminance signal. In this way, not only can the position of each block of the color difference signal be specified, but also the relationship between the positions of the luminance signal block and the color difference signal block can be clarified only by comparing the coordinate values.

(About prediction block groups)
A group composed of a plurality of prediction blocks is defined as a prediction block group. 5, FIG. 6, FIG. 7 and FIG. 8 are diagrams for explaining a prediction block group adjacent to a prediction block to be encoded / decoded within the same picture as the prediction block to be encoded / decoded. FIG. 9 shows an already encoded / decoded picture that exists at the same position as or near the predicted block to be encoded / decoded in a picture that has been encoded / decoded that is temporally different from the predicted block to be encoded / decoded. It is a figure explaining this prediction block group. The prediction block group will be described with reference to FIGS. 5, 6, 7, 8, and 9.

  As shown in FIG. 5, the prediction block A1 adjacent to the left side of the prediction block to be encoded / decoded within the same picture as the prediction block to be encoded / decoded, and the prediction block to be encoded / decoded A first prediction block group including a prediction block A0 adjacent to the lower left vertex is defined as a prediction block group adjacent to the left side.

  As shown in FIG. 6, even when the size of the prediction block adjacent to the left side of the prediction block to be encoded / decoded is larger than the prediction block to be encoded / decoded, the prediction adjacent to the left side according to the above condition is used. If the block A is adjacent to the left side of the prediction block to be encoded / decoded, the prediction block A1 is used. If the block A is adjacent to the lower left vertex of the prediction block to be encoded / decoded, the prediction block A0 is set. . In the example of FIG. 6, the prediction block A0 and the prediction block A1 are the same prediction block.

  In addition, as shown in FIG. 7, when the size of the prediction block adjacent to the left side of the prediction block to be encoded / decoded is smaller than the prediction block to be encoded / decoded and there are a plurality of prediction blocks, this embodiment In the embodiment, only the lowest prediction block A10 among the prediction blocks adjacent to the left side is set as the prediction block A1 adjacent to the left side.

  In almost the same way as defining the prediction block group adjacent to the left side, the prediction block B1 adjacent to the upper side of the prediction block to be encoded / decoded within the same picture as the prediction block to be encoded / decoded is encoded. The second prediction block group composed of the prediction block B0 adjacent to the upper right vertex of the prediction block to be encoded / decoded and the prediction block B2 adjacent to the upper left vertex of the prediction block to be encoded / decoded are It is defined as an adjacent prediction block group.

  In addition, as shown in FIG. 8, even when the size of the prediction block adjacent to the upper side of the prediction block to be encoded / decoded is larger than the prediction block to be encoded / decoded, the prediction adjacent to the upper side according to the above condition. If the block B is adjacent to the upper side of the prediction block to be encoded / decoded, the prediction block B1 is set. If the block B is adjacent to the upper right vertex of the prediction block to be encoded / decoded, the prediction block B0 is set. If it is adjacent to the top left vertex of the prediction block to be encoded / decoded, the prediction block is B2. In the example of FIG. 8, the prediction block B0, the prediction block B1, and the prediction block B2 are the same prediction block.

  As shown in FIG. 7, when the size of the prediction block adjacent to the upper side of the prediction block to be encoded / decoded is smaller than the prediction block to be encoded / decoded, and there are a plurality of prediction blocks, the embodiment In FIG. 4, only the rightmost prediction block B10 among the prediction blocks adjacent on the upper side is set as the prediction block B1 adjacent on the upper side.

  The difference from the prediction block group adjacent to the left side is that the prediction block B2 adjacent to the upper left vertex of the prediction block to be encoded / decoded is included in the prediction block group adjacent to the upper side. The upper left prediction block B2 may be included in any prediction block group, but here is included in the prediction block group adjacent on the upper side. Therefore, the prediction block group adjacent on the upper side has a larger number of prediction blocks than the prediction block group adjacent on the left side.

  As shown in FIG. 9, in an encoded / decoded picture that is temporally different from the prediction block to be encoded / decoded, an existing code that exists at the same position as the prediction block to be encoded / decoded or a position in the vicinity thereof. A third prediction block group composed of prediction / decoding prediction block groups T0 and T1 is defined as a prediction block group at different times.

(Inter prediction mode, reference list)
In the embodiment of the present invention, in inter prediction in which prediction is performed from an image signal of a coded / decoded picture, a plurality of decoded pictures can be used as reference pictures. In order to identify a reference picture selected from a plurality of reference pictures, a reference index is attached to each prediction block. In the B slice, any two reference pictures can be selected for each prediction block, and inter prediction can be performed. As inter prediction modes, there are L0 prediction (Pred_L0), L1 prediction (Pred_L1), and bi-prediction (Pred_BI). . The reference picture is managed by L0 (reference list 0) and L1 (reference list 1) of the list structure, and the reference picture can be specified by designating the reference index of L0 or L1. L0 prediction (Pred_L0) is inter prediction that refers to a reference picture managed in L0, L1 prediction (Pred_L1) is inter prediction that refers to a reference picture managed in L1, and bi-prediction (Pred_BI) is This is inter prediction in which both L0 prediction and L1 prediction are performed and one reference picture managed by each of L0 and L1 is referred to. Only L0 prediction can be used in inter prediction of P slice, and L0 prediction, L1 prediction, and bi-prediction (Pred_BI) that averages or weights and adds L0 prediction and L1 prediction can be used in inter prediction of B slice. In the subsequent processing, regarding constants and variables with subscript LX (X is 0 or 1) in the output, it is assumed that processing is performed for each of L0 and L1.

(About POC)
POC is a variable associated with the picture to be encoded, and is set to a value that increases by 1 in the picture output / display order. Based on the POC value, it is possible to determine whether they are the same picture, to determine the front-to-back relationship between pictures in the output / display order, or to derive the distance between pictures. For example, if the POCs of two pictures have the same value, it can be determined that they are the same picture. When the POCs of two pictures have different values, it can be determined that the picture with the smaller POC value is a picture that is output / displayed earlier in time, and the difference between the POCs of the two pictures is the time axis direction. Indicates the distance between pictures.

(Embodiment 1)
Embodiment 1 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 coding apparatus according to an embodiment of the present invention. The moving image encoding apparatus according to the embodiment includes an image memory 101, a motion vector detection unit 102, a difference motion vector calculation unit 103, an inter prediction information estimation unit 104, a motion compensation prediction unit 105, a prediction method determination unit 106, a residual signal. Generation unit 107, orthogonal transform / quantization unit 108, first encoded bit string generation unit 109, second encoded bit string generation unit 110, multiplexing unit 111, inverse quantization / inverse orthogonal transform unit 112, decoded image signal A superimposing unit 113, an encoded information storage memory 114, and a decoded image memory 115 are provided.

  The image memory 101 temporarily stores image signals of pictures to be encoded supplied in order of display time. The image memory 101 supplies the stored image signal of the picture to be encoded to the motion vector detection unit 102, the prediction method determination unit 106, and the residual signal generation unit 107 in units of predetermined pixel blocks. At this time, the image signals of the pictures stored in the order of display time are rearranged in the encoding order and output from the image memory 101 in units of pixel blocks.

  The motion vector detecting unit 102 uses the motion vector of each prediction block size and each prediction mode by block matching between the image signal supplied from the image memory 101 and the reference picture supplied from the decoded image memory 115 in units of prediction blocks. And the detected motion vector is supplied to the motion compensation prediction unit 105, the difference motion vector calculation unit 103, and the prediction method determination unit 106.

  The difference motion vector calculation unit 103 calculates a plurality of motion vector predictor candidates by using the encoded information of the already encoded image signal stored in the encoded information storage memory 114, and will be described later. The motion vector detection unit 102 selects an optimal motion vector predictor from a plurality of motion vector predictor candidates registered in the motion vector list and registered in the motion vector predictor list, and a motion vector difference is detected from the motion vector detected by the motion vector detection unit 102 and the motion vector predictor. And the calculated difference motion vector is supplied to the prediction method determination unit 106. Furthermore, a prediction motion vector index that identifies a prediction motion vector selected from prediction motion vector candidates registered in the prediction motion vector list is supplied to the prediction method determination unit 106. The detailed configuration and operation of the difference motion vector calculation unit 103 will be described later.

  The inter prediction information estimation unit 104 estimates inter prediction information in merge mode. The merge mode refers to inter prediction information such as a prediction mode of the prediction block, a reference index (information for specifying a reference picture used for motion compensation prediction from a plurality of reference pictures registered in the reference list), a motion vector, and the like. Is a mode in which inter prediction information of an adjacent inter-predicted prediction block that has been encoded or an inter-predicted prediction block of a different picture is used. A plurality of merge candidates (inter prediction information candidates) are calculated using the encoded information of the already encoded prediction block stored in the encoded information storage memory 114, registered in the merge candidate list, and merged. An optimal merge candidate is selected from a plurality of merge candidates registered in the candidate list, and inter prediction information such as a prediction mode, a reference index, and a motion vector of the selected merge candidate is supplied to the motion compensation prediction unit 105. The merge index that identifies the selected merge candidate is supplied to the prediction method determination unit 106.

  The motion compensation prediction unit 105 generates a prediction image signal by motion compensation prediction from a reference picture using the motion vectors detected by the motion vector detection unit 102 and the inter prediction information estimation unit 104, and uses the prediction image signal as a prediction method determination unit. 106. In L0 prediction and L1 prediction, one-way prediction is performed. In the case of bi-prediction (Pred_BI), bi-directional prediction is performed, the inter-predicted signals of L0 prediction and L1 prediction are adaptively multiplied by weighting factors, offset values are added and superimposed, and finally A predicted image signal is generated.

  The prediction method determination unit 106 evaluates the coding amount of the difference motion vector of a plurality of prediction methods, the distortion amount between the prediction image signal and the image signal, and the like, thereby performing inter prediction (PRED_INTER) or intra prediction for each coding block. (PRED_INTRA) prediction mode PredMode and split mode PartMode are determined, and inter prediction (PRED_INTER) determines the prediction method such as whether or not the merge mode for each prediction block. In merge mode, merge index, merge When the mode is not the mode, the inter prediction flag, the prediction motion vector index, the reference indexes of L0 and L1, the difference motion vector, and the like are determined, and the encoded information corresponding to the determination is supplied to the first encoded bit string generation unit 109.

  Further, the prediction method determination unit 106 stores information indicating the determined prediction method and encoded information including a motion vector according to the determined prediction method in the encoded information storage memory 114. The encoded information stored here is the motion of the reference indexes refIdxL0, refIdxL1, L0, and L1 of the flags predFlagL0, predFlagL1, L0, and L1 indicating whether to use the prediction mode PredMode, the partition mode PartMode, the L0 prediction, and the L1 prediction Vectors mvL0, mvL1, etc. Here, when the prediction mode PredMode is intra prediction (MODE_INTRA), a flag predFlagL0 indicating whether to use L0 prediction and a flag predFlagL1 indicating whether to use L1 prediction are both 0. On the other hand, when the prediction mode PredMode is inter prediction (MODE_INTER) and the inter prediction mode is L0 prediction (Pred_L0), a flag predFlagL0 indicating whether to use L0 prediction is 1, and a flag predFlagL1 indicating whether to use L1 prediction Is 0. When the inter prediction mode is L1 prediction (Pred_L1), the flag predFlagL0 indicating whether to use L0 prediction is 0, and the flag predFlagL1 indicating whether to use L1 prediction is 1. When the inter prediction mode is bi-prediction (Pred_BI), a flag predFlagL0 indicating whether to use L0 prediction and a flag predFlagL1 indicating whether to use L1 prediction are both 1. The prediction method determination unit 106 supplies a prediction image signal corresponding to the determined prediction mode to the residual signal generation unit 107 and the decoded image signal superposition unit 113.

  The residual signal generation unit 107 generates a residual signal by performing subtraction between the image signal to be encoded and the predicted image signal, and supplies the residual signal to the orthogonal transform / quantization unit 108.

  The orthogonal transform / quantization unit 108 performs orthogonal transform and quantization on the residual signal in accordance with the quantization parameter to generate an orthogonal transform / quantized residual signal, and a second encoded bit string generation unit 110 and the inverse quantization / inverse orthogonal transform unit 112. Further, the orthogonal transform / quantization unit 108 stores the quantization parameter in the encoded information storage memory 114.

  The first encoded bit string generation unit 109 adds information corresponding to the prediction method determined by the prediction method determination unit 106 for each encoding block and prediction block, in addition to the information in units of sequences, pictures, slices, and encoding blocks. Encoding information is encoded. Specifically, in the prediction mode PredMode, partition mode PartMode, and inter prediction (PRED_INTER) for each coding block, a flag that determines whether or not the mode is merge mode, merge index in the case of merge mode, and inter prediction in the case of not in merge mode Encoding information such as information on the mode, predicted motion vector index, and difference motion vector is entropy encoded according to a prescribed syntax rule described later to generate a first encoded bit string, which is supplied to the multiplexing unit 111.

  The second encoded bit string generation unit 110 generates a second encoded bit string by entropy-encoding the orthogonally transformed and quantized residual signal according to a prescribed syntax rule, and supplies the second encoded bit string to the multiplexing unit 111. The multiplexing unit 111 multiplexes the first encoded bit string and the second encoded bit string in accordance with a prescribed syntax rule, and outputs a bit stream.

The inverse quantization / inverse orthogonal transform unit 112 calculates a residual signal by performing inverse quantization and inverse orthogonal transform on the orthogonal transform / quantized residual signal supplied from the orthogonal transform / quantization unit 108 to perform decoding. This is supplied to the image signal superimposing unit 113.
The decoded image signal superimposing unit 113 superimposes the predicted image signal according to the determination by the prediction method determining unit 106 and the residual signal that has been inverse quantized and inverse orthogonal transformed by the inverse quantization / inverse orthogonal transform unit 112 to decode the decoded image. Is generated and stored in the decoded image memory 115. The decoded image may be stored in the decoded image memory 115 after filtering processing for reducing distortion such as block distortion due to encoding.

  FIG. 2 is a block diagram showing the configuration of the moving picture decoding apparatus according to the embodiment of the present invention corresponding to the moving picture encoding apparatus of FIG. The moving picture decoding apparatus according to the embodiment includes a separation unit 201, a first encoded bit string decoding unit 202, a second encoded bit string decoding unit 203, a motion vector calculation unit 204, an inter prediction information estimation unit 205, and a motion compensation prediction unit 206. , An inverse quantization / inverse orthogonal transform unit 207, a decoded image signal superimposing unit 208, an encoded information storage memory 209, and a decoded image memory 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 the motion compensation prediction unit 206 in FIG. The configurations of the inverse orthogonal transform unit 207, the decoded image signal superimposing unit 208, the encoded information storage memory 209, and the decoded image memory 210 are the same as those of the motion compensation prediction unit 105, the inverse quantization / inverse of the moving image encoding device in FIG. The orthogonal transform unit 112, the decoded image signal superimposing unit 113, the encoded information storage memory 114, and the decoded image memory 115 have functions corresponding to the respective configurations.

  The bit stream supplied to the separation unit 201 is separated according to a rule of a prescribed syntax, and the separated encoded bit string is supplied to the first encoded bit string decoding unit 202 and the second encoded bit string decoding unit 203.

  The first encoded bit string decoding unit 202 decodes the supplied encoded bit string to obtain sequence, picture, slice, encoded block unit information, and predicted block unit encoded information. Specifically, in the prediction mode PredMode for determining whether the prediction is inter prediction (PRED_INTER) or intra prediction (PRED_INTRA) for each coding block, split mode PartMode, and inter prediction (PRED_INTER), a flag for determining whether the mode is merge mode, When the merge mode is selected, the merge index is decoded. When the merge mode is not selected, the encoded information related to the inter prediction mode, the predicted motion vector index, the difference motion vector, and the like is decoded according to a predetermined syntax rule to be described later, and the encoded information is a motion vector calculation unit. 204 or the inter prediction information estimation unit 205.

  The second encoded bit string decoding unit 203 calculates a residual signal that has been orthogonally transformed / quantized by decoding the supplied encoded bit string, and dequantized / inverted the residual signal that has been orthogonally transformed / quantized. This is supplied to the orthogonal transform unit 207.

  When the prediction block to be decoded is not in the merge mode, the motion vector calculation unit 204 uses the encoded information of the already decoded image signal stored in the encoded information storage memory 209 to generate a plurality of prediction motion vector candidates. And the motion vector is registered in a motion vector predictor list, which will be described later, and a motion vector predictor that is decoded and supplied from the plurality of motion vector predictor candidates registered in the motion vector predictor list by the first encoded bit string decoding unit 202. A prediction motion vector corresponding to the index is selected, a motion vector is calculated from the difference vector decoded by the first encoded bit string decoding unit 202 and the selected prediction motion vector, and the motion compensated prediction unit 206 together with other encoded information. And is stored in the encoded information storage memory 209. Here, the encoding information of the prediction block to be supplied / stored includes the prediction indexes PredFlagL0, predFlagL1, L0, and L1 reference indexes refIdxL0, refIdxL1, indicating whether to use the prediction mode PredMode, the partition mode PartMode, the L0 prediction, and the L1 prediction L0, L1 motion vectors mvL0, mvL1, etc. Here, when the prediction mode PredMode is intra prediction (MODE_INTRA), a flag predFlagL0 indicating whether to use L0 prediction and a flag predFlagL1 indicating whether to use L1 prediction are both 0. On the other hand, when the prediction mode PredMode is inter prediction (MODE_INTER) and the inter prediction mode is L0 prediction (Pred_L0), a flag predFlagL0 indicating whether to use L0 prediction is 1, and a flag predFlagL1 indicating whether to use L1 prediction Is 0. When the inter prediction mode is L1 prediction (Pred_L1), the flag predFlagL0 indicating whether to use L0 prediction is 0, and the flag predFlagL1 indicating whether to use L1 prediction is 1. When the inter prediction mode is bi-prediction (Pred_BI), a flag predFlagL0 indicating whether to use L0 prediction and a flag predFlagL1 indicating whether to use L1 prediction are both 1. The detailed configuration and operation of the motion vector calculation unit 204 will be described later.

  The inter prediction information estimation unit 205 estimates the inter prediction information in the merge mode when the prediction block to be decoded is in the merge mode. Using the encoded information of the already decoded prediction block stored in the encoded information storage memory 114, a plurality of merge candidates are calculated and registered in the merge candidate list, and the plurality of merge candidates registered in the merge candidate list The merge candidate corresponding to the merge index decoded and supplied by the first encoded bit string decoding unit 202 is selected from the merge candidates, and the prediction mode PredMode, split mode PartMode, L0 prediction, and L1 prediction of the selected merge candidate are selected. Inter prediction information such as a flag indicating whether or not to use, reference indexes of L0 and L1, motion vectors of L0 and L1, and the like are supplied to the motion compensation prediction unit 206 and stored in the encoded information storage memory 209.

  The motion compensation prediction unit 206 generates a prediction image signal by motion compensation prediction from the reference picture using the motion vector calculated by the motion vector calculation unit 204, and supplies the prediction image signal to the decoded image signal superimposition unit 208. In the case of bi-prediction (Pred_BI), a weighted coefficient is adaptively multiplied and superimposed on the two motion-compensated predicted image signals of L0 prediction and L1 prediction to generate a final predicted image signal.

  The inverse quantization / inverse orthogonal transform unit 207 performs inverse orthogonal transform and inverse quantization on the orthogonal transform / quantized residual signal decoded by the first encoded bit string decoding unit 202, and performs inverse orthogonal transform / An inverse quantized residual signal is obtained.

  The decoded image signal superimposing unit 208 superimposes the predicted image signal subjected to motion compensation prediction by the motion compensation prediction unit 206 and the residual signal subjected to inverse orthogonal transform / inverse quantization by the inverse quantization / inverse orthogonal transform unit 207. As a result, the decoded image signal is decoded and stored in the decoded image memory 210. When stored in the decoded image memory 210, the decoded image may be stored in the decoded image memory 210 after filtering processing for reducing block distortion or the like due to encoding is performed on the decoded image.

(About syntax)
Next, a syntax that is a common rule for encoding and decoding a bit stream of a moving image that is encoded by a moving image encoding device including the motion vector prediction method according to the present embodiment and decoded by the decoding device explain.

  FIG. 10 shows a first syntax structure described in the slice header in units of slices of the bitstream generated according to the present embodiment. However, only syntax elements relevant to the present embodiment are shown. When the slice type is B, a flag indicating whether a picture colPic at a different time used for calculating a temporal motion vector predictor candidate or a merge candidate is defined in the L0 reference list or the L1 reference list collocated_from_l0_flag is set. Details of the flag collocated_from_l0_flag will be described later.

  Note that the above syntax elements may be placed in a picture parameter set describing syntax elements set in units of pictures.

  FIG. 11 shows a syntax pattern described in units of prediction blocks. When the value of the prediction mode PredMode of the prediction block is inter prediction (MODE_INTER), merge_flag [x0] [y0] indicating whether the mode is the merge mode is set. Here, x0 and y0 are indices indicating the position of the upper left pixel of the prediction block in the picture of the luminance signal, and merge_flag [x0] [y0] is the prediction block located at (x0, y0) in the picture It is a flag indicating whether or not merge mode.

  Next, when merge_flag [x0] [y0] is 1, it indicates merge mode, and a merge list index syntax element merge_idx [x0] [y0] is set, which is a list of merge candidates to be referred to. . Here, x0 and y0 are indices indicating the position of the upper left pixel of the prediction block in the picture, and merge_idx [x0] [y0] is the merge index of the prediction block located at (x0, y0) in the picture. is there.

  On the other hand, when merge_flag [x0] [y0] is 0, this indicates that the mode is not merge mode. When the slice type is B, a syntax element inter_pred_flag [x0] [y0] for identifying the inter prediction mode is installed. The L0 prediction (Pred_L0), L1 prediction (Pred_L1), and bi-prediction (Pred_BI) are identified by the tax element. According to the inter prediction mode, the reference index syntax elements ref_idx_l0 [x0] [y0], ref_idx_l1 [x0] [y0] for specifying the reference picture are obtained by motion vector detection for each of L0 and L1. The differential motion vector syntax elements mvd_l0 [x0] [y0] [j] and mvd_l1 [x0] [y0] [j], which are the differences between the motion vector of the prediction block and the prediction motion vector, are installed. Here, x0 and y0 are indexes indicating the position of the upper left pixel of the prediction block in the picture, and ref_idx_l0 [x0] [y0] and mvd_l0 [x0] [y0] [j] are respectively (x0 , y0) is the reference index of L0 of the prediction block and the differential motion vector, and ref_idx_l1 [x0] [y0] and mvd_l1 [x0] [y0] [j] are respectively (x0, y0) in the picture It is the reference index of L1 of the prediction block located, and a difference motion vector. Further, j represents a differential motion vector component, j represents 0 as an x component, and j represents 1 as a y component. Next, syntax elements mvp_idx_l0 [x0] [y0] and mvp_idx_l1 [x0] [y0] of an index of a predicted motion vector list that is a list of predicted motion vector candidates to be referred to are set. Here, x0 and y0 are indices indicating the position of the upper left pixel of the prediction block in the picture, and mvp_idx_l0 [x0] [y0] and mvp_idx_l1 [x0] [y0] are in (x0, y0) in the picture It is the prediction motion vector index of L0 and L1 of the prediction block located.

  The motion vector prediction method according to the embodiment is implemented in the difference motion vector calculation unit 103 of the video encoding device in FIG. 1 and the motion vector calculation unit 204 of the video decoding device in FIG.

  A motion vector prediction method according to an embodiment will be described with reference to the drawings. The motion vector prediction method is performed in any of the encoding and decoding processes for each prediction block constituting the encoding block. When the prediction mode of the prediction block is inter prediction (MODE_INTER) and not the merge mode, in the case of encoding, the encoded motion vector used when calculating the differential motion vector to be encoded from the motion vector to be encoded is used. In the case of deriving the predicted motion vector, the decoding is performed when the predicted motion vector is derived using the decoded motion vector used when calculating the motion vector to be decoded.

(Prediction of motion vector in encoding)
The operation of the motion vector prediction method according to the embodiment in the video encoding device that encodes a video bitstream based on the above-described syntax will be described. The motion vector prediction method is a method in which motion compensated prediction is performed in units of slices, that is, when the slice type is a P slice (unidirectional prediction slice) or a B slice (bidirectional prediction slice). The prediction mode is an inter prediction (MODE_INTER) and is applied to a prediction block that encodes or decodes a differential motion vector that is not a merge mode.

  FIG. 12 is a diagram showing a detailed configuration of the differential motion vector calculation unit 103 of the moving picture encoding device of FIG. A portion surrounded by a thick frame line in FIG. 12 indicates the differential motion vector calculation unit 103.

  Further, the part surrounded by a thick dotted line inside thereof shows an operation unit of a motion vector prediction method described later, and is similarly installed in a video decoding device corresponding to the video encoding device of the embodiment, The same determination result that does not contradict between encoding and decoding can be obtained. Hereinafter, a motion vector prediction method in encoding will be described with reference to FIG.

  The difference motion vector calculation unit 103 includes a prediction motion vector candidate generation unit 121, a prediction motion vector candidate registration unit 122, a prediction motion vector redundant candidate deletion unit 123, a prediction motion vector candidate number limit unit 124, and a prediction motion vector candidate code amount calculation unit. 125, a predicted motion vector selection unit 126, and a motion vector subtraction unit 127.

  In the differential motion vector calculation processing in the differential motion vector calculation unit 103, the differential motion vector of the motion vector used in the inter prediction method selected in the encoding target block is calculated for each of L0 and L1. Specifically, when the prediction mode PredMode of the encoding target block is inter prediction (MODE_INTER) and the inter prediction mode of the encoding target block is L0 prediction (Pred_L0), the prediction motion vector list mvpListL0 of L0 is calculated and predicted. A motion vector mvpL0 is selected, and a differential motion vector mvdL0 of the L0 motion vector is calculated. When the inter prediction mode (Pred_L1) of the encoding target block is L1 prediction, the L1 predicted motion vector list mvpListL1 is calculated, the predicted motion vector mvpL1 is selected, and the differential motion vector mvdL1 of the L1 motion vector is calculated. When the inter prediction mode of the encoding target block is bi-prediction (Pred_BI), both L0 prediction and L1 prediction are performed, a prediction motion vector list mvpListL0 of L0 is calculated, a prediction motion vector mvpL0 of L0 is selected, and L0 A difference motion vector mvdL0 of the L1 motion vector mvLL, a prediction motion vector list mvpListL1 of the L1, a prediction motion vector mvpL1 of the L1, and a difference motion vector mvdL1 of the motion vector mvL1 of the L1. .

  Difference motion vector calculation processing is performed for each of L0 and L1, but both L0 and L1 are common processing. Therefore, in the following description, L0 and L1 are represented as a common LX. In the process of calculating the differential motion vector of L0, X is 0, and in the process of calculating the differential motion vector of L1, X is 1. Further, when the information of the other list is referred to during the process of calculating the LX differential motion vector, the other list is represented as LY.

  For each of L0 and L1, the motion vector predictor candidate generating unit 121 has a prediction block group adjacent to the left side (a prediction block group adjacent to the left side of the prediction block in the same picture as the prediction block to be encoded: A0 in FIG. 5). , A1), prediction block group adjacent to the upper side (prediction block group adjacent to the upper side of the prediction block in the same picture as the prediction block to be encoded: B0, B1, B2 in FIG. 5), prediction blocks at different times Three prediction blocks in a group (predicted block groups that have already been encoded and are located in the same position as the prediction block in a temporally different picture from the encoding target prediction block: T0 and T1 in FIG. 9) From the group, one motion vector mvLXA, mvLXB, mvLXCol is derived for each prediction block group. To the predicted motion vector candidates, and supplies the predicted motion vector candidate registration unit 122. Hereinafter, mvLXA and mvLXB are referred to as spatial prediction motion vectors, and mvLXCol is referred to as a temporal prediction motion vector. In calculating the motion vector predictor candidates, the reference index and POC of the prediction block to be encoded, the prediction mode of the encoded prediction block stored in the encoding information storage memory 114, and the reference index for each reference list The encoding information such as the POC of the reference picture and the motion vector is used.

  These predicted motion vector candidates mvLXA, mvLXB, and mvLXCol may be derived by scaling according to the relationship between the POC of the current picture and the POC of the reference picture.

  The motion vector predictor candidate generation unit 121 performs a later-described condition determination on the prediction blocks in each prediction block group in a predetermined order for each prediction block group, and the motion of the prediction block that first meets the conditions A vector is selected and set as the motion vector candidates mvLXA, mvLXB, and mvLXCol.

  When calculating a prediction motion vector from a prediction block group adjacent to the left side, predictions adjacent to the upper side in the order from the bottom to the top of the prediction block group adjacent to the left side (order of A0 to A0 and A1 in FIG. 5). When calculating a prediction motion vector from a block group, prediction is performed from prediction block groups at different times in the order from right to left of the prediction block group adjacent to the upper side (order B0 to B0, B1, and B2 in FIG. 5). When calculating the motion vector, condition determination to be described later is performed for each prediction block in the order of T0 to T0 and T1 in FIG. 9, and the motion vector of the prediction block that first matches the condition is selected. Predicted motion vector candidates are mvLXA, mvLXB, and mvLXCol, respectively.

  That is, in the left adjacent prediction block group, the lowest prediction block has the highest priority, and the priority is given from the bottom to the top. In the upper adjacent prediction block group, the rightmost prediction block Blocks have the highest priority and are prioritized from right to left. In the prediction block group at different times, the prediction block of T0 has the highest priority, and the priority is given in the order of T0 and T1.

  This is because in the case of selecting a plurality of motion vector predictor candidates, it is effective that each motion vector predictor candidate has a motion vector having a different value as much as possible, and that the range of motion vector predictor selection is widened on the encoding side. This is because it leads to the selection of a predictive motion vector.

(Explanation of loop for condition judgment of spatial prediction block)
For the adjacent prediction block group on the left, the respective condition determinations are applied in the priority order of the following condition determinations 1 and 2. On the other hand, the respective condition determinations are applied to the upper adjacent prediction block group in the following priority order of condition determinations 1, 2, 3, and 4.

<Condition judgment>
Condition determination 1: Prediction using the same reference index, that is, the same reference picture is performed in the adjacent prediction block with the same reference list LX as the motion vector of the LX that is the difference motion vector calculation target of the prediction block to be encoded / decoded. Yes.
Condition determination 2: Although the reference list LY is different from the LX motion vector of the differential motion vector calculation target of the prediction block to be encoded / decoded, prediction using the same reference picture is performed in the adjacent prediction block.
Condition determination 3: Prediction using a different reference picture is performed in the adjacent prediction block in the same reference list LX as the LX motion vector of the differential motion vector calculation target of the prediction block to be encoded / decoded.
Condition determination 4: Prediction using a different reference picture in a reference list LY different from a motion vector of an LX that is a difference motion vector calculation target of a prediction block to be encoded / decoded is performed in an adjacent prediction block.

  If any of these conditions is met, a motion vector that matches the condition is adopted as a prediction vector candidate for the prediction block, and the subsequent condition determination is not performed. Note that, when the condition of condition determination 1 or condition determination 2 is met, the motion vector of the corresponding adjacent prediction block is the same as the reference picture of the encoding target prediction block, so that it is used as a prediction motion vector candidate as it is. If the condition of decision 3 or condition decision 4 is met, the motion vector of the corresponding adjacent prediction block is different from the reference picture of the prediction block to be encoded, so that the motion vector is scaled based on the distance between the reference images and predicted Candidate motion vectors.

  In the above four condition determinations, the prediction vector is determined by scanning the prediction block in the prediction block group in the following order.

<Prediction vector scan procedure>
Priority is given to prediction motion vector determination that does not require scaling operations using the same reference picture, and two condition determinations are performed for each prediction block among the four condition determinations. If the condition is not satisfied, the condition determination for the next prediction block is performed. Move on. First, the condition determination of condition determination 1 and condition determination 2 is performed on each prediction block in the adjacent prediction block group, and then the condition determination of condition determination 3 and condition determination 4 is performed on each prediction block in the adjacent prediction block group. Against.

Specifically, the condition determination is performed in the following priority order. (However, N is A or B)
1. Condition determination 1 of the prediction block N0 (same reference list LX, same reference picture)
2. Prediction block N0 condition determination 2 (different reference list LY, same reference picture)
3. Condition determination 1 of the prediction block N1 (same reference list LX, same reference picture)
4). Condition determination 2 of the prediction block N1 (different reference list LY, same reference picture)
5. Prediction block N2 condition determination 1 (same reference list LX, same reference picture)
6). Condition determination 2 of the prediction block N2 (different reference list LY, same reference picture)
7). Condition determination 3 of the prediction block N0 (same reference list LX, different reference pictures)
8). Prediction block N0 condition decision 4 (different reference list LY, different reference picture)
9. Condition determination 3 of the prediction block N1 (same reference list LX, different reference pictures)
10. Prediction block N1 condition determination 4 (different reference list LY, different reference picture)
11. Condition determination 3 of the prediction block N2 (same reference list LX, different reference pictures)
12 Prediction block N2 condition determination 4 (different reference list LY, different reference picture)

  By scanning as described above, a prediction motion vector with high prediction accuracy that does not require a scaling operation using the same reference picture is easily selected, so that the coding amount of the difference motion vector is reduced and the encoding efficiency is improved. is there.

Here, in the adjacent prediction block group on the left side, since the group is configured by two prediction blocks, 5.6. And 11.12. This procedure is not performed.
In the adjacent prediction block group on the left side, only the steps 1 to 4 of the scan are performed, and the steps 7 to 10 are not performed (condition determinations 3 and 4 are not performed). That is, in the adjacent prediction block group on the left, scanning under conditions that require a scaling operation is omitted.

  Thereby, the scaling operation that requires division of the left adjacent prediction block group can be omitted, and the scaling operation can be limited only to the prediction vector candidates of the upper adjacent prediction block group. The circuit scale can be reduced by having only one scaling operation circuit only for the upper adjacent prediction block group. Also, compared with the case where the scaling operation is adaptively limited to one time for the motion vector predictor from either the prediction block group adjacent to the left side or the prediction block group adjacent to the upper side, the scaling circuit is There is an effect that it can be fixedly arranged only on the prediction block group side adjacent to.

  The flow of the above process will be described in detail later using the flowcharts of FIGS.

  The calculation of the prediction vector candidate mvLXCol from the prediction block groups at different times will be described later in detail using the flowcharts of FIGS.

  Returning to the description of FIG. Next, the predicted motion vector candidate registration unit 122 stores the calculated predicted motion vector candidates mvLXA, mvLXB, and mvLXCol in the predicted motion vector list mvpListLX.

  Next, the predicted motion vector redundancy candidate deletion unit 123 compares the motion vector values of the spatial motion vector predictor candidates stored in the LX motion vector predictor list mvpListLX, and uses the same motion from the motion vector predictor candidates. Predicted motion vectors by determining those with vector values, deleting the first one of the predicted motion vector candidates determined to have the same motion vector value, and deleting the others from the predicted motion vector list mvpListLX And the motion vector predictor list mvpListLX is updated. Redundant candidate deletion is not performed between the spatial prediction motion vector candidate and the temporal prediction motion vector candidate. The reason why redundant candidate deletion of temporal prediction motion vector candidates is not performed will be described later.

  In the motion vector predictor candidate number limiting unit 124, the number of motion vector predictor candidates registered in the motion vector list mvpListLX of LX is counted, and the motion vector candidate number numMVPCandLX of LX is set to the counted candidate value. Is done.

  Further, the motion vector predictor candidate number limiting unit 124 fixes the motion vector vector candidate number numMVPCandLX registered in the motion vector list mvpListLX of LX to the specified final candidate number finalNumMVPCand.

  In the present embodiment, the final candidate number finalNumMVPCand is defined as 2. The reason for fixing the number of prediction vector candidates is that when the number of prediction motion vector candidates registered in the prediction motion vector list fluctuates according to the construction state of the prediction motion vector list, the decoding side constructs the prediction motion vector list. Otherwise, entropy decoding of the motion vector predictor index cannot be performed, so that a dependency relationship occurs between the construction of the motion vector list and the entropy decoding, which may cause a waiting time for the entropy decoding. . Furthermore, if entropy decoding depends on the construction state of the motion vector predictor list including the motion vector candidate mvLXCol derived from the prediction block of the picture at different time, when an error occurs when decoding the encoded bit string of another picture The encoded bit string of the current picture is also affected by the error, and there is a problem that entropy decoding cannot be continued normally. If the final candidate number finalNumMVPCand is defined as a fixed number, the prediction motion vector index can be entropy-decoded independently of the construction of the prediction motion vector list, and an error occurs when decoding the encoded bit string of another picture. However, it is possible to continue the entropy decoding of the coded bit string of the current picture without being affected by this.

  When the number of predicted motion vector candidates numMVPCandLX is smaller than the specified final candidate number finalNumMVPCand, the predicted motion vector candidate number limit unit 124 (0 until the predicted motion vector candidate number numMVPCandLX becomes the same value as the final candidate number finalNumMVPCand. , 0) is added to the motion vector list mvpListLX to limit the number of motion vector predictor candidates to a specified value. In this case, a motion vector having a value of (0, 0) may be added in duplicate, but on the decoding side, how the predicted motion vector index falls within the range from 0 to (specified number of candidates −1). Even if a large value is taken, the predicted motion vector can be determined. On the other hand, when the number of predicted motion vector candidates for LX numMVPCandLX is larger than the final number of final candidates defined, finalNumMVPCand, all elements registered at an index larger than finalNumMVPCand-1 are deleted from the predicted motion vector list mvpListLX, and the predicted motion vector for LX By setting the number of candidates numMVPCandLX to the same value as the final candidate number finalNumMVPCand, the number of motion vector predictor candidates is limited to a specified value. The updated prediction motion vector list mvpListLX is supplied to the prediction motion vector candidate code amount calculation unit 125 and the prediction motion vector selection unit 126.

  On the other hand, a motion vector mvLX of LX (X = 0 or 1) is detected for each prediction block by the motion vector detection unit 102 in FIG. These motion vectors mvLX are supplied to the prediction motion vector candidate code amount calculation unit 125 together with the prediction motion vector candidates of the updated prediction motion vector list mvpListLX.

  The motion vector predictor candidate code amount calculation unit 125 calculates the difference from the motion vector mvLX for each motion vector candidate mvpListLX [i] stored in the motion vector list mvpListLX of LX (X = 0 or 1). The motion vectors are calculated, the code amounts when the differential motion vectors are encoded are calculated, and supplied to the predicted motion vector selection unit 126.

  The motion vector predictor selection unit 126 selects a motion vector candidate mvpListLX [i] that has the smallest code amount for each motion vector predictor candidate from among the elements registered in the motion vector list mvpListLX of LX. Select as the predicted motion vector mvpLX. In the case where there are a plurality of motion vector predictor candidates having the smallest generated code amount in the motion vector predictor list mvpListLX, the motion vector predictor candidates represented by the index i in the motion vector list mvpListLX with a small number mvpListLX [i] is selected as the LX optimum predicted motion vector mvpLX. The selected predicted motion vector mvpLX is supplied to the motion vector subtraction unit 127. Further, the index i in the predicted motion vector list corresponding to the selected predicted motion vector mvpLX is output as a predicted motion vector index mvpIdxLX of LX.

Finally, the motion vector subtraction unit 127 calculates the LX differential motion vector mvdLX by subtracting the selected LX predicted motion vector mvpLX from the LX motion vector mvLX, and outputs the differential motion vector mvdLX.
mvdLX = mvLX-mvpLX

  Returning to FIG. 1, the motion compensation prediction unit 105 refers to the image signal of the decoded picture stored in the decoded image memory 115, and the motion vector of LX (X = 0 or 1) supplied from the motion vector detection unit 102. Motion compensation is performed according to mvLX to obtain a predicted image signal, which is supplied to the prediction method determination unit 106.

  The prediction method determination unit 106 determines a prediction method. The code amount and the coding distortion are calculated for each prediction mode, and the prediction block size and the prediction mode for the smallest generated code amount and coding distortion are determined. The LX differential motion vector mvdLX supplied from the motion vector subtraction unit 127 of the differential motion vector calculation unit 103 and the LX prediction motion vector index mvpIdxLX representing the prediction motion vector supplied from the prediction motion vector selection unit 126 are encoded. The code amount of motion information is calculated. Further, the code amount of the prediction residual signal obtained by encoding the prediction residual signal between the prediction image signal supplied from the motion compensation prediction unit 105 and the image signal to be encoded supplied from the image memory 101 is calculated. The total generated code amount obtained by adding the code amount of the motion information and the code amount of the prediction residual signal is calculated and used as the first evaluation value.

  Further, after encoding such a prediction residual signal, it is decoded for distortion amount evaluation, and the encoding distortion is calculated as a ratio representing an error from the original image signal caused by the encoding. By comparing the total generated code amount and the coding distortion for each motion compensation, the prediction block size and the prediction mode that generate the least generated code amount and the coding distortion are determined. The motion vector prediction method described above is performed on the motion vector mvLX corresponding to the prediction mode of the determined prediction block size, and the index mvpIdxLX representing the prediction motion vector is a second syntax pattern in units of prediction blocks. It is encoded as the represented syntax element mvp_idx_lX [i]. The generated code amount calculated here is preferably a simulation of the encoding process, but can be approximated or approximated easily.

(Prediction of motion vector in decoding)
The operation of the motion vector prediction method according to the present invention in the video decoding device that decodes the encoded video bitstream based on the above-described syntax will be described.

  When the motion vector prediction method according to the embodiment is performed, processing is performed by the motion vector calculation unit 204 of the video decoding device in FIG. FIG. 13 is a diagram illustrating a detailed configuration of the motion vector calculation unit 204 of the video decoding device of FIG. 2 corresponding to the video encoding device of the embodiment. A portion surrounded by a thick frame line in FIG. 13 indicates the motion vector calculation unit 204. Further, the part surrounded by the thick dotted line inside shows the operation part of the motion vector prediction method described later, which is also installed in the corresponding moving picture coding apparatus in the same way and has the same consistency between coding and decoding. Judgment results are obtained. Hereinafter, a motion vector prediction method in decoding will be described with reference to FIG.

  The motion vector calculation unit 204 includes a predicted motion vector candidate generation unit 221, a predicted motion vector candidate registration unit 222, a predicted motion vector redundancy candidate deletion unit 223, a predicted motion vector candidate number limit unit 224, a predicted motion vector selection unit 225, and a motion vector. An adder 226 is included.

  In the motion vector calculation process in the motion vector calculation unit 204, a motion vector used in inter prediction is calculated for each of L0 and L1. Specifically, when the prediction mode PredMode of the encoding target block is inter prediction (MODE_INTER) and the inter prediction mode of the encoding target block is L0 prediction (Pred_L0), the prediction motion vector list mvpListL0 of L0 is calculated and predicted. A motion vector mvpL0 is selected, and a motion vector mvL0 of L0 is calculated. When the inter prediction mode of the encoding target block is L1 prediction (Pred_L1), the L1 predicted motion vector list mvpListL1 is calculated, the predicted motion vector mvpL1 is selected, and the L1 motion vector mvL1 is calculated. When the inter prediction mode of the encoding target block is bi-prediction (Pred_BI), both L0 prediction and L1 prediction are performed, a prediction motion vector list mvpListL0 of L0 is calculated, a prediction motion vector mvpL0 of L0 is selected, and L0 Motion vector mvL0, L1 predicted motion vector list mvpListL1 is calculated, L1 predicted motion vector mvpL1 is calculated, and L1 motion vector mvL1 is calculated.

  Similar to the encoding side, the motion vector calculation processing is performed for each of L0 and L1 on the decoding side, but both L0 and L1 are common processing. Therefore, in the following description, L0 and L1 are represented as a common LX. X is 0 in the process of calculating the L0 motion vector, and X is 1 in the process of calculating the L1 motion vector. In addition, when the information of the other list is referred to instead of LX during the process of calculating the LX motion vector, the other list is represented as LY.

  The motion vector calculation unit 204 includes a motion vector predictor candidate generation unit 221, a motion vector predictor candidate registration unit 222, a motion vector predictor redundant candidate deletion unit 223, and a motion vector predictor candidate number restriction unit 224, which are differential motion vectors on the encoding side. It is specified that the motion vector predictor candidate generating unit 121, the motion vector predictor candidate registering unit 122, the motion vector predictor redundant candidate deleting unit 123, and the motion vector predictor candidate number limiting unit 124 in the calculation unit 103 perform the same operation. Thus, the same motion vector predictor candidates that are consistent between encoding and decoding can be obtained on the encoding side and the decoding side.

  The motion vector predictor candidate generation unit 221 performs the same processing as that of the motion vector predictor candidate generation unit 121 on the encoding side in FIG. The same position as or near the decoded prediction block adjacent to the decoding target block in the same picture as the decoding target block and the decoding target block in a different picture recorded in the encoded information storage memory 209 after decoding. Are read out from the encoded information storage memory 209. Predicted motion vector candidates mvLXA, mvLXB, and mvLXCol are generated for each predicted block group from the motion vectors of other decoded blocks read from the encoded information storage memory 209 and supplied to the predicted motion vector candidate registration unit 222. .

  These predicted motion vector candidates mvLXA, mvLXB, and mvLXCol may be calculated by scaling according to the reference index. Note that the motion vector predictor candidate generation unit 221 performs the same processing as the motion vector predictor candidate generation unit 121 on the encoding side in FIG. 12, and thus detailed description thereof is omitted here.

  Next, the motion vector predictor candidate registration unit 222 performs the same processing as the motion vector predictor candidate registration unit 122 on the encoding side in FIG. The calculated prediction motion vector candidates mvLXA, mvLXB, and mvLXCol are stored in the LX prediction motion vector list mvpListLX.

  Next, the motion vector redundancy candidate deletion unit 223 performs the same processing as the motion vector redundancy candidate deletion unit 123 on the encoding side in FIG. Predictive motion vector candidates determined to have the same motion vector value from among the spatial prediction motion vector candidates stored in the LX predicted motion vector list mvpListLX. The remaining one is deleted from the predicted motion vector list mvpListLX, the predicted motion vector candidates are not duplicated, and the predicted motion vector list mvpListLX is updated. Redundant candidate deletion is not performed between the spatial prediction motion vector candidate and the temporal prediction motion vector candidate.

  Next, the motion vector predictor candidate number limiting unit 224 performs the same processing as the motion vector predictor candidate number limiting unit 124 on the encoding side in FIG. In the motion vector predictor candidate number limiting unit 224, the number of elements registered in the motion vector predictor list mvpListLX is counted, and the motion vector candidate number numMVPCandLX of LX is set to the value of the counted candidate number.

  Further, the motion vector predictor candidate number restriction unit 224 fixes the motion vector vector candidate number numMVPCandLX of LX registered in the motion vector predictor list mvpListLX to the final candidate number finalNumMVPCand for the reason explained on the encoding side.

  In the present embodiment, the final candidate number finalNumMVPCand is defined as 2. When the number of predicted motion vector candidates for LX numMVPCandLX is smaller than the specified final candidate number finalNumMVPCand, a motion vector having a value of (0, 0) is obtained until the predicted motion vector candidate number numMVPCandLX becomes the same value as the final candidate number finalNumMVPCand. By adding to the motion vector predictor list mvpListLX, the number of motion vector predictor candidates is limited to a specified value. In this case, a motion vector having a value of (0, 0) may be added in duplicate, but on the decoding side, any value within the range of the predicted motion vector index from 0 to the specified value −1 is taken. The predicted motion vector can be determined. On the other hand, when the number of predicted motion vector candidates for LX numMVPCandLX is larger than the final number of final candidates defined, finalNumMVPCand, all elements registered at an index larger than finalNumMVPCand-1 are deleted from the predicted motion vector list mvpListLX, and the predicted motion vector for LX By setting the number of candidates numMVPCandLX to the same value as the final candidate number finalNumMVPCand, the number of motion vector predictor candidates is limited to a specified value. The updated motion vector predictor list mvpListLX is supplied to the motion vector predictor selection unit 225.

  On the other hand, the LX (X = 0 or 1) differential motion vector mvdLX for each prediction block decoded by the first encoded bit string decoding unit 202 is supplied to the motion vector addition unit 226. The predicted motion vector index mvpIdxLX of the LX predicted motion vector decoded by the first encoded bit string decoding unit 202 is supplied to the predicted motion vector selection unit 225.

  The prediction motion vector selection unit 225 is supplied with the prediction motion vector index mvpIdxLX of the LX decoded by the first encoded bit string decoding unit 202, and selects a prediction motion vector candidate mvpListLX [mvpIdxLX] corresponding to the supplied index mvpIdxLX. It extracts from the motion vector list mvpListLX as the motion vector motion vector mvpLX of LX. The supplied motion vector predictor candidate is supplied to the motion vector adding unit 226 as a motion vector predictor mvpLX.

Finally, the motion vector addition unit 226 calculates the LX motion vector mv by adding the LX differential motion vector mvdLX and the LX prediction motion vector mvpLX that are supplied after being decoded by the first encoded bit string decoding unit 202. And outputs a motion vector mvLX.
mvLX = mvpLX + mvdLX

  As described above, the LX motion vector mvLX is calculated for each prediction block. A predicted image signal is generated by motion compensation using this motion vector, and is added to the decoded prediction residual signal to generate a decoded image signal.

  Processing procedures of the difference motion vector calculation unit 103 of the moving image encoding device and the motion vector calculation unit 204 of the moving image decoding device will be described with reference to the flowcharts of FIGS. 14 and 15, respectively. FIG. 14 is a flowchart showing a difference motion vector calculation processing procedure by the moving image encoding device, and FIG. 15 is a flowchart showing a motion vector calculation processing procedure by the moving image decoding device.

  With reference to FIG. 14, the differential motion vector calculation processing procedure on the encoding side will be described. On the encoding side, first, the differential motion vector calculation unit 103 sets the final candidate number finalNumMVCand of motion vector predictor candidates (S100). In the present embodiment, the final candidate number finalNumMVPCand is defined as 2.

  Subsequently, the differential motion vector calculation unit 103 calculates the differential motion vector of the motion vector used in the inter prediction selected in the encoding target block for each of L0 and L1 (S101 to S106).

  When the LX differential motion vector mvdLX is calculated (YES in S102), the LX predicted motion vector candidates are calculated to construct the LX predicted motion vector list mvpListLX (S103). The motion vector predictor candidate generation unit 121 in the motion vector difference calculation unit 103 calculates a plurality of motion vector predictor candidates, and the motion vector predictor candidate registration unit 122 calculates the motion vector predictor candidates calculated in the motion vector predictor list mvpListLX. In addition, the motion vector redundancy candidate deletion unit 123 deletes unnecessary motion vector predictor candidates, and the motion vector predictor candidate number restriction unit 124 predicts the number of motion vector candidates in the motion vector list mvpListLX. The prediction motion vector list mvpListLX is constructed by fixing the number of motion vector candidates to numMVPCandLX. The detailed processing procedure of step S103 will be described later with reference to the flowchart of FIG.

  Subsequently, the predicted motion vector candidate code amount calculation unit 125 and the predicted motion vector selection unit 126 select the LX predicted motion vector mvpLX from the LX predicted motion vector list mvpListLX (S104). First, the motion vector predictor candidate code amount calculation unit 125 calculates each motion vector difference which is a difference between the motion vector mvLX and each motion vector candidate mvpListLX [i] stored in the motion vector predictor list mvpListLX. Then, the coding amount when the difference motion vector is encoded is calculated for each element of the prediction motion vector list mvpListLX, and the prediction motion vector selection unit 126 among the elements registered in the prediction motion vector list mvpListLX, A predicted motion vector candidate mvpListLX [i] that minimizes the code amount for each predicted motion vector candidate is selected as a predicted motion vector mvpLX. In the case where there are a plurality of motion vector predictor candidates having the smallest generated code amount in the motion vector predictor list mvpListLX, the motion vector predictor candidates represented by the index i in the motion vector list mvpListLX with a small number Select mvpListLX [i] as the optimal prediction motion vector mvpLX.

Subsequently, the motion vector subtraction unit 127 calculates an LX differential motion vector mvdLX by subtracting the selected LX motion vector mvpLX from the LX motion vector mvLX (S105).
mvdLX = mvLX-mvpLX

  Next, the decoding-side motion vector calculation processing procedure will be described with reference to FIG. Also on the decoding side, the motion vector calculation unit 204 calculates motion vectors used for inter prediction for each of L0 and L1 (S201 to S206).

  When calculating the LX motion vector mvdLX (YES in S202), the LX predicted motion vector candidates are calculated to construct the LX predicted motion vector list mvpListLX (S203). The motion vector calculation unit 204 includes a motion vector predictor candidate generation unit 221 that calculates a plurality of motion vector predictor candidates, and the motion vector predictor candidate registration unit 222 adds the motion vector predictor candidates calculated to the motion vector predictor list mvpListLX. Then, the motion vector redundancy candidate deletion unit 223 deletes unnecessary motion vector predictor candidates, and the motion vector predictor candidate number limiting unit 224 stores the motion vector vector candidate number in the motion vector list mvpListLX. The prediction motion vector list mvpListLX is constructed by fixing the number of vector candidates to numMVPCandLX. The detailed processing procedure of step S203 will be described later with reference to the flowchart of FIG.

  Subsequently, a predicted motion vector candidate mvpListLX [mvpIdxLX] corresponding to the predicted motion vector index mvpIdxLX decoded by the first encoded bit string decoding unit 202 is selected from the predicted motion vector list mvpListLX by the predicted motion vector selection unit 225. The predicted motion vector mvpLX is extracted (S204).

Subsequently, the motion vector addition unit 226 calculates the LX motion vector mvLX by adding the LX differential motion vector mvdLX decoded and supplied by the first encoded bit string decoding unit 202 and the LX predicted motion vector mvpLX. (S205).
mvLX = mvpLX + mvdLX

  Next, a procedure for derivation of a prediction motion vector candidate common to S103 of FIG. 14 and S203 of FIG.

(Motion vector prediction method)
FIG. 16 shows prediction motion vector candidate generation units 121 and 221 having a function common to the difference motion vector calculation unit 103 of the video encoding device and the motion vector calculation unit 204 of the video decoding device, and a prediction motion vector candidate registration unit 122. And 222, and the motion vector redundancy candidate deletion units 123 and 223 and the motion vector predictor candidate number limiting units 124 and 224 are flowcharts.

  The motion vector predictor candidate generation units 121 and 221 derive motion vector predictor candidates from the prediction block adjacent on the left side, a flag availableFlagLXA indicating whether the motion vector candidate of the prediction block adjacent on the left side can be used, and motion A vector mvLXA, a reference index refIdxA, and a list ListA are derived (S301 in FIG. 16). Note that X is 0 when L0 and X is 1 when L1 (the same applies hereinafter). Subsequently, the motion vector predictor candidate generation units 121 and 221 derive a motion vector predictor candidate from the upper adjacent prediction block, and a flag availableFlagLXB indicating whether the motion vector predictor candidate of the upper adjacent prediction block can be used. , And a motion vector mvLXB, a reference index refIdxB, and a list ListB are derived (S302 in FIG. 16). The processes in steps S301 and S302 in FIG. 16 are the same except that the position and number of adjacent blocks to be referenced are different and the scanning procedure of the adjacent blocks is different, and whether or not the motion vector predictor candidate of the prediction block can be used. A flag availableFlagLXN, a motion vector mvLXN, a reference index refIdxN, and ListN (N is A or B, and so on) are derived. The processing procedure of S301 and S302 will be described in detail later using the flowcharts of FIGS.

  Subsequently, the motion vector predictor candidate generation units 121 and 221 derive motion vector predictor candidates from the prediction blocks of the pictures at different times, and indicate whether the motion vector predictor candidates of the motion pictures at different times can be used. A flag availableFlagLXCol, a motion vector mvLXCol, a reference index refIdxCol, and a list ListCol are derived (S303 in FIG. 16). The derivation processing procedure of step S303 will be described in detail later with reference to the flowcharts of FIGS.

  Subsequently, the motion vector predictor candidate registration units 122 and 222 create a motion vector predictor list mvpListLX, and add the respective motion vector candidates mvLXA, mvLXB, and mvLXCol of LX (S304 in FIG. 16). The registration processing procedure in step S304 will be described in detail later using the flowchart in FIG.

  Next, the motion vector redundancy candidate deletion units 123 and 223 are derived from the spatial motion vector candidate mvLXA derived from the prediction block group adjacent on the left side in the motion vector predictor list mvpListLX and the prediction block group adjacent to the upper side. When the spatial prediction motion vector candidate mvLXB has the same value, one of them is determined as a redundant motion vector candidate, and the redundant motion with the smallest order i.e., the motion vector candidate mvLXA with the smallest index i is left. The vector candidate mvLXB is removed (S305 in FIG. 16). The redundancy candidate deletion comparison unit does not perform motion vector redundancy comparison between the temporal prediction motion vector candidate mvLXCol and the spatial prediction motion vector candidate mvLXA or mvLXB. This is because temporal prediction motion vector candidates and spatial prediction motion vector candidates are generated in parallel because there are different motion information memories to be accessed and different processing amounts. If motion vector comparison between vector candidates is performed, it is necessary to speed up synchronization when performing parallel processing, and waiting time occurs for either spatial prediction candidate generation processing or temporal candidate generation processing. It is. Therefore, since it is desirable to separate the temporal prediction motion vector candidate and the spatial prediction motion vector candidate derivation process as much as possible, in the present invention, the motion vector redundancy comparison between the temporal prediction motion vector candidate and the spatial prediction motion vector candidate is performed. Absent. A temporal motion vector predictor and a spatial motion vector predictor are different in the motion vector derivation process, so that motion vector predictors having different properties are derived. Therefore, the probability that the temporal prediction motion vector and the spatial prediction motion vector are the same is very low, and there is no reduction in coding efficiency without performing a motion vector comparison between the temporal prediction motion vector candidate and the spatial prediction motion vector candidate. Also, even if the temporal prediction candidate is deleted when the temporal prediction candidate is the same motion vector as the spatial prediction candidate, the newly added prediction motion vector candidate is only the (0, 0) motion vector. Efficiency improvement is not expected. The deletion processing procedure in step S305 will be described in detail later using the flowchart of FIG.

  Subsequently, the motion vector predictor candidate number limiting units 124 and 224 count the number of elements added in the motion vector predictor list mvpListLX, and the number is set as the motion vector vector candidate number numMVPCandLX of the LX. The number of predicted motion vector candidates for LX numMVPCandLX registered in mvpListLX is limited to the specified final candidate number finalNumMVPCand. (S306 in FIG. 16). The deletion processing procedure in step S306 will be described in detail later using the flowchart of FIG.

  A procedure for deriving motion vector predictor candidates from prediction blocks adjacent to the left or upper side of S301 and S302 in FIG. 16 will be described in detail.

  As shown in FIGS. 5, 6, 7, and 8, a prediction block (in FIG. 5, FIG. 6, FIG. 7, and FIG. 8) that is defined for motion compensation within a coding block in the same picture. A motion vector predictor candidate is derived from surrounding prediction blocks adjacent to the processing target prediction block.

  The motion vector predictor candidates are the prediction block Ak (k = 0, 1) adjacent to the left side of the prediction block to be processed, that is, the prediction block group A composed of A0 and A1, and the prediction block Bk (k adjacent to the upper side). = 0, 1, 2), that is, predictive motion vector candidates are selected from the prediction block group B composed of B0, B1, and B2.

  Next, a method for deriving prediction motion vector candidates from the prediction block group N adjacent on the left and upper sides, which is the processing procedure of S301 and S302 in FIG. 16, will be described. FIG. 17 is a flowchart showing the prediction motion vector candidate derivation processing procedure of S301 and S302 of FIG. 16 by the above-described scanning method. Subscript X contains 0 or 1 representing the reference list, and N represents A (left side) or B (upper side) representing the area of the adjacent prediction block group.

  In order to derive a motion vector predictor candidate from the left prediction block (S301 in FIG. 16), N is A in FIG. 17, and prediction is performed from prediction blocks A0 and A1 adjacent to the left side of the prediction block to be encoded / decoded. In order to derive a motion vector candidate from the upper prediction block (S302 in FIG. 16), a motion vector predictor is predicted from the B0, B1, and B2 prediction blocks adjacent to the upper side with N as B in FIG. The candidates are calculated according to the following procedure.

  First, a prediction block adjacent to a prediction block to be encoded / decoded is specified, and encoding is performed when each prediction block Nk (k = 0, 1, 2, where 2 is only the upper prediction block group) can be used. The encoded information stored in the information storage memory 114 or the encoded information storage memory 209 is acquired. The encoding information of the adjacent prediction block Nk acquired here includes the flags predFlagLX [xNk] [yNk] indicating whether to use the prediction modes PredMode and LX, the LX reference indexes refIdxLX [xNk] [yNk], and LX. The motion vector mvLX [xNk] [yNk]. In the case of a prediction block group (N = A) adjacent to the left side of the prediction block to be encoded / decoded (YES in S1101), the prediction block A0 adjacent to the lower left and the prediction block A1 adjacent to the left are specified and encoded. Information is acquired (S1102, S1103), and in the case of a prediction block group (N = B) adjacent to the upper side of the prediction block to be encoded / decoded (NO in S1101), adjacent to the prediction block B0 adjacent to the upper right Coding information is acquired by specifying the prediction block B1 to be predicted and the prediction block B2 adjacent to the upper left (S1104, S1105, S1106). It can be used when the adjacent prediction block Nk is located inside the slice including the prediction block to be encoded / decoded, and cannot be used when located outside.

  Next, a flag availableFlagLXN indicating whether or not a prediction motion vector is selected from the prediction block group N is set to 0, and a motion vector mvLXN representing the prediction block group N is set to (0, 0) (S1107).

  Subsequently, a motion vector predictor candidate that matches Condition Determination 1 or Condition Determination 2 described above is derived (S1108). Reference list that is currently targeted by the prediction block to be encoded / decoded in the adjacent prediction blocks N0, N1, and N2 (N2 is only the upper adjacent prediction block group B) of the prediction block group N (N is A or B) The same reference list LX as LX, or the reference list LY opposite to the reference list LX that is the current target in the encoding / decoding target prediction block (Y =! X: the opposite reference list when the current target reference list is L0) Is L1, and when the current reference list is L1, the opposite reference list is L0), and the motion vector of the same reference picture is searched for and the motion vector is set as a motion vector predictor candidate.

  FIG. 18 is a flowchart showing the derivation process procedure of step S1108 of FIG. For the adjacent prediction block Nk (k = 0, 1, 2, where 2 is only the upper adjacent prediction block group), the following processing is performed in the order of k, 0, 1, 2 (S1201 to S1207). ). When N is A, the following processing is performed in the order of A0 and A1 from bottom to top, and when N is B, the following processing is performed in the order of B0, B1, and B2 from right to left.

  When the adjacent prediction block Nk can be used and the prediction mode PredMode of the prediction block Nk is not intra (MODE_INTRA) (YES in S1202), the condition determination of the above-described condition determination 1 is performed (S1203). The flag predFlagLX [xNk] [yNk] indicating whether or not to use the LX of the adjacent prediction block Nk is 1, that is, the adjacent prediction block Nk is inter-predicted using the same LX motion vector as the calculation target and adjacent When the LX reference index refIdxLX [xNk] [yNk] of the prediction block Nk and the index refIdxLX of the prediction block to be processed are the same, that is, when adjacent prediction blocks Nk are inter-predicted using the same reference picture in LX prediction ( If YES in step S1203, the process advances to step S1204. If not (NO in step S1203), the condition determination in step S1205 is performed.

  When step S1203 is YES, the flag availableFlagLXN is set to 1, the prediction motion vector mvLXN of the prediction block group N is set to the same value as the LX motion vector mvLX [xNk] [yNk] of the adjacent prediction block Nk, and prediction is performed. The reference index refIdxN of the block group N is set to the same value as the LX reference index refIdxLX [xNk] [yNk] of the adjacent prediction block Nk, and the reference list ListN of the prediction block group N is set to LX (S1204). The predicted motion vector candidate calculation process is terminated.

  On the other hand, when step S1203 is NO, the condition determination of the above-described condition determination 2 is performed (S1205). The flag predFlagLY indicating whether to use the LY of the adjacent prediction block Nk is 1, that is, the adjacent prediction block Nk is inter-predicted using a motion vector of LY different from the calculation target, and the current target of the adjacent prediction block Nk Inter prediction using the same reference picture in the LY prediction where the POC of the reference picture in the reference list LY opposite to the reference list LX is the same as the POC of the reference picture of the LX of the prediction block to be processed If YES in step S1205, the process advances to step S1206, the flag availableFlagLXN is set to 1, and the prediction motion vector mvLXN of the prediction block group N is the LY motion vector mvLY [xNk] [yNk] of the adjacent prediction block Nk. And the reference index refIdxN of the prediction block group N is It is set to the same value as the LY reference index refIdxLY [xNk] [yNk] of the adjacent prediction block Nk, the reference list ListN of the prediction block group N is set to LY (S1206), and this motion vector predictor candidate calculation process ends. To do.

  If these conditions are not met, that is, if NO in S1203 and NO in S1205, k is incremented by 1 and the next adjacent prediction block is processed (S1201 to S1207), and availableFlagLXN becomes 1 or adjacent block Repeat until A1 or B2 is finished.

  Subsequently, returning to the flowchart of FIG. 17, when the prediction block group is adjacent on the upper side (N == A) (YES in S1109), a motion vector predictor candidate that matches the above-described condition determination 3 or condition determination 4 is calculated. Therefore, the process proceeds to the condition determination in S1110, and when the prediction block group is not adjacent on the upper side (N ≠ A) (NO in S1109), the prediction motion vector candidate calculation process is terminated.

  When availableFlagLXB is 0 (YES in S1110), that is, when a motion vector predictor candidate cannot be calculated in step S1108 in the prediction block group adjacent on the upper side, a motion vector predictor that matches the above-described condition determination 3 or condition determination 4 Candidates are calculated (S1113).

  When availableFlagLXB is not 0 (NO in S1110), that is, when a motion vector predictor candidate can be calculated in step S1108 in the upper adjacent prediction block group, a motion vector predictor candidate is further detected in the prediction block group adjacent on the left side. It is determined whether or not the calculation was successful (S1111).

  When the prediction motion vector of the prediction block group adjacent on the left side cannot be calculated (availableFlagLXA == 0) (YES in S1111), the prediction motion vector detected in condition determination 1 or condition determination 2 of the prediction block group adjacent on the upper side And the reference index is set to availableFlagLA = 1, mvLXA = mvLXB, refIdxA = refIdxB, and the motion vector of the adjacent prediction block group and availableFlagLB = 0, mvLXB = (0,0) and refIdxB = -1 are set (S1112).

  As a result, when a motion vector predictor candidate from the prediction block group adjacent on the left side cannot be calculated, a plurality of motion vector predictor candidates can be calculated from the prediction block group adjacent on the upper side, so the number of motion vector motion candidate candidates decreases. However, it is possible to prevent the problem that the prediction efficiency of the motion vector is lowered.

  In addition, the prediction motion vector candidates from the prediction block group adjacent to the left side having a high registration priority in the prediction list of motion vector predictors are only candidates that do not require scaling, and the registration priority in the prediction motion vector candidate list is By allowing candidates that require a scaling operation only to the predicted motion vector candidates from the lower adjacent prediction block group, the rank of the motion vector candidate list becomes a predicted motion vector candidate that does not require scaling, and scaling is required. Predicted motion vector candidates that are registered in the order of predicted motion vector candidates and do not require reliable scaling can be registered in the candidate list of predicted motion vectors first. Will improve.

  FIG. 19 is a flowchart showing the derivation process procedure of step S1110 of FIG. The following processes are performed on the adjacent prediction blocks Nk (k = 0, 1, 2) in the order of k, 0, 1, 2 (S1301 to S1307).

  When the adjacent prediction block Nk can be used and the prediction mode PredMode of the prediction block Nk is not intra (MODE_INTRA) (YES in S1302), the condition determination of the above-described condition determination 3 is performed (S1303). When the flag predFlagLX [xNk] [yNk] indicating whether or not to use the LX of the adjacent prediction block Nk is 1, that is, when the adjacent prediction block Nk is inter-predicted using the same LX motion vector as the calculation target, step The process proceeds to S1304, and if not (NO in S1303), the condition is determined in Step S1305.

  When step S1303 is YES, the flag availableFlagLXN is set to 1, the mvLXN of the prediction block group N is set to the same value as the LX motion vector mvLX [xNk] [yNk] of the adjacent prediction block Nk, and the prediction block group N Is set to the same value as the LX reference index refIdxLX [xNk] [yNk] of the adjacent prediction block Nk, the reference list ListN of the prediction block group N is set to LX (S1304), and the process proceeds to step S1308. .

  When step S1303 is NO, the condition determination of the above-described condition determination 4 is performed (S1305). When the flag predFlagLY indicating whether to use the LY of the adjacent prediction block Nk is 1, that is, when the adjacent prediction block Nk is inter-predicted using a motion vector of LY different from the calculation target (YES in S1305), the flag availableFlagLXN Is set to 1, the prediction motion vector mvLXN of the prediction block group N is set to the same value as the LY motion vector mvLY [xNk] [yNk] of the adjacent prediction block Nk, and the prediction motion vector mvLXN of the prediction block group N is The LY motion vector mvLY [xNk] [yNk] of the adjacent prediction block Nk is set to the same value, and the reference index refIdxN of the prediction block group N is the LY reference index refIdxLY [xNk] [yNk] of the adjacent prediction block Nk , The reference list ListN of the prediction block group N is set to LY (S1306), step The process proceeds to S1308.

  On the other hand, if these conditions are not met (NO in S1303 and NO in S1305), k is incremented by 1 and the next adjacent prediction block is processed (S1301 to S1307), and availableFlagLXN becomes 1, The process is repeated until the processing of the adjacent block A1 is completed, and the process proceeds to step S1308.

  Subsequently, when availableFlagLXN is 1 (YES in S1308), mvLXN is scaled (S1309). The scaling calculation processing procedure of the spatial prediction motion vector candidate in step S1309 will be described with reference to FIGS.

  FIG. 20 is a flowchart showing the motion vector scaling processing procedure in step S1309 of FIG.

The inter-picture distance td is calculated by subtracting the POC of the reference picture referenced by the adjacent prediction block list ListN from the POC of the current encoding / decoding target picture (S1601). When the reference picture referred to in the adjacent prediction block list ListN is earlier in the display order than the current encoding / decoding target picture, the inter-picture distance td becomes a positive value, and the current encoding / decoding target When the reference picture referenced in the reference list ListN of the adjacent prediction block is later in the display order than the picture, the inter-picture distance td is a negative value.
td = POC of current encoding / decoding target picture—POC of reference picture referenced in reference list ListN of adjacent prediction block

Next, the inter-picture distance tb is calculated by subtracting the POC of the reference picture referenced by the current encoding / decoding target picture list LX from the POC of the current encoding / decoding target picture (S1602). When the reference picture referred to in the current encoding / decoding target picture list LX is earlier in the display order than the current encoding / decoding target picture, the inter-picture distance tb becomes a positive value, When the reference picture referred to in the encoding / decoding target picture list LX is later in the display order, the inter-picture distance tb is a negative value.
tb = POC of current encoding / decoding target picture-POC of reference picture referenced in reference list LX of current encoding / decoding target picture

Subsequently, scaling operation processing is performed by multiplying mvLXN by the scaling coefficient tb / td according to the following equation (S1603) to obtain a scaled motion vector mvLXN.
mvLXN = tb / td * mvLXN

  FIG. 21 shows an example of the case where the scaling operation in step S1603 is performed with integer precision arithmetic. The processing in steps S1604 to S1606 in FIG. 21 corresponds to the processing in step S1603 in FIG.

  First, as in the flowchart of FIG. 20, the inter-picture distance td and the inter-picture distance tb are calculated (S1601, S1602).

Subsequently, the variable tx is calculated by the following equation (S1603).
tx = (16384 + Abs (td / 2)) / td

Subsequently, the scaling coefficient DistScaleFactor is calculated by the following equation (S1605).
DistScaleFactor = (tb * tx + 32) >> 6

Subsequently, a scaled motion vector mvLXN is obtained by the following equation (S1606).
mvLXN = ClipMv (Sign (DistScaleFactor * mvLXN) * ((Abs (DistScaleFactor * mvLXN) + 127) >> 8))

  Next, a method of deriving motion vector predictor candidates from the prediction blocks of pictures at different times in S303 in FIG. 16 will be described in detail. FIG. 24 is a flowchart for explaining the prediction motion vector candidate derivation processing procedure in step S303 in FIG.

  First, based on slice_type and collocated_from_l0_flag, it is determined whether the picture colPic at a different time is defined as the L0 reference list or the L1 reference list (S2101 in FIG. 24).

  FIG. 25 is a flowchart for explaining the procedure for deriving the picture colPic at different times in step S2101 of FIG. When slice_type is B and the above-described flag collocated_from_l0_flag is 0 (YES in S2201 in FIG. 25, YES in S2202), RefPicList1 [0], that is, a picture with a reference index 0 in the reference list L1 is a picture colPic at a different time. (S2203 in FIG. 25). Otherwise, that is, when the above-described flag collocated_from_l0_flag is 1 (NO in S2201, NO in S2202, NO in S2204 in FIG. 25), RefPicList0 [0], that is, pictures with a reference index 0 in the reference list L0 are different times. The picture is colPic (S2205 in FIG. 25).

  Next, returning to the flowchart of FIG. 24, a prediction block colPU at a different time is calculated, and encoded information is acquired (S2102 of FIG. 24).

  FIG. 26 is a flowchart for explaining the procedure for deriving the prediction block colPU of the picture colPic at different times in step S2102 of FIG.

  First, a prediction block located in the lower right (outside) of the same position as the processing target prediction block in the picture colPic at a different time is set as a prediction block colPU at a different time (S2301 in FIG. 26). This prediction block corresponds to the prediction block T0 in FIG.

  Next, when it is outside the tree block of the prediction block coding block at different time or when PredMode of colPU is MODE_INTRA (S2302 and S2303 in FIG. 26), it is the same as the prediction block to be processed in the picture colPic at different time The prediction block located at the lower right center of the position is reset to the prediction block colPU at a different time (S2304 in FIG. 26). This prediction block corresponds to the prediction block T1 in FIG. When colPu is outside the tree block, the reason why colPu is changed to the same center lower right as the processing target prediction block is because colPU information acquisition from outside the tree block has a high memory access cost .

  Next, returning to the flowchart of FIG. 24, the LX prediction motion vector mvLXCol calculated from the prediction block of another picture at the same position as the prediction block to be encoded / decoded and the encoding information of the reference list LX of the prediction block group Col The flag availableFlagLXCol indicating whether or not is valid is calculated (S2103 in FIG. 24).

  FIG. 27 is a flowchart illustrating the inter prediction information derivation process in step S2103 of FIG.

  When PredMode of the prediction block colPU at different times is MODE_INTRA or cannot be used (NO in S2401 in FIG. 27, NO in S2402), availableFlagLXCol is set to 0, mvLXCol is set to (0, 0) (S2403, S2404 in FIG. 27), and processing Exit.

  When the prediction block colPU can be used and PredMode is not MODE_INTRA (YES in S2401 and YES in S2402 in FIG. 27), mvCol and refIdxCol are calculated in the following procedure.

  When the flag PredFlagL0 [xPCol] [yPCol] indicating whether or not the L0 prediction of the prediction block colPU is used (YES in S2405 in FIG. 27), since the prediction mode of the prediction block colPU is Pred_L1, the motion vector mvCol Is set to the same value as MvL1 [xPCol] [yPCol] which is the L1 motion vector of the prediction block colPU (S2406 in FIG. 27), and the reference index refIdxCol is set to the same value as the reference index RefIdxL1 [xPCol] [yPCol] of L1 It is set (S2407 in FIG. 27), and the list ListCol is set to L1 (S2408 in FIG. 27).

  On the other hand, if the L0 prediction flag PredFlagL0 [xPCol] [yPCol] of the prediction block colPU is not 0 (NO in S2405 in FIG. 27), it is determined whether the L1 prediction flag PredFlagL1 [xPCol] [yPCol] of the prediction block colPU is 0. . When the L1 prediction flag PredFlagL1 [xPCol] [yPCol] of the prediction block colPU is 0 (YES in S2409 in FIG. 27), the motion vector mvCol is the same as MvL0 [xPCol] [yPCol], which is the L0 motion vector of the prediction block colPU. 27 (S2410 in FIG. 27), the reference index refIdxCol is set to the same value as the reference index RefIdxL0 [xPCol] [yPCol] of L0 (S2411 in FIG. 27), and the list ListCol is set to L0 (FIG. 27). S2412).

  If both the L0 prediction flag PredFlagL0 [xPCol] [yPCol] of the prediction block colPU and the L1 prediction flag PredFlagL1 [xPCol] [yPCol] of the prediction block colPU are not 0 (NO in S2405 in FIG. 27, NO in S2409), the prediction block colPU Since the inter prediction mode is bi-prediction (Pred_BI), one of the two motion vectors L0 and L1 is selected (S2413 in FIG. 27).

  FIG. 28 is a flowchart illustrating a procedure for deriving inter prediction information of a prediction block when the inter prediction mode of the prediction block colPU is bi-prediction (Pred_BI).

  First, it is determined whether the POCs of all the pictures registered in all the reference lists are smaller than the POC of the current encoding / decoding target picture (S2501), and L0 and all the reference lists of the prediction block colPU are determined. When the POC of all the pictures registered in L1 is smaller than the POC of the current encoding / decoding target picture (YES in S2501), X is 0, that is, the L0 motion vector of the prediction block to be encoded / decoding When the prediction vector candidate is calculated (YES in S2502), the inter prediction information of L0 of the prediction block colPU is selected, and X is 1, that is, the motion vector of L1 of the prediction block to be encoded / decoded. When the prediction vector candidate is calculated (NO in S2502), the inter prediction information on the L1 side of the prediction block colPU is selected. . On the other hand, when at least one of the POCs of pictures registered in all the reference lists L0 and L1 of the prediction block colPU is larger than the POC of the current encoding / decoding target picture (NO in S2501), and the flag collocated_from_l0_flag is 0 (YES in S2503), the L0 inter prediction information of the prediction block colPU is selected. If the flag collocated_from_l0_flag is 1 (NO in S2503), the L1 inter prediction information of the prediction block colPU is selected. That is, when colPU is defined using L1, L0 is selected as mvCol, and when colPU is defined using L0, L1 is selected as mvCol. By reversing the selection list of colPU and the selection list of mvCol, it is possible to calculate a prediction vector with high accuracy that can perform interpolation prediction using a motion vector for a reference image directed from colPic toward the encoding target picture. .

  When the inter prediction information of L0 of the prediction block colPU is selected (YES in S2502, YES in S2503), the motion vector mvCol is set to the same value as MvL0 [xPCol] [yPCol] (S2504), and the reference index refIdxCol is set. The same value as RefIdxL0 [xPCol] [yPCol] is set (S2505), and the list ListCol is set to L0 (S2506).

  When the inter prediction information of L1 of the prediction block colPU is selected (NO in S2502 and NO in S2503), the motion vector mvCol is set to the same value as MvL1 [xPCol] [yPCol] (S2507), and the reference index refIdxCol is set. The same value as RefIdxL1 [xPCol] [yPCol] is set (S2508), and the list ListCol is set to L1 (S2509).

  Returning to FIG. 27, if inter prediction information can be acquired from the prediction block colPU, the availableFlagLXCol is set to 1 (S2414 in FIG. 27).

  Next, returning to the flowchart of FIG. 24, when availableFlagLXCol is 1 (YES in S2104 of FIG. 24), mvLXCol is scaled as necessary (S2105 of FIG. 24). A procedure for scaling operation of the motion vector mvLXCol will be described with reference to FIGS.

  FIG. 29 is a flowchart showing the motion vector scaling calculation processing procedure in step S2105 of FIG.

The inter-picture distance td is calculated by subtracting the POC of the reference picture referenced in the list ListCol of the prediction block colPU from the POC of the picture colPic at a different time (S2601). Note that when the POC of the reference picture referenced in the list Col of the prediction block colPU is earlier in the display order than the picture colPic at a different time, the inter-picture distance td is a positive value, and the prediction is more accurate than the picture colPic at a different time. When the POC of the reference picture referenced in the list ListCol of the block colPU is later in the display order, the inter-picture distance td is a negative value.
td = POC of picture colPic at different time-POC of reference picture referenced by list ListCol of prediction block colPU

The inter-picture distance tb is calculated by subtracting the POC of the reference picture referenced by the current encoding / decoding target picture list LX from the POC of the current encoding / decoding target picture (S2602). When the reference picture referred to in the current encoding / decoding target picture list LX is earlier in the display order than the current encoding / decoding target picture, the inter-picture distance tb becomes a positive value, When the reference picture referred to in the encoding / decoding target picture list LX is later in the display order, the inter-picture distance tb is a negative value.
tb = POC of current encoding / decoding target picture-POC of reference picture referenced in reference list LX of current encoding / decoding target picture

Subsequently, the inter-picture distances td and tb are compared (S2603). If the inter-picture distances td and tb are equal (YES in S2603), the scaling calculation process ends. When the inter-picture distances td and tb are not equal (NO in S2603), scaling operation processing is performed by multiplying mvLXCol by the scaling coefficient tb / td according to the following equation (S2604) to obtain a scaled motion vector mvLXCol.
mvLXCol = tb / td * mvLXCol

  FIG. 30 shows an example of the case where the scaling operation in step S2604 is performed with integer precision arithmetic. The processing in steps S2605 to S2607 in FIG. 30 corresponds to the processing in step S2604 in FIG.

  First, as in the flowchart of FIG. 29, the inter-picture distance td and the inter-picture distance tb are calculated (S2601, S2602).

Subsequently, the inter-picture distances td and tb are compared (S2603). If the inter-picture distances td and tb are equal (YES in S2603), the scaling calculation process ends. If the inter-picture distances td and tb are not equal (NO in S2603), a variable tx is calculated by the following equation (S2605).
tx = (16384 + Abs (td / 2)) / td

Subsequently, the scaling coefficient DistScaleFactor is calculated by the following equation (S2606).
DistScaleFactor = (tb * tx + 32) >> 6

Subsequently, a scaled motion vector mvLXN is obtained by the following equation (S2607).
mvLXN = ClipMv (Sign (DistScaleFactor * mvLXN) * ((Abs (DistScaleFactor * mvLXN) + 127) >> 8))

  The procedure for adding the predicted motion vector candidate in S304 of FIG. 16 to the predicted motion vector list will be described in detail.

  The candidate motion vector candidates mvLXA, mvLXB, and mvLXCol of LX calculated in S301, S302, and S303 of FIG. 16 are added to the LX motion vector predictor list mvpListLX (S304). In the embodiment of the present invention, by assigning priorities and registering predicted motion vector candidates mvLXA, mvLXB, and mvLXCol in the LX predicted motion vector list mvpListLX in order of priority, elements having higher priorities are registered. It is arranged in front of the predicted motion vector list. When the number of elements in the motion vector predictor list is limited in step S306 in FIG. 16, the elements placed at the rear of the motion vector predictor list are removed from the motion vector predictor list to leave elements with higher priority.

  The motion vector predictor list mvpListLX has a list structure, is managed by an index i indicating the location inside the motion vector predictor list, and has a storage area for storing a motion vector predictor candidate corresponding to the index i as an element. The number of the index i starts from 0, and motion vector predictor candidates are stored in the storage area of the motion vector predictor list mvpListLX. In the subsequent processing, a motion vector predictor candidate that is an element of the index i registered in the motion vector predictor list mvpListLX is represented by mvpListLX [i].

  Next, the registration processing method of LX predicted motion vector candidates mvLXA, mvLXB, and mvLXCol to the LX predicted motion vector list mvpListLX in S304 of FIG. 16 will be described in detail. FIG. 31 is a flowchart showing the motion vector predictor candidate registration processing procedure in step S304 of FIG.

  First, the index i of the motion vector predictor list mvpListLX is set to 0 (S3101). When availableFlagLXA is 1 (YES in S3102), the LX predicted motion vector candidate mvLXA is registered at the position corresponding to the index i in the predicted motion vector list mvpListLX (S3103), and is updated by adding 1 to the index i (S3103). S3104).

  Subsequently, when availableFlagLXB is 1 (YES in S3105), an LX predicted motion vector candidate mvLXB is registered at a position corresponding to the index i of the predicted motion vector list mvpListLXmvpListLX (S3106), and 1 is added to the index i. Update (S3107).

  Subsequently, when availableFlagLXCol is 1 (YES in S3108), the LX predicted motion vector candidate mvLXCol is registered at the position corresponding to the index i of the predicted motion vector list mvpListLX (S3109), and the registration process to the predicted motion vector list is performed. finish.

  The prediction motion vector list is registered in the order of the prediction candidate of the prediction block group adjacent on the left side, the prediction candidate of the prediction block group adjacent on the upper side, and the temporal prediction candidate. This is because the spatial prediction candidates that do not need scaling have higher prediction efficiency than the prediction candidates that require scaling, and when the number of motion vector predictor candidates to be described later is limited, the motion that is registered earlier in the motion vector predictor list is predicted motion. This is because they tend to remain as vector candidates. In the present embodiment, the number of motion vector predictors is limited to two from a total of three motion vector predictor candidates, that is, two spatial prediction candidates and one temporal prediction candidate. One reason for not detecting a motion vector that needs to be scaled in the detection of the motion vector predictor of the left prediction block group that is the point of the present embodiment is that a prediction candidate that does not require scaling is more likely than a prediction candidate that requires scaling. This is because prediction efficiency is high as a prediction vector candidate. This is because the prediction candidate that does not require scaling is a prediction candidate from the same reference image as the reference image of the motion vector to be encoded, so the most reliable difference compared to the motion vector prediction from other reference images This is because the probability that the motion vector becomes small is high. That is, if a prediction candidate that requires scaling is detected and registered in the prediction block group adjacent to the left side, the prediction motion vector of the prediction motion vector is detected when a prediction candidate that does not require scaling is detected in the prediction block group adjacent to the upper side. Since the registration order is the order of prediction candidates that require scaling and the prediction candidates that do not require scaling, the priority of prediction candidates that do not require scaling that is more easily selected is reduced. The present embodiment avoids the problem that the prediction efficiency decreases.

  Next, a method for deleting redundant prediction motion vector candidates in the prediction motion vector list in S305 of FIG. 16 will be described in detail. FIG. 32 is a flowchart showing the predicted motion vector redundant candidate deletion processing procedure in S305 of FIG.

  The motion vector redundancy candidate deletion units 123 and 223 compare spatial motion vector candidates registered in the motion vector predictor list mvpListLX (S4101), and a plurality of motion vector candidates have the same value. In this case (YES in S4102), the candidate is determined as a redundant motion vector candidate, and the redundant motion vector candidate is removed except for the motion vector candidate with the smallest order, that is, the smallest index i (S4103). Here, the motion vector candidates are not compared between the temporal prediction motion vector candidates and the spatial prediction motion vector candidates. Generation of temporal prediction motion vector candidates and spatial prediction motion vector candidates may be implemented by parallel processing because of different motion information memories to be accessed and different processing amounts. If the motion vector comparison between the temporal motion vector predictor candidate and the spatial motion vector predictor candidate is performed, a waiting time occurs when performing parallel processing. Therefore, since it is desirable that there is no dependency between the temporal prediction motion vector candidate and the spatial prediction motion vector candidate, the present invention does not perform motion vector redundancy comparison between the temporal prediction motion vector candidate and the spatial prediction motion vector candidate. Here, in order to further reduce the processing amount, it is also possible not to determine whether the prediction vectors between the spatial prediction motion vector candidates are the same. After the redundant motion vector candidates are removed, the predicted motion vector list mvpListLX has an empty storage area for the deleted predicted motion vector candidates. Are packed in the order of the predicted motion vector candidates with the smallest (S4104). For example, if the motion vector candidate mvListLX [1] with the index i = 1 is deleted and the motion vector candidates mvListLX [0], mvListLX [2] with the index i 0 and 2 remain, the index i is 2 The predicted motion vector candidate mvListLX [2] is newly changed to a predicted motion vector candidate mvListLX [1] whose index i is 1. Further, it is set that the motion vector candidate mvListLX [2] having the index i of 2 does not exist.

  Next, the method for limiting the number of motion vector candidates in the predicted motion vector list in S306 of FIG. 16 will be described in detail. FIG. 33 is a flowchart for explaining the predicted motion vector candidate number limit processing procedure in S306 of FIG.

  The predicted motion vector candidate number limit unit 124 and the predicted motion vector candidate number limit unit 224 limit the LX predicted motion vector candidate number numMVPCandLX registered in the LX predicted motion vector list mvpListLX to the defined final candidate number finalNumMVPCand.

  First, the number of elements registered in the LX prediction motion vector list mvpListLX is counted and set to the number of LX prediction motion vector candidates numMVPCandLX (S5101).

  Subsequently, when the number of predicted motion vectors numMVPCandLX of LX is smaller than the number of final candidates finalNumMVPCand (YES in S5103), the index i of the predicted motion vector list mvpListLX of LX has a value of (0, 0) at the position corresponding to numMVPCandLX. Is registered (S5104), and 1 is added to numMVPCandLX (S5105). If numMVPCandLX and finalNumMVPCand have the same value (YES in S5105), this restriction process is terminated. If numMVPCandLX and finalNumMVPCand are not the same value (NO in S5105), the processes in steps S5104 and S5105 are repeated until they have the same value. That is, MVs having a value of (0, 0) are registered until the LX predicted motion vector candidate number numMVPCandLX reaches the final candidate number finalNumMVPCand. In this way, by registering MVs having a value of (0, 0) overlappingly, the predicted motion vector is determined regardless of the value of the predicted motion vector index within the range of 0 or more and less than the final candidate number finalNumMVPCand. be able to.

  On the other hand, when the number of predicted motion vectors numMVPCandLX for LX is larger than the number of final candidates finalNumMVPCand (NO in S5103, YES in S5107), all predicted motion vectors for which the index i of the predicted motion vector list mvpListLX for LX is greater than finalNumMVPCand-1 The candidate is deleted (S5108), the same value as finalNumMVPCand is set in numMVPCandLX (S5109), and this restriction process ends.

  When the number of predicted motion vectors numMVPCandLX of LX is equal to the number of final candidates finalNumMVPCand (NO in S5103, NO in S5107), the limiting process is terminated without performing the limiting process.

  In the present embodiment, the prediction block is classified into two groups, that is, the prediction block group adjacent to the left side and the prediction block group adjacent to the upper side, and the first motion vector list registration order is prioritized. Candidates that need to be scaled for the group (predicted block group adjacent to the upper side) that does not generate a candidate that requires scaling for the group (predicted block group adjacent to the left side) and that do not prioritize the list registration order of the predicted motion vector However, the present invention is not limited to this, and can be applied to the case where the prediction blocks are classified into three or more groups. For example, the first group is a prediction block adjacent to the left side, the second group is a prediction block adjacent to the upper side, the third group is a prediction block adjacent to the upper left, and the first group, the second group, and the third group In the case of sequentially registering in the list of motion vector predictors, the number of scaling operations is not generated by generating candidates that require scaling for the second group and the third group without generating candidates that require scaling only for the first group. Reduce the number of scaling operations by generating candidates that require scaling for the third group without generating candidates that require scaling for the first group and the second group. Is also possible.

(Embodiment 2)
In the first embodiment, in the prediction block group adjacent to the left side, scanning of conditions that require scaling (condition determinations 3 and 4) is omitted, but in the second embodiment, it is adjacent to the left side. Even in a prediction block group to be predicted, a motion vector scan is performed under a condition that requires scaling, but the scaling operation is omitted even when the detected motion vector predictor is not a motion vector from the same reference image.

  In the second embodiment, in the prediction block group adjacent to the left side, the scanning of the condition requiring the scaling (condition determinations 3 and 4) is not limited, and the scaling is performed even when the condition requiring the scaling is used. Not performed. This simplifies the circuit configuration by having only one scaling operation circuit only for the upper adjacent prediction block group, and the prediction block group adjacent to the left side has only motion vectors that require scaling. Even in this case, it can be registered as a prediction vector candidate. Processes other than the scanning procedure of the prediction block group adjacent to the left side and the scaling calculation restriction are the same as those in the first embodiment.

  FIG. 22 is a flowchart showing the predicted motion vector candidate derivation process procedure of S301 and S302 of FIG. 16 in the second embodiment. The difference from FIG. 17 which is the first embodiment is only the processing after S1109.

  Regarding S1109, if there is no predicted motion vector candidate that matches the above-described condition determination 1 or condition determination 2 in the prediction block group adjacent to the left side in addition to the prediction block group adjacent to the upper side, the above-described condition determination is performed. In order to change to proceed to 3 or condition determination 4, the S1109 processing branch is deleted.

  A flowchart of S1113 is shown in FIG. The difference from FIG. 19 which is the first embodiment is only S1308. In S1308, the scaling operation is permitted only for the prediction vector candidate of the prediction block group adjacent on the upper side, and the prediction vector candidate of the prediction block group adjacent on the left side is a prediction vector that satisfies the above-described condition 3 and condition 4. Also, no scaling operation is performed.

  As a result, as in the first embodiment, since the scaling operation that requires division can be limited to only the prediction vector candidates of the upper adjacent prediction block group, the prediction of the prediction group adjacent to the upper side as the spatial prediction vector candidate can be performed. The processing is simplified by having only one scaling operation circuit only for the motion vector calculation processing.

(Embodiment 3)
In the third embodiment, the method for scanning a motion vector predictor candidate is changed in the prediction block group adjacent to the left side as compared with the second embodiment.
Specifically, the condition determination is performed on the prediction motion vector candidates of the prediction block group adjacent on the left side in the following priority order.
1. Prediction block A0 condition determination (same reference list LX)
2. Prediction block A0 condition determination (different reference list LY)
3. Prediction block A1 condition determination (same reference list LX)
4). Prediction block A1 condition determination (different reference list LY)

  The intention of the above processing is that the prediction block group adjacent to the left side does not perform the scaling operation regardless of whether the reference picture is the same or the same, thereby eliminating the condition determination as to whether or not the reference picture is the same. It is only to determine whether the reference lists are the same.

  A pattern that may have a different result from the second embodiment is that a predicted motion vector from a reference picture with a different prediction block A0 is a candidate, and a predicted motion vector from the same reference picture is a candidate for a prediction block A1. This is only the case. In any case of the second embodiment and the third embodiment, since the scaling operation is not performed, the prediction efficiency is hardly lowered.

  Compared to the third embodiment and the second embodiment, it is possible to halve the condition judgment of the prediction block group adjacent to the left side where the scaling operation is not performed, and the reference image index is used to calculate the motion vector predictor candidate. Since the processing amount can be greatly reduced compared to the prediction block group adjacent to the upper side where the scaling operation is performed, the prediction block group adjacent to the left side and the prediction block group adjacent to the upper side are parallel. There is no need to consider processing and the like, and the circuit scale can be reduced.

  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 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 recorded 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. .

  DESCRIPTION OF SYMBOLS 101 Image memory, 102 Motion vector detection part, 103 Difference motion vector calculation part, 104 Inter prediction information estimation part, 105 Motion compensation prediction part, 106 Prediction method determination part, 107 Residual signal generation part, 108 Orthogonal transformation and quantization part 109 first encoded bit string generation unit, 110 second encoded bit string generation unit, 111 multiplexing unit, 112 inverse quantization / inverse orthogonal transform unit, 113 decoded image signal superimposing unit, 114 encoded information storage memory, 115 decoded image memory, 121 prediction motion vector candidate generation unit, 122 prediction motion vector candidate registration unit, 123 prediction motion vector redundant candidate deletion unit, 124 prediction motion vector candidate number limit unit, 125 prediction motion vector candidate code amount calculation unit, 126 Predicted motion vector selection unit, 127 motion Coulter subtraction unit, 201 separation unit, 202 first encoded bit sequence decoding unit, 203 second encoded bit sequence decoding unit, 204 motion vector calculation unit, 205 inter prediction information estimation unit, 206 motion compensation prediction unit, 207 inverse quantization / Inverse orthogonal transform unit, 208 decoded image signal superimposing unit, 209 encoded information storage memory, 210 decoded image memory, 221 prediction motion vector candidate generation unit, 222 prediction motion vector candidate registration unit, 223 prediction motion vector redundant candidate deletion unit, 224 A motion vector predictor candidate limit unit; 225 motion vector predictor selection unit; and 226 motion vector addition unit.

Claims (8)

A moving picture decoding apparatus for decoding a coded bit string obtained by coding the moving picture using motion compensation in units of blocks obtained by dividing each picture of the moving picture,
A decoding unit that decodes information indicating a predicted motion vector to be selected together with a difference motion vector;
A prediction motion vector candidate generation unit that generates prediction motion vector candidates by predicting from motion vectors of decoded adjacent blocks spatially or temporally adjacent to the decoding target block;
A prediction motion vector selection unit that selects a prediction motion vector from the plurality of prediction motion vector candidates based on the decoded information indicating the prediction motion vector to be selected;
A motion vector calculator that calculates a motion vector used for motion compensated prediction by adding the selected motion vector predictor and the difference motion vector;
The motion vector predictor candidate generation unit, for a group having the highest priority of at least the motion vector list registration order among a plurality of groups formed by spatially adjacent decoded adjacent blocks, Predictive motion vector candidates that do not derive scaling motion vectors from motion vectors and that have the lowest priority in the motion vector list registration order at least for predicted motion vectors that use scaling operations from adjacent block motion vectors A moving picture decoding apparatus characterized by permitting derivation of a vector candidate.   The motion vector predictor candidate generation unit performs motion of the adjacent block for a group of which priority is given to the registered motion vector list among the two groups formed by spatially adjacent decoded adjacent blocks. Predicted motion vector candidates using scaling operations are not derived from the vectors. For groups that do not prioritize the motion vector list registration order, predicted motion vector candidates that use scaling operations from adjacent block motion vectors. The video decoding device according to claim 1, wherein derivation of the video is permitted.   The predicted motion vector candidate generation unit prioritizes the list registration order of the motion vector predictor when the predicted motion vector of the group of which priority is given to the list motion motion vector list among the two groups cannot be used. The moving picture decoding apparatus according to claim 2, wherein registration of a plurality of motion vector predictor candidates derived from a group that is not performed is permitted.   The predicted motion vector candidate generation unit prioritizes the list registration order of the motion vector predictor when the predicted motion vector of the group of which priority is given to the list motion motion vector list among the two groups cannot be used. 4. The first motion vector predictor candidate derived from the group not to be used is substituted as a motion vector predictor candidate for the group in which the list registration order of the motion vector predictor is prioritized. The moving picture decoding apparatus described. A moving picture decoding method for decoding a coded bit string obtained by coding the moving picture using motion compensation in units of blocks obtained by dividing each picture of the moving picture,
A decoding step of decoding information indicating a predicted motion vector to be selected together with a difference motion vector;
A prediction motion vector candidate generation step of generating a plurality of motion vector predictor candidates by predicting from motion vectors of decoded adjacent blocks spatially or temporally adjacent to the decoding target block;
A predicted motion vector selection step of selecting a predicted motion vector from the plurality of predicted motion vector candidates based on the decoded information indicating the predicted motion vector to be selected;
A motion vector calculating step of calculating a motion vector used for motion compensated prediction by adding the selected motion vector predictor and the difference motion vector;
In the prediction motion vector candidate generation step, at least a group having a highest priority in the list registration order of prediction motion vectors among a plurality of groups formed by spatially adjacent decoded adjacent blocks Predictive motion vector candidates that do not derive scaling motion vectors from motion vectors and that have the lowest priority in the motion vector list registration order at least for predicted motion vectors that use scaling operations from adjacent block motion vectors A moving picture decoding method characterized by permitting derivation of a vector candidate.   In the predicted motion vector candidate generation step, if the predicted motion vector of the group in which the list registration order of the predicted motion vector is prioritized among the two groups cannot be used, the list registration order of the predicted motion vector has priority. A plurality of motion vector predictor candidates derived from a group not to be registered are registered, and the motion vector list registration order of the motion vector predictor is not prioritized and the first motion vector predictor candidate derived from the group to which priority is not given 6. The moving picture decoding method according to claim 5, wherein the predicted motion vector list registration order is used as a candidate for a predicted motion vector of a group having priority. A moving picture decoding program for decoding a coded bit string obtained by coding the moving picture using motion compensation in units of blocks obtained by dividing each picture of the moving picture,
A decoding step of decoding information indicating a predicted motion vector to be selected together with a difference motion vector;
A prediction motion vector candidate generation step of generating a plurality of motion vector predictor candidates by predicting from motion vectors of decoded adjacent blocks spatially or temporally adjacent to the decoding target block;
A predicted motion vector selection step of selecting a predicted motion vector from the plurality of predicted motion vector candidates based on the decoded information indicating the predicted motion vector to be selected;
Causing the computer to execute a motion vector calculating step of calculating a motion vector used for motion compensated prediction by adding the selected predicted motion vector and the difference motion vector;
In the prediction motion vector candidate generation step, at least a group having a highest priority in the list registration order of prediction motion vectors among a plurality of groups formed by spatially adjacent decoded adjacent blocks Predictive motion vector candidates that do not derive scaling motion vectors from motion vectors and that have the lowest priority in the motion vector list registration order at least for predicted motion vectors that use scaling operations from adjacent block motion vectors A moving picture decoding program characterized by permitting derivation of vector candidates.   In the predicted motion vector candidate generation step, if the predicted motion vector of the group in which the list registration order of the predicted motion vector is prioritized among the two groups cannot be used, the list registration order of the predicted motion vector has priority. A plurality of motion vector predictor candidates derived from a group not to be registered are registered, and the motion vector list registration order of the motion vector predictor is not prioritized and the first motion vector predictor candidate derived from the group to which priority is not given 8. The moving picture decoding program according to claim 7, wherein the moving picture decoding program is used as a candidate for a motion vector predictor of a group whose priority is a list registration order of motion vector predictors.

Images