JP2014135744A - Moving image decoding method, moving image decoder and moving image decoding program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve prediction accuracy of a prediction vector candidate.SOLUTION: In a moving image decoding method, when motion vector information including a motion vector mvc (mvcx, mvcy) of a block which is spatially or temporally adjacent to picture information of a picture to which a processing object block can refer and the processing object block and a reference picture identifier showing a picture to which the motion vector mvc refers is used, and a prediction vector candidate mvc' to the motion vector mv of the processing object block is scaled by correcting the prediction vector candidate in a zero direction by using a prescribed mathematical formula when the motion vector mvc of the adjacent block is scaled and calculated and when a Scale coefficient has N bits as accuracy or a decimal point or below and N=8 bits, and a difference with the motion vector of the processing object block is calculated by using the scaled prediction vector candidate mvc'.

Description

  The present invention relates to a moving picture decoding method, a moving picture decoding apparatus, and a moving picture decoding program that divide each picture into a plurality of blocks and perform motion compensation for each block.

  In recent video coding, a high compression rate is achieved by dividing an image into blocks, predicting pixels included in the block, and coding a prediction difference. A prediction mode that configures a prediction pixel from pixels in a picture to be encoded is called intra prediction, and a prediction mode that configures a prediction pixel from a previously encoded reference image called motion compensation is called inter prediction.

  In the inter-prediction apparatus, in inter prediction, a region referred to as a prediction pixel is expressed by two-dimensional coordinate data of horizontal and vertical components called motion vectors, and motion vector and pixel prediction difference data are encoded. In order to suppress the coding amount of the motion vector, a prediction vector is generated from a motion vector of a block adjacent to the encoding target block, and a difference vector between the motion vector and the prediction vector is encoded. By assigning a smaller code amount to a smaller difference vector, the code amount of the motion vector can be reduced, and the code efficiency can be improved.

  Also in the video decoding apparatus, the same prediction vector as that of the video encoding apparatus is determined for each block, and the motion vector is restored by adding the encoded difference vector and the prediction vector. Therefore, the moving image encoding device and the moving image decoding device include the same motion vector prediction unit.

  In the moving image decoding apparatus, each block is generally decoded in the order of raster scan or z scan from the upper left to the lower right of the image. Therefore, the motion vectors of peripheral blocks that can be used for prediction by the motion vector prediction unit in the video encoding device and the video decoding device are already decoded when the processing target block is decoded by the video decoding device. It becomes the motion vector of the block adjacent to the top.

  Furthermore, MPEG (Moving Picture Experts Group) -4 AVC / H. In H.264 (hereinafter also referred to as H.264), a prediction vector may be determined using a motion vector of a reference picture that has been encoded and decoded in the past instead of a processing target picture (Non-patent Document 1).

  As a conventional technique of a prediction vector determination method, a technique of HEVC (High Efficiency Video Coding) which is an international standardization organization ISO / IEC and ITU-T jointly studying standardization is disclosed (non-patent document). Reference 2). As reference software, HM Software (Version 3.0) is disclosed.

  Below, the outline | summary description regarding HEVC is given. In HEVC, there are two lists, L0 and L1, as lists of pictures that can be referred to (hereinafter also referred to as reference picture lists). Each block can use a maximum of two reference picture areas for inter prediction by motion vectors corresponding to L0 and L1, respectively.

  L0 and L1 generally correspond to the direction of display time, L0 is a reference list of past pictures with respect to the processing target picture, and L1 is a reference list of future pictures. Each entry of the reference picture list has information including a storage position of pixel data and display time information POC (Picture Order Count) value of the picture.

The POC is an integer value representing the display time relative to the display order of each picture. When the display time of a picture with a POC value of 0 is assumed to be 0, the display time of a picture can be represented by a constant multiple of the POC value of that picture. For example, the display time of a picture in which the frame display period (Hz) is fr and the POC value is p can be expressed by Expression (1). Thereby, POC can be regarded as a display time in units of a certain constant (second).
Display time = p × (fr / 2) (1)
When the number of entries in one reference picture list is 2 or more, each motion vector specifies which reference picture is to be referred to by an index number (reference index) in the reference picture list. In particular, when the number of entries in the reference picture list includes only one picture, the reference index of the motion vector corresponding to that list is automatically 0, so there is no need to explicitly specify the reference index.

  That is, the motion vector of the block includes an L0 / L1 list identifier, a reference index, and vector data (Vx, Vy). A reference picture is designated by the L0 / L1 list identifier and the reference index. An area in the reference picture is specified by (Vx, Vy). Vx and Vy are the differences between the coordinates of the reference area and the coordinates of the processing target block (also referred to as the current block) in the horizontal direction and the vertical direction, respectively. The L0 / L1 list identifier and the reference index are referred to as a reference picture identifier, (Vx, Vy) is referred to as vector data, and the (0, 0) vector is referred to as a 0 vector.

  A method for determining a prediction vector in HEVC will be described. The prediction vector is determined for each reference picture specified by the L0 / L1 list identifier and the reference index. When determining vector data mvp of a prediction vector for a motion vector that refers to a reference picture specified by a reference picture list of LX and a reference index of refidx, a maximum of three vector data are calculated as prediction vector candidates.

  Blocks adjacent to the processing target block in the spatial direction and time direction are classified into three blocks: a block adjacent in the left direction, a block adjacent in the upward direction, and a block adjacent in the time direction. A maximum of one prediction vector candidate is selected from each of these three groups.

  The selected prediction vector candidates are listed in the priority order of the group adjacent in the time direction, the group adjacent to the left, and the group adjacent above. This list is an array mvp_cand. If no prediction vector candidate exists in all groups, the 0 vector is added to mvp_cand.

  Further, the prediction candidate index mvp_idx is used as an identifier of which prediction vector candidate in the candidate list is used as the prediction vector. The prediction vector mvp is vector data of a prediction vector candidate entered in the mvp_idxth of mvp_cand.

In the moving image encoding device, when the motion vector referring to the LX refidx of the block to be encoded is mv, a prediction vector candidate closest to mv is searched for in mvp_cand, and its index is set to mvp_idx. Further, the difference vector mvd is calculated by the equation (2), and refidx, mvd, and mvp_idx are encoded as the motion vector information of the list LX.
mvd = mv−mvp (2)
The moving picture decoding apparatus decodes refidx, mvd, and mvp_idx, determines mvp_cand based on refidx, and sets the prediction vector mvp as the mvp_idxth prediction vector candidate of mvp_cand. The moving image decoding apparatus restores the motion vector mv of the processing target block based on Expression (3).
mv = mvd + mvp (3)
Next, blocks adjacent in the spatial direction will be described. FIG. 1 is a diagram for explaining the prior art (part 1). In the example illustrated in FIG. 1, a procedure for selecting a prediction vector candidate from the left adjacent block and the upper adjacent block will be described.

  First, a procedure for selecting a prediction vector candidate from the left adjacent block will be described. Search is performed in the order of block I and block H adjacent to the left side of the processing target block until a motion vector having a reference index equal to refidx with a reference list of LX is found. If a motion vector 1 having a reference index equal to refidx in the reference list is found, this motion vector 1 is selected.

  If motion vector 1 is not found, if there is motion vector 2 that refers to the same reference picture as the reference picture indicated by refidx in reference list LX in reference list LY that is not LX, motion vector 2 is selected. .

  If motion vector 2 is not found, if motion vector 3 by inter prediction exists, motion vector 3 is selected. When a motion vector that refers to the same reference picture as the reference picture indicated by refidx in the reference list LX is not selected, scaling processing described later is performed.

  Next, a procedure for selecting a prediction vector candidate from the upper adjacent block will be described. Search is performed in the order of block E, block D, and block A adjacent to the upper side of the processing target block until a motion vector 1 having a reference index equal to refidx is found in the reference list LX. If a motion vector 1 having a reference index equal to refidx in the reference list is found, this motion vector 1 is selected.

  If motion vector 1 is not found, if there is motion vector 2 that refers to the same reference picture as the reference picture indicated by refidx in reference list LX in reference list LY that is not LX, motion vector 2 is selected. .

  If motion vector 2 is not found, if motion vector 3 by inter prediction exists, motion vector 3 is selected. When a motion vector that refers to the same reference picture as the reference picture indicated by refidx in the reference list LX is not selected, scaling processing described later is performed.

  Next, blocks adjacent in the time direction will be described. FIG. 2 is a diagram for explaining the prior art (part 2). In the example illustrated in FIG. 2, a procedure for selecting a prediction vector candidate from blocks adjacent in the time direction will be described.

  First, a reference picture 20 adjacent in the time direction called Collated Picture (hereinafter also referred to as ColPic) is designated as a picture including blocks adjacent in the time direction. ColPic20 is a reference picture with reference index 0 in the reference list of either L0 or L1. Normally, reference index 0 of L1 is ColPic.

  A motion vector included in a block (Col block) 21 at the same position as the processing target block 11 in the ColPic 20 is set as mvCol22, and the mvCol22 is scaled by a scaling method described later to generate a prediction vector candidate.

  A motion vector scaling method is described. The input motion vector is mvc = (mvcx, mvcy), and the output vector (predicted vector candidate) is mvc ′ = (mvcx ′, mvcy ′). Next, a case where mvc is mvCol will be described as an example.

  A picture referred to by mvc is referred to as ColRefPic23. The POC value of the picture 20 having mvc is ColPicPoc, and the POC value of ColRefPic23 is ColRefPoc. The POC value of the current processing target picture 10 is CurPoc, and the POC value of the picture 25 specified by RefPicList LX and RefIdx is CurrRefPoc.

  Note that when the motion vector to be scaled is a motion vector of a block adjacent in the spatial direction, ColPicPoc is equal to CurrPOC. In addition, when the motion vector to be scaled is a motion vector of a block adjacent in the time direction, ColPicPoc is equal to the POC value of ColPic.

mvc is scaled based on the ratio of the time intervals of pictures as shown in equations (4) and (5).
mvcx ′ = mvcx × (CurrPoc−CurRefPoc) ÷ (ColPicPoc−ColRefPoc) (4)
mvcy ′ = mvcy × (CurrPoc−CurRefPoc) ÷ (ColPicPoc−ColRefPoc) (5)
However, since division requires a large amount of calculation, for example, mvc ′ is calculated by approximation by multiplication and shift as follows.
DiffPocD = ColPicPOC-ColRefPOC (6)
DiffPocB = CurrPOC-CurrRefPOC (7)
Then,
TDB = Clip3 (−128, 127, DiffPocB) (8)
TDD = Clip3 (−128, 127, DiffPocD) (9)
iX = (0x4000 + abs (TDD / 2)) / TDD (10)
Scale = Clip3 (−1024, 1023, (TDB × iX + 32) >> 6) (11)
abs (•): Function that returns absolute value Clip3 (x, y, z): Function that returns median value of x, y, z >>: Arithmetic right shift The finally obtained Scale is used as a scaling factor. In this example, when Scale is 256, it means that the coefficient is one time, that is, it is not scaled. At this time, the Scale coefficient has 8 bits as the precision after the decimal point, and the motion vector is expanded by 8 bits below the decimal point by multiplying by the scaling coefficient.

Next, the scaling operation performed based on the scaling coefficient is calculated as follows.
mvcx ′ = (Scale × mvcx + 128) >> 8 Expression (12)
mvcy ′ = (Scale × mvcy + 128) >> 8 Expression (13)
As is obvious to those skilled in the art, adding 2 ^N-1 and shifting N bits to the right means rounding the value of N bits after the decimal point to the nearest integer. In addition, this scaling processing is performed by H.264. The same processing is performed in H.264 (Non-Patent Document 1). This mvc ′ is a prediction vector candidate.

ISO / IEC 14496-10 (MPEG-4 Part 10) / ITU-T Rec.H.264 Thomas Wiegand, Woo-Jin Han, Benjamin Bross, Jens-Rainer Ohm, Gary J. Sullivan, "WD3: Working Draft 3 of High-Efficiency Video Coding" JCTVC-E603, JCT-VC 5th Meeting, March 2011

  HEVC and H.C. In H.264, the motion between frames is represented by a motion vector for each block. Usually, when a prediction vector is generated from a block adjacent in the time direction, the motion vector of the block adjacent in the time direction is scaled. At this time, the difference T1 between the display time of the processing target picture and the display time of the picture referred to by the processing target block, the display time of the picture to which the block adjacent in the time direction belongs, and the motion vector of the block adjacent in the time direction are The difference T2 from the display time of the picture to be referenced is different.

  Here, in order to make the motion amount per unit time constant, the motion vector is scaled by the display time difference ratio (T1 ÷ T2). At this time, for example, as shown in equations (12) and (13), scaling is performed with a scaling coefficient having a precision below a predetermined decimal point, and a vector having an integer value closest to the scaled vector is assumed as a prediction vector. There is a problem that the accuracy of the prediction vector does not increase.

  Therefore, the disclosed technique aims to improve the accuracy of a prediction vector.

A moving image decoding method according to an aspect of the disclosure includes a moving image decoding apparatus that performs a decoding process using a motion vector of a processing target block and a prediction vector candidate of the motion vector for each processing target block. A motion vector mvc having picture information of a picture that can be referred to by the processing target block, and a horizontal component mvcx and a vertical component mvcy of a block spatially or temporally adjacent to the processing target block; Using motion vector information including a reference picture identifier indicating a picture to which a motion vector mvc refers, a predicted vector candidate mvc ′ for the motion vector mv of the processing target block is scaled with a motion vector mvc of the adjacent block In the case of calculation, the Scale coefficient should be accurate to the decimal point. N bits and N = 8 bits
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 128) >> 8}
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 128) >> 8}
abs (·): function that returns an absolute value sign (·): a function that returns a sign (1 or −1), and a predetermined amount a is set in the 0 direction within a range of 1 ≦ a ≦ 2 ^N−2. And the difference from the motion vector of the block to be processed is calculated using the scaled prediction vector candidate mvc ′ = (mvcx ′, mvcy ′).

  According to the disclosed technique, the accuracy of a prediction vector can be improved.

The figure for demonstrating a prior art (the 1). The figure for demonstrating a prior art (the 2). The figure which shows the relationship between mvp 'and mvCol. The figure which shows the appearance probability distribution of mv in case mvp 'is positive. The figure which shows the appearance probability distribution of mv in case mvp 'is negative. 1 is a block diagram illustrating an example of a configuration of a video decoding device in Embodiment 1. FIG. FIG. 3 is a block diagram illustrating an example of a configuration of a prediction vector generation unit according to the first embodiment. FIG. 3 is a block diagram illustrating an example of a configuration of a scaling calculation unit according to the first embodiment. The block diagram which shows an example of the structure (the 1) of each part of a scaling calculating part. The block diagram which shows an example of the structure (the 2) of each part of a scaling calculating part. The figure for demonstrating the process of the scaling calculating part using a specific example. The block diagram which shows an example of the structure (the 3) of each part of a scaling calculating part. 5 is a flowchart illustrating an example of processing of the video decoding device in the first embodiment. 9 is a flowchart illustrating an example of a process (part 1) by a prediction vector generation unit according to the first embodiment. 9 is a flowchart illustrating an example of a process (part 2) by a prediction vector generation unit according to the first embodiment. FIG. 10 is a block diagram illustrating an example of a configuration of a prediction vector generation unit according to the second embodiment. 10 is a flowchart illustrating an example of processing performed by a prediction vector generation unit according to the second embodiment. FIG. 10 is a block diagram illustrating an example of a configuration of a prediction vector generation unit according to a third embodiment. 10 is a flowchart illustrating an example of a process (part 1) by a prediction vector generation unit according to the third embodiment. 10 is a flowchart illustrating an example of a process (part 2) by a prediction vector generation unit according to the third embodiment. FIG. 10 is a block diagram illustrating an example of a configuration of a prediction vector generation unit according to a fourth embodiment. 10 is a flowchart illustrating an example of a process (part 1) by a prediction vector generation unit according to the fourth embodiment. 10 is a flowchart illustrating an example of a process (part 2) performed by a prediction vector generation unit according to the fourth embodiment. FIG. 10 is a block diagram illustrating an example of a configuration of a moving image encoding device according to a fifth embodiment. The flowchart which shows an example of a process of a moving image encoder. 1 is a block diagram illustrating an example of a configuration of an image processing device.

  First, verification regarding the accuracy of the prediction vector by the inventors will be described. FIG. 3 is a diagram illustrating the relationship between mvp ′ and mvCol. As illustrated in FIG. 3, the motion vector of a block (Col block) 21 adjacent to the processing target block 11 in the time direction is mvCol22, and the motion vector of the processing target block is mv.

In addition, a motion vector (predicted vector candidate) obtained by scaling mvCol22 with infinite real number precision at the ratio of (T1 ÷ T2) described above is mvp ′.
mvp ′ = mvCol × (T1 ÷ T2) Expression (14)
When a prediction vector equal to mv is selected from the prediction vector candidates, the difference vector becomes 0 and the encoding efficiency is improved. Therefore, it is important to increase the coding efficiency that mvp ′ is equal to or approximate to mv. Here, the inventors examined the error between the scaled mvp ′ and mv.

  FIG. 4 is a diagram illustrating an appearance probability distribution of mv when mvp ′ is positive. The probability distribution shown in FIG. 4 shows a case where attention is paid to the horizontal component of the vector. FIG. 5 is a diagram showing an application probability distribution of mv when mvp ′ is negative. The probability distribution shown in FIG. 5 also shows a case where attention is paid to the horizontal component of the vector.

  The inventors have found that when mvp ′ and mv calculated from the motion vector to be scaled are compared, the appearance frequency of mv increases in the 0 vector direction as shown in FIGS. 4 and 5. Therefore, in the embodiment described below, the prediction vector candidate is corrected in the direction of the zero vector when scaling with the scaling coefficient. Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

[Example 1]
<Configuration>
FIG. 6 is a block diagram illustrating an example of the configuration of the video decoding device 100 according to the first embodiment. A moving picture decoding apparatus 100 shown in FIG. 6 includes an entropy decoding unit 101, a reference picture list storage unit 102, a motion vector information storage unit 103, a prediction vector generation unit 104, a motion vector restoration unit 105, a prediction pixel generation unit 106, an inverse quantum. A conversion unit 107, an inverse orthogonal transform unit 108, a decoded pixel generation unit 109, and a decoded image storage unit 110.

  The entropy decoding unit 101 performs entropy decoding on the compressed stream, and decodes the L0 and L1 reference indexes, the difference vector, the prediction candidate index, and the orthogonal transform coefficient of the processing target block.

  The reference picture list storage unit 102 stores picture information including a POC of a picture that can be referred to by the processing target block, a storage position of image data, and the like.

  The motion vector information storage unit 103 stores motion vector information including motion vectors of temporally and spatially adjacent blocks of the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers. The motion vector information is generated by the motion vector restoration unit 105.

  The prediction vector generation unit 104 acquires a reference index (reference picture identifier) between L0 and L1 from the entropy decoding unit 101, and generates a prediction vector candidate list for the motion vector of the processing target block. Details of the prediction vector generation unit 104 will be described later.

  The motion vector restoration unit 105 acquires the prediction candidate index and difference vector of L0 and L1 from the entropy decoding unit 101, adds the prediction vector candidate and the difference vector indicated by the prediction candidate index, and restores the motion vector.

  The predicted pixel generation unit 106 generates a predicted pixel signal using the restored motion vector and the decoded image stored in the decoded image storage unit 110.

  The inverse quantization unit 107 performs an inverse quantization process on the orthogonal transform coefficient acquired from the entropy decoding unit 101. The inverse orthogonal transform unit 108 performs an inverse orthogonal transform process on the inversely quantized output signal to generate a prediction error signal. The prediction error signal is output to the decoded pixel generation unit 109.

  The decoded pixel generation unit 109 adds the prediction pixel signal and the prediction error signal to generate a decoded pixel.

  The decoded image storage unit 110 stores a decoded image including the decoded pixel generated by the decoded pixel generation unit 109. The decoded image stored in the decoded image storage unit 110 is output to a display unit such as a display.

  Next, the prediction vector generation unit 104 will be described in detail. FIG. 7 is a block diagram illustrating an example of the configuration of the prediction vector generation unit 104 according to the first embodiment. The prediction vector generation unit 104 illustrated in FIG. 7 includes a scaling coefficient calculation unit 201, a vector information acquisition unit 202, and a scaling calculation unit 203.

  The prediction vector generation unit 104 receives the reference picture identifier of the processing target block and the POC information of the processing target block. Here, the reference list identifier included in the reference picture identifier of the processing target block is LX, and the reference index is refidx.

  The motion vector information storage unit 103 holds motion vector information of blocks processed in the past. The motion vector information stores the identifier of the picture to which the block having each motion vector belongs, the identifier of the picture that becomes the reference destination of the motion vector (reference picture identifier), and the values of the horizontal and vertical components of the motion vector. .

  The vector information acquisition unit 202 acquires motion vector information of a block adjacent to the processing target block from the motion vector information storage unit 103. The motion vector information includes a motion vector, an identifier of a picture to which a block having the motion vector belongs, and a reference picture identifier of a motion vector reference destination.

  The vector information acquisition unit 202 sequentially acquires motion vector information of blocks adjacent to the processing target block in the spatial direction or the time direction. In this process, as described above, first, a motion vector of a block adjacent in the left direction is searched. The vector information acquisition unit 202 searches until a motion vector having a reference index equal to refidx in the reference list is found, and if found, selects the motion vector.

  If this motion vector is not found, the vector information acquisition unit 202, if there is a motion vector that refers to the same reference picture as the reference picture indicated by the refidx of the reference list LX in the reference list LY that is not LX, Select.

  Furthermore, if this motion vector is not found, the vector information acquisition unit 202 selects the motion vector if a motion vector exists. When a motion vector that refers to the same reference picture as the reference picture indicated by refidx in the reference list LX is not selected, scaling processing is performed next. The vector information acquisition unit 202 outputs the acquired motion vector information to the scaling coefficient calculation unit 201.

  The scaling coefficient calculation unit 201 acquires the motion vector information from the vector information acquisition unit 202, acquires the POC value of the picture to which the motion vector belongs from the reference picture list storage unit 102, and calculates the scaling coefficient.

  Here, the POC value of the processing target picture is CurrPOC. The scaling coefficient calculation unit 201 receives, from the reference picture list storage unit 102, the POC value CurrRefPoc of the picture referenced by the processing target block, the POC value ColPicPoc of the picture to which the motion vector to be scaled belongs, and the picture referenced by this motion vector. The POC value ColRefPoc is acquired.

The scaling coefficient calculation unit 201 calculates the scaling coefficient by the following formula.
The scaling factor is calculated as follows:
DiffPocD = ColPicPOC-ColRefPOC (6)
DiffPocB = CurrPOC-CurrRefPOC (7)
TDB = Clip3 (−128, 127, DiffPocB) (8)
TDD = Clip3 (−128, 127, DiffPocD) (9)
iX = (0x4000 + abs (TDD / 2)) / TDD (10)
Scale = Clip3 (−1024, 1023, (TDB * iX + 32) >> 6) Expression (11)
abs (x): a function that returns the absolute value of x Clip3 (x, y, z): a function that returns the median of x, y, z >>: arithmetic right shift The calculated scaling factor is the precision after the decimal point Has 8-bit precision. The scaling coefficient calculation unit 201 outputs the calculated scale coefficient Scale to the scaling calculation unit 203.

  The scaling calculation unit 203 scales the motion vector based on the motion vector information acquired from the vector information acquisition unit 202 and the scaling coefficient acquired from the scaling coefficient calculation unit 201.

  FIG. 8 is a block diagram illustrating an example of the configuration of the scaling calculation unit 203 according to the first embodiment. The scaling operation unit 203 receives the scaling coefficient calculated by the scaling coefficient calculation unit 201 and the scaling target motion vector (mvcx, mvcy) acquired by the vector information acquisition unit 202. The motion vector to be scaled is a prediction vector candidate. The scaling operation unit 203 outputs the motion vector (mvcx ′, mvcy ′) after scaling. The output motion vector becomes the final prediction vector candidate.

  A scaling calculation unit 203 illustrated in FIG. 8 includes a scaling unit 301, a correction unit 302, and an adjustment unit 303. The scaling coefficient has accuracy below a predetermined decimal point, and the scaling unit 301 obtains a scaled prediction vector candidate by multiplying the prediction vector candidate by the scaling coefficient. The scaled prediction vector candidate is expanded to the precision below the decimal point of the scaling coefficient.

The correction unit 302 corrects the prediction vector candidate scaled by the scaling unit 301 by a predetermined amount in the 0 direction. The adjustment unit 303 converts the predicted vector candidate that has been scaled by the scaling factor having the precision below a predetermined decimal point by the scaling unit 301 and corrected by the correction unit 302 into a motion vector having the closest integer value. Hereinafter, detailed operations of the scaling unit 301, the correction unit 302, and the adjustment unit 303 will be described.
The scaling unit 301 calculates a scaling coefficient Scale × motion vector (mvcx, mvcy). When the scaling coefficient has N-bit precision as the precision after the decimal point, the precision after the decimal point of the motion vector after scaling obtained by multiplying the scaling coefficient is extended to N bits.

The correction unit 302 subtracts the predetermined amount a after calculating the absolute value in order to correct the scaled motion vector in the 0 direction. The adjustment unit 303 adds 2 ^N−1 and then right shifts N bits to round the subtracted value to the nearest integer value. Multiply by the sign of the scaled motion vector that has been converted to an absolute value after shifting to the right.

In summary, the scaling calculation unit 203 performs scaling calculation on the motion vector to be scaled by the following equation.
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 2 ^N−1 ) >> N} Expression (15)
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 2 ^N−1 ) >> N} (16)
abs (·): function that returns an absolute value sign (·): function that returns a sign (1 or −1) The reason why the absolute value is obtained by the equations (15) and (16) is that (Scale × mvcx) is positive or This is because even in the negative case, correction is made in the 0 direction by subtracting a. Expressions (15) and (16) represent that the motion vector after scaling is corrected by a predetermined amount a in the 0 vector direction. By correcting a by 0 in the 0 direction, the average value of the prediction vector candidates output from the scaling calculation unit 203 can be brought close to 0.

For example, when N in Expression (15) and Expression (16) is 8, Expression (15) and Expression (16) become Expression (17) and Expression (18), respectively.
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 128) >> 8} Expression (17)
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 128) >> 8} Expression (18)
The inventors conducted experiments on the predetermined amount a and found that the encoding efficiency is improved when the predetermined amount a is in the range of 1 ≦ a ≦ 2 ^N−2 . Therefore, for example, when N = 8, the predetermined amount a is any value of 1 ≦ a ≦ 64.

  Here, the value of the predetermined amount a may be a fixed value in the above range. In addition, an optimum value may be dynamically obtained for each scene or according to the scaling factor, and that value may be used. An example of dynamically changing the value according to the scaling factor will be described later.

  FIG. 9 is a block diagram illustrating an example of the configuration (part 1) of each unit of the scaling operation unit 203. In the example illustrated in FIG. 9, the scaling unit 301 calculates (Scale × mvcx) of Expression (15) and Expression (16).

  The correction unit 302 performs an absolute value calculation abs (Scale × mvcx) on (Scale × mvcx), and subtracts a predetermined amount a from this value. Further, the correction unit 302 calculates sign (Scale × mvcx) for (Scale × mvcx) to obtain a sign.

The adjustment unit 303 adds the predetermined amount 2 ^N-1 to the value obtained by subtracting the predetermined amount a, and shifts by N bits. The adjustment unit 303 multiplies the value shifted by N bits by a sign, and outputs the scaled motion vector (mvcx ′, mvcy ′) as a prediction vector candidate.

FIG. 10 is a block diagram illustrating an example of the configuration (part 2) of each unit of the scaling operation unit 203. In the example illustrated in FIG. 10, the correction unit 304 adds 2 ^N−1 −a to abs (Scale × mvcx).

  The adjustment unit 305 performs an N-bit shift operation on the value output from the correction unit 304 and multiplies the code. The processing of the scaling unit 301 is the same as the processing of the scaling unit 301 shown in FIG.

  FIG. 11 is a diagram for explaining the processing of the scaling operation unit 203 using a specific example. In the example shown in FIG. 11, it is assumed that a moving image in which an object whose input stream is stationary is copied is compressed. However, even a moving image in which a stationary object is captured, a small vector other than 0 may be selected as a motion vector due to noise or the like riding on each image.

  Here, it is assumed that the input stream is completely stationary and a zero vector is expected to be predicted. When the motion vector adjacent in the time direction is not 0, for example, when (mvcx, mvcy) = (2, 0) (the motion of (2/4 pixel, 0 pixel) in pixel units), Scale = 64 Consider an example of scaling by a factor of 1/4. At this time, since mvcx / 4 = 0.5, there are two possible ways to select mvcx ′ = 0 or mvcx ′ = 1 as a predicted vector candidate to be output.

  At this time, in the scaling calculation method according to the equations (12) and (13), mvcx ′ = 1 (predicted vector candidate 2 shown in FIG. 11). In the scaling operation unit 203 according to the formulas (15) and (16) of the embodiment, the scaling unit 301 first outputs 2 (mvcx) × 64 (Scale) = 128. The correction unit 302 sets 128−a + 128 = 256−a. In the range a described above, the adjustment unit 303 shifts by 8 bits and becomes mvcx ′ = 0 (predicted vector candidate 1 shown in FIG. 11).

  Therefore, (mvcx ′, mvcy ′) = (0, 0), and the scaling operation unit 203 can obtain an expected stationary prediction vector candidate.

  Here, when the frequency with mv being (0, 0) is N0 times and the frequency with mv being (1, 0) is N1 times, if N0> N1, the prediction vector is 0 vector. The number of times increases as compared with the prior art, the code amount of the difference vector decreases, and the coding efficiency improves.

Further, the scaling calculation method may be the following method. FIG. 12 is a block diagram illustrating an example of the configuration (No. 3) of each unit of the scaling calculation unit 203. In the example shown in FIG. 12, the predetermined amount a is calculated based on the size of the scaling coefficient. 2 ^N-1 -a is called an offset.

The correction unit 306 illustrated in FIG. 12 includes an offset calculation unit 361. The offset calculation unit 361 acquires the scaling coefficient from the scaling unit 301, calculates a predetermined amount a based on the magnitude of the scaling coefficient, and calculates an offset (2 ^N−1 −a). For example, the predetermined amount a is calculated by, for example, the following formula.
a = MIN (2 ^N−2 , abs (Scale) >> 3) (19)
MIN (x, y): a function that returns the smaller of x and y When the absolute value of the scaling coefficient Scale is large, the predetermined amount a increases, and therefore the degree of rounding in the 0 vector direction increases. Expression (19) is an expression representing that the value of a increases as the scaling coefficient increases, and the upper limit value of a is 2 ^N−2 .

  As described above, when the prediction vector candidate is scaled, the accuracy of the prediction vector can be improved by correcting in the 0 vector direction.

<Operation>
Next, the operation of the video decoding device 100 according to the first embodiment will be described. FIG. 13 is a flowchart illustrating an example of processing of the video decoding device in the first embodiment. The process shown in FIG. 13 is a decoding process in one processing unit block.

  In step S101, the entropy decoding unit 101 performs entropy decoding on the input stream data. The entropy decoding unit 101 decodes the L0 reference index and the difference vector, the prediction candidate index, the L1 reference index and the difference vector, the prediction candidate index, and the orthogonal transformation coefficient of the processing target block.

  In step S102, the prediction vector generation unit 104 calculates a list of prediction vector candidates of L0 and L1, using the decoded reference indexes of L0 and L1, motion vector information, and the like.

  In step S <b> 103, the motion vector restoration unit 105 acquires the L0 and L1 prediction candidate indexes and difference vector information decoded by the entropy decoding unit 101. The motion vector restoration unit 105 calculates each of the prediction vectors identified by the prediction candidate index from the prediction vector candidate list, L0 and L1. The motion vector restoration unit 105 restores the motion vectors of L0 and L1 by adding the prediction vector and the difference vector.

  In step S104, the motion vector restoration unit 105 stores the restored L0 and L1 reference indexes and the motion vector information in the motion vector information storage unit 103. These pieces of information are used in the subsequent block decoding process.

  In step S105, the prediction pixel generation unit 106 acquires the L0 motion vector and the L1 motion vector, acquires pixel data of an area referred to by the motion vector from the decoded image storage unit 110, and generates a prediction pixel signal.

  In step S106, the inverse quantization unit 107 acquires the orthogonal transform coefficient decoded by the entropy decoding unit 101, and performs an inverse quantization process.

  In step S107, the inverse orthogonal transform unit 108 performs inverse orthogonal transform processing on the inversely quantized signal. Determination of the inverse orthogonal transform process and a prediction error signal are generated.

  Note that the processes in steps S102 to S104 and the processes in steps S106 and S07 are performed in parallel and are not in any order.

  In step S108, the decoded pixel generation unit 109 adds the predicted pixel signal and the prediction error signal to generate a decoded pixel.

  In step S109, the decoded image storage unit 110 stores a decoded image including decoded pixels. Thus, the block decoding process ends, and the process proceeds to the next block decoding process.

(Predicted vector candidates for blocks adjacent in the spatial direction)
Next, the operation of the prediction vector generation unit 104 will be described. First, a process for generating a prediction vector candidate of a block adjacent to the processing target block in the spatial direction will be described. FIG. 14 is a flowchart illustrating an example of a process (part 1) performed by the prediction vector generation unit according to the first embodiment. In step S201 illustrated in FIG. 14, the vector information acquisition unit 202 sequentially acquires motion vector information of blocks adjacent to the processing target block in the spatial direction (upper adjacent block and left adjacent block). The acquisition of motion vector information is as described above.

  In step S202, the vector information acquisition unit 202 determines whether a motion vector (also referred to as a predetermined motion vector) that refers to the same reference picture as the reference picture indicated by refidx in the reference list LX has been selected. If a predetermined motion vector has been selected (step S202—YES), the process proceeds to step S205, and if a predetermined motion vector has not been selected (step S202—NO), the process proceeds to step S203.

  In step S203, the scaling coefficient calculation unit 201 calculates a scaling coefficient using the above-described equations (5) to (10).

  In step S204, the scaling calculation unit 203 performs a bit shift calculation by correcting a predetermined amount in the 0 vector direction for the scaled motion vector (predicted vector candidate).

  In step S205, the scaling calculator 203 outputs the corrected motion vector after scaling as a prediction vector candidate. Further, the scaling operation unit 203 outputs a predetermined motion vector as a prediction vector candidate without performing the scaling operation if it is a predetermined motion vector.

(Predicted vector candidates for blocks adjacent in the time direction)
Next, a process for generating a prediction vector candidate for a block adjacent to the processing target block in the time direction will be described. FIG. 15 is a flowchart illustrating an example of a process (part 2) performed by the prediction vector generation unit according to the first embodiment.

  In step S301 illustrated in FIG. 15, the vector information acquisition unit 202 acquires motion vector information of a block adjacent to the processing target block in the time direction. The acquisition of motion vector information is as described above.

  In step S302, the scaling coefficient calculation unit 201 calculates the scaling coefficient using the above-described equations (5) to (10).

  In step S303, the scaling coefficient calculator 201 determines whether the scaling coefficient Scale is not 1. If the scaling coefficient is not 1 (step S303—YES), the process proceeds to step S304. If the scaling coefficient is 1 (step S303—NO), the process proceeds to step S305.

  In step S304, the scaling calculation unit 203 performs a bit shift calculation by correcting a predetermined amount in the 0 vector direction for the scaled motion vector (predicted vector candidate).

  In step S305, the scaling calculation unit 203 outputs the corrected motion vector after scaling as a prediction vector candidate. Further, if the scaling coefficient is 1, the scaling operation unit 203 outputs the motion vector as a prediction vector candidate without performing the scaling operation.

  As described above, according to the first embodiment, the prediction vector candidate is corrected in the 0 vector direction based on the discovery of the inventors, thereby improving the accuracy of the prediction vector, reducing the code amount of the difference vector, and encoding efficiency. Can be improved.

[Example 2]
Next, a moving picture decoding apparatus according to the second embodiment will be described. In the second embodiment, the scaling operation is switched depending on whether the picture to which the motion vector of the prediction vector candidate belongs is adjacent to the processing target block in the spatial direction or in the temporal direction.

<Configuration>
Since the configuration of the moving picture decoding apparatus in the second embodiment is the same as that of the first embodiment except for the prediction vector generation unit, the prediction vector generation unit will be described below.

  FIG. 16 is a block diagram illustrating an example of the configuration of the prediction vector generation unit 400 according to the second embodiment. In the example illustrated in FIG. 16, the prediction vector generation unit 400 has a plurality of scaling calculation units and is configured to adaptively switch between these. In the configuration shown in FIG. 16, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted.

  The block determination unit 401 determines whether a block having a motion vector used for generating a prediction vector candidate is adjacent to the processing target block in the spatial direction or in the temporal direction. The block determination unit 401 switches to the scaling calculation unit A402 if adjacent to the time direction and to the scaling calculation unit B403 if adjacent to the spatial direction.

  The scaling calculation unit A402 performs processing similar to that of the scaling calculation unit 203 of the first embodiment (scaling calculation using Expression (15) and Expression (16)), and corrects the prediction vector candidate to be scaled.

  The scaling operation unit B403 scales the prediction vector candidate using Expression (12) and Expression (13). The motion vector generated by each scaling calculation unit is a prediction vector candidate.

  This is because, when the block to which the motion vector used to generate the prediction vector candidate belongs is adjacent to the processing target block in the spatial direction or in the temporal direction, the appearance frequency bias differs. It is possible. Information indicating whether adjacent in the time direction or adjacent in the spatial direction is referred to as adjacent information.

  For example, since the scaling is often performed when adjacent in the time direction, the encoding efficiency can be improved by performing the correction described in the first embodiment. On the other hand, when adjacent to each other in the spatial direction, scaling is rarely performed in the first place. Therefore, there is no problem even if scaling is performed using Expression (12) and Expression (13).

  In the second embodiment, an example in which a plurality of scaling calculators are used has been described. However, a single scaling calculator can also be used. For example, the block determination unit 401 notifies adjacent information to the scaling calculation unit.

  The scaling operation unit performs scaling using the equations (15) and (16) if the neighboring information indicates the time direction, and if the neighboring information indicates the spatial direction, the scaling calculation unit performs the operations of the equations (15) and (16). Scaling may be performed by omitting the process of subtracting the fixed amount a.

  Thus, the second embodiment can be implemented with one scaling calculation unit by switching whether or not the predetermined amount a is subtracted according to the determination result of the block determination unit 401.

<Operation>
Next, processing of the video decoding device in the second embodiment will be described. The decoding process is the same as the process shown in FIG.

  FIG. 17 is a flowchart illustrating an example of processing performed by the prediction vector generation unit according to the second embodiment.

  In step S401, the prediction vector generation unit 400 determines whether the block having the motion vector of the prediction vector candidate is adjacent to the processing target block in the spatial direction or in the temporal direction. This determination can be made by referring to the picture identifier. If it is adjacent in the time direction (step S401-YES), the process proceeds to step S402. If it is adjacent in the space direction (step S401-NO), the process proceeds to step S407.

  Steps S402 to S406 are the same as steps S301 to S305 described in FIG. The scaling calculation A is performed by the scaling calculation unit A402.

  Steps S407 to S409 and S411 are the same as steps S201 to S203 and S205 described in FIG.

  In step S410, the scaling calculation unit B403 performs the scaling calculation shown in Expression (12) and Expression (13).

  As described above, according to the second embodiment, it is possible to adaptively switch the scaling operation processing according to the adjacent information of the block to which the motion vector of the prediction vector candidate belongs, and to improve the accuracy of the prediction vector.

[Example 3]
Next, a moving picture decoding apparatus according to the third embodiment will be described. In the third embodiment, the scaling operation is switched based on the magnitude of the motion vector of the prediction vector candidate.

<Configuration>
Since the configuration of the moving picture decoding apparatus in the third embodiment is the same as that of the first embodiment except for the prediction vector generation unit, the prediction vector generation unit will be described below.

  FIG. 18 is a block diagram illustrating an example of a configuration of the prediction vector generation unit 500 according to the third embodiment. In the example illustrated in FIG. 18, the prediction vector generation unit 500 has a plurality of scaling operation units and is configured to adaptively switch between these. In the configuration shown in FIG. 18, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted.

  The motion vector determination unit 501 switches the scaling operation unit according to the magnitude of the input pre-scaling motion vector (predicted vector candidate). For example, the motion vector determination unit 501 determines whether the magnitude of the acquired motion vector is a predetermined value (for example, 16) or less.

  The motion vector determination unit 501 switches to the scaling calculation unit A502 if the magnitude of the motion vector is less than or equal to a predetermined value, and switches to the scaling calculation unit B503 if the magnitude of the motion vector is greater than the predetermined value.

  The scaling calculation unit A502 performs processing similar to that of the scaling calculation unit 203 of the first embodiment (scaling calculation using Expression (15) and Expression (16)) to correct the prediction vector candidate to be scaled.

  The scaling calculation unit B503 scales the prediction vector candidate using the equations (12) and (13). The motion vector generated by each scaling calculation unit becomes the final prediction vector candidate.

  This is because the bias in the appearance frequency of mv may vary depending on the size of the motion vector (predicted vector candidate) before scaling. The smaller the absolute value of the motion vector, the greater the influence of the error from the prediction vector. Therefore, when the absolute value of the motion vector is small, the correction processing described in the first embodiment is performed.

  In the third embodiment, an example in which a plurality of scaling calculation units are used has been described. However, a single scaling calculation unit can be used. For example, the motion vector determination unit 501 notifies the scaling calculation unit of the determination result as to whether or not the magnitude of the motion vector is a predetermined value or less.

  The scaling operation unit performs scaling using Expression (15) and Expression (16) if the determination result indicates that the determination result is equal to or less than the predetermined value, and if the determination result indicates that the determination result is greater than the predetermined value, the expression ( 15) Scaling may be performed by omitting the process of subtracting the predetermined amount a in Expression (16).

  Thus, the third embodiment can be implemented with one scaling calculation unit by switching whether or not the predetermined amount a is subtracted according to the determination result of the motion vector determination unit 501.

<Operation>
Next, processing of the video decoding device in the third embodiment will be described. The decoding process is the same as the process shown in FIG. The process by the prediction vector production | generation part in Example 3 is demonstrated.

(Predicted vector candidates for blocks adjacent in the spatial direction)
FIG. 19 is a flowchart illustrating an example of a process (part 1) performed by the prediction vector generation unit according to the third embodiment. The processes in steps S501 to S503 shown in FIG. 19 are the same as the processes in steps S201 to S203 shown in FIG.

  In step S504, the motion vector determination unit 501 determines whether the magnitude of the motion vector (predicted vector candidate) is equal to or less than a predetermined value. If the magnitude of the motion vector is equal to or smaller than the predetermined value (step S504—YES), the process proceeds to step S505. If the magnitude of the motion vector is greater than the predetermined value (step S504—NO), the process proceeds to step S506.

  In step S505, the scaling calculation unit A502 performs a scaling calculation that corrects by a predetermined amount a using Expressions (15) and (16).

  In step S506, the scaling calculation unit B503 performs scaling calculation using Expression (12) and Expression (13).

  In step S507, the prediction vector generation unit 500 outputs the motion vector calculated by each scaling calculation unit as a prediction vector candidate. Moreover, if it is a predetermined motion vector, the prediction vector production | generation part 500 will output a predetermined motion vector as a prediction vector candidate, without performing a scaling calculation.

(Predicted vector candidates for blocks adjacent in the time direction)
FIG. 20 is a flowchart illustrating an example of a process (part 2) performed by the prediction vector generation unit according to the third embodiment.

  Steps S601 to S603 shown in FIG. 20 are the same as the processes of steps S301 to S303 shown in FIG.

  In step S604, the motion vector determination unit 501 determines whether the magnitude of the motion vector (predicted vector candidate) is equal to or less than a predetermined value. If the magnitude of the motion vector is equal to or smaller than the predetermined value (step S604—YES), the process proceeds to step S605, and if the magnitude of the motion vector is greater than the predetermined value (step S604—NO), the process proceeds to step S606.

  In step S <b> 605, the scaling calculation unit A <b> 502 performs a scaling calculation that corrects by a predetermined amount a using Expression (15) and Expression (16).

  In step S606, the scaling calculation unit B503 performs scaling calculation using Expression (12) and Expression (13).

  In step S607, the prediction vector generation unit 500 outputs the motion vector calculated by each scaling calculation unit as a prediction vector candidate. Further, if the scaling coefficient is 1, the prediction vector generation unit 500 outputs the motion vector as a prediction vector candidate without performing the scaling operation.

  As described above, according to the third embodiment, it is possible to adaptively switch the scaling calculation processing according to the size of the motion vector of the prediction vector candidate and improve the accuracy of the prediction vector.

[Example 4]
Next, a moving picture decoding apparatus according to the fourth embodiment will be described. In the fourth embodiment, the scaling operation is switched based on the difference between the display time of the picture to which the motion vector (prediction vector candidate) before scaling belongs and the display time of the picture referenced by the motion vector.

<Configuration>
Since the configuration of the moving picture decoding apparatus according to the fourth embodiment is the same as that of the first embodiment except for the prediction vector generation unit, the prediction vector generation unit will be described below.

  FIG. 21 is a block diagram illustrating an example of a configuration of the prediction vector generation unit 600 according to the fourth embodiment. In the example illustrated in FIG. 21, the predicted vector generation unit 600 has a plurality of scaling calculation units and is configured to adaptively switch between these. In the configuration shown in FIG. 21, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted.

  The time difference determination unit 601 switches the scaling operation unit according to the difference between the display time of the picture to which the motion vector (prediction vector candidate) before scaling belongs and the display time of the picture referenced by the motion vector.

  For example, the time difference determination unit 601 acquires the reference picture identifier of the processing target block and the POC value of the processing target picture. Further, the time difference determination unit 601 acquires the identifier of the picture to which the motion vector acquired by the vector information acquisition unit 202 belongs and the identifier of the picture referred to by the motion vector.

  The time difference determination unit 601 acquires the POC value of each acquired picture from the reference picture list storage unit 102, and the POC value of the picture to which the motion vector (prediction vector candidate) before scaling belongs and the POC of the picture referenced by the motion vector The difference from the value is calculated. The time difference determination unit 601 determines whether or not the time difference (interval) calculated using this POC is a predetermined value (for example, 4) or less.

  If the time difference (interval) of the POC is less than or equal to a predetermined value, the time difference determination unit 601 switches to the scaling operation unit A602, and if the time difference of the POC is greater than the predetermined value, the time difference determination unit 601 switches to the scaling operation unit B603.

  The scaling calculation unit A602 performs the same processing as the scaling calculation unit 203 of the first embodiment (scaling calculation using Expression (15) and Expression (16)), and corrects the predicted vector candidate after scaling.

  The scaling operation unit B603 scales the prediction vector candidate using Expression (12) and Expression (13). The motion vector generated by each scaling calculation unit becomes the final prediction vector candidate.

  This is because the bias in the frequency of appearance of mv may differ depending on the difference between the display time of the picture to which the motion vector (predicted vector candidate) before scaling belongs and the display time of the picture referenced by the motion vector. It is.

  In the fourth embodiment, an example in which a plurality of scaling calculation units are used has been described. However, a single scaling calculation unit can be used. For example, the time difference determination unit 601 notifies the scaling calculation unit of the determination result as to whether or not the calculated time difference is equal to or less than a predetermined value.

  The scaling operation unit performs scaling using Expression (15) and Expression (16) if the determination result indicates that the determination result is equal to or less than the predetermined value, and if the determination result indicates that the determination result is greater than the predetermined value, the expression ( 15) Scaling may be performed by omitting the process of subtracting the predetermined amount a in Expression (16).

  Accordingly, the fourth embodiment can be implemented with one scaling calculation unit by switching whether or not the predetermined amount a is subtracted according to the determination result of the time difference determination unit 601.

<Operation>
Next, processing of the video decoding device in the fourth embodiment will be described. The decoding process is the same as the process shown in FIG. The process by the prediction vector production | generation part in Example 4 is demonstrated.

(Predicted vector candidates for blocks adjacent in the spatial direction)
FIG. 22 is a flowchart illustrating an example of a process (part 1) performed by the prediction vector generation unit according to the fourth embodiment. The processes in steps S701 to S703 shown in FIG. 22 are the same as the processes in steps S201 to S203 shown in FIG.

  In step S704, the time difference determination unit 601 determines whether or not the difference between the display time of the picture to which the motion vector (prediction vector candidate) before scaling belongs and the display time of the picture referred to by the motion vector is equal to or smaller than a predetermined value. judge. If the time difference is less than or equal to the predetermined value (step S704-YES), the process proceeds to step S705, and if the time difference is greater than the predetermined value (step S704-NO), the process proceeds to step S706.

  In step S <b> 705, the scaling calculation unit A <b> 602 performs a scaling calculation that corrects by a predetermined amount a using Expression (15) and Expression (16).

  In step S706, the scaling calculation unit B603 performs scaling calculation using Expression (12) and Expression (13).

  In step S707, the prediction vector generation unit 600 outputs the motion vector calculated by each scaling calculation unit as a prediction vector candidate. Moreover, if it is a predetermined motion vector, the prediction vector production | generation part 600 will output a predetermined motion vector as a prediction vector candidate, without performing a scaling calculation.

(Predicted vector candidates for blocks adjacent in the time direction)
FIG. 23 is a flowchart illustrating an example of the process (part 2) performed by the prediction vector generation unit according to the fourth embodiment.

  Steps S801 to S803 shown in FIG. 23 are the same as the processes of steps S301 to S303 shown in FIG.

  In step S804, the time difference determination unit 601 determines whether or not the difference between the display time of the picture to which the motion vector (prediction vector candidate) before scaling belongs and the display time of the picture referenced by the motion vector is equal to or smaller than a predetermined value. judge. If the time difference is less than or equal to the predetermined value (step S804-YES), the process proceeds to step S805, and if the time difference is greater than the predetermined value (step S804-NO), the process proceeds to step S806.

  In step S805, the scaling calculation unit A602 performs scaling calculation for performing correction by the predetermined amount a using Expressions (15) and (16).

  In step S806, the scaling calculation unit B603 performs scaling calculation using Expression (12) and Expression (13).

  In step S807, the prediction vector generation unit 600 outputs the motion vector calculated by each scaling calculation unit as a prediction vector candidate. Further, if the scaling coefficient is 1, the predicted vector generation unit 600 outputs the motion vector as a predicted vector candidate without performing the scaling operation.

  As described above, according to the fourth embodiment, the scaling operation processing is adaptively switched according to the difference between the display time of the picture to which the motion vector before scaling belongs and the display time of the picture referred to by the motion vector. Accuracy can be improved.

[Example 5]
Next, a moving picture encoding apparatus in Embodiment 5 will be described. A moving picture encoding apparatus according to the fifth embodiment is a moving picture encoding apparatus including the prediction vector generation unit according to any one of the first to fourth embodiments.

<Configuration>
FIG. 24 is a block diagram illustrating an example of a configuration of the moving image encoding apparatus 700 according to the fifth embodiment. A moving image encoding apparatus 700 illustrated in FIG. 24 includes a motion detection unit 701, a reference picture list storage unit 702, a decoded image storage unit 703, a motion vector information storage unit 704, a prediction vector generation unit 705, and a difference vector calculation unit 706. .

  In addition, the moving image encoding apparatus 700 includes a prediction pixel generation unit 707, a prediction error generation unit 708, an orthogonal transformation unit 709, a quantization unit 710, an inverse quantization unit 711, an inverse orthogonal transformation unit 712, a decoded pixel generation unit 713, An entropy encoding unit 714 is included.

  The motion detection unit 701 acquires the original image, acquires the storage position of the reference picture from the reference picture list storage unit 702, and acquires pixel data of the reference picture from the decoded image storage unit 703. The motion detection unit 701 detects the reference indexes and motion vectors of L0 and L1. The motion detection unit 701 outputs the region position information of the reference image referred to by the detected motion vector to the predicted pixel generation unit 707.

  The reference picture list storage unit 702 stores a reference picture storage location and picture information including POC information of a picture that can be referred to by the processing target block.

  The decoded image storage unit 703 stores a picture that has been encoded in the past and locally decoded in the moving image encoding device in order to be used as a reference picture for motion compensation.

  The motion vector information storage unit 704 stores motion vector information including the motion vector detected by the motion detection unit 701 and the reference index information of L0 and L1. The motion vector information storage unit 704 stores, for example, motion vector information including a motion vector of a block spatially and temporally adjacent to the processing target block and a reference picture identifier indicating a picture to which the motion vector refers.

  The prediction vector generation unit 705 generates a prediction vector candidate list using L0 and L1. The process of generating a prediction vector candidate is the same as the process described in the first to fourth embodiments.

  The difference vector calculation unit 706 acquires the L0 and L1 motion vectors from the motion detection unit 701, acquires the L0 and L1 prediction vector candidate lists from the prediction vector generation unit 705, and calculates the respective difference vectors.

  For example, the difference vector calculation unit 706 selects prediction vectors closest to the L0 and L1 motion vectors from the prediction vector candidate list, and determines the prediction vectors and prediction vector candidate indexes of L0 and L1, respectively.

  Further, the difference vector calculation unit 706 generates an L0 difference vector by subtracting the L0 prediction vector from the L0 motion vector, and generates an L1 difference vector by subtracting the L1 prediction vector from the L1 motion vector. .

  The prediction pixel generation unit 707 acquires a reference pixel from the decoded image storage unit 703 based on the input region position information of the reference image, and generates a prediction pixel signal.

  The prediction error generation unit 708 acquires an original image and a prediction pixel signal, and generates a prediction error signal by calculating a difference between the original image and the prediction pixel signal.

  The orthogonal transform unit 709 performs orthogonal transform processing such as discrete cosine transform on the prediction error signal, and outputs the orthogonal transform coefficient to the quantization unit 710. The quantization unit 710 quantizes the orthogonal transform coefficient.

  The inverse quantization unit 711 performs inverse quantization processing on the quantized orthogonal transform coefficient. The inverse orthogonal transform unit 712 performs an inverse orthogonal transform process on the inversely quantized coefficient.

  The decoded pixel generation unit 713 generates a decoded pixel by adding the prediction error signal and the prediction pixel signal. The decoded image including the generated decoded pixel is stored in the decoded image storage unit 703.

  The entropy encoding unit 714 performs entropy encoding on the reference index of L0 and L1, the difference vector, the prediction candidate index, and the quantized orthogonal transform coefficient acquired from the difference vector calculation unit 706 and the quantization unit 710. Do. The entropy encoding unit 714 outputs the data after entropy encoding as a stream.

<Operation>
Next, the operation of the moving picture coding apparatus 700 according to the fifth embodiment will be described. FIG. 25 is a flowchart illustrating an example of processing of the video encoding device. The encoding process shown in FIG. 25 shows the encoding process of one process unit block.

  In step S901, the motion vector detection unit 701 acquires an original image, acquires pixel data of a reference picture, and detects reference indexes and motion vectors of L0 and L1.

  In step S902, the prediction vector generation unit 705 calculates prediction vector candidate lists for L0 and L1, respectively. At this time, the prediction vector generation unit 705 corrects the prediction vector candidate by a predetermined amount a in the zero vector direction when scaling the prediction vector candidate, as in any of the first to fourth embodiments.

  In step S903, the difference vector calculation unit 706 selects a prediction vector closest to the L0 and L1 motion vectors from the prediction vector candidate list, and determines a prediction vector and a prediction candidate index for L0 and L1, respectively.

  Further, the difference vector calculation unit 706 generates an L0 difference vector by subtracting the L0 prediction vector from the L0 motion vector, and generates an L1 difference vector by subtracting the L1 prediction vector from the L1 motion vector. .

  In step S904, the predicted pixel generation unit 707 acquires a reference pixel from the decoded image storage unit 703 based on the input region position information of the reference image, and generates a predicted pixel signal.

  In step S905, the prediction error generation unit 708 receives the original image and the prediction pixel signal, and generates a prediction error signal by calculating a difference between the original image and the prediction pixel signal.

  In step S906, the orthogonal transform unit 709 performs orthogonal transform processing on the prediction error signal to generate orthogonal transform coefficients.

  In step S907, the quantization unit 710 performs a quantization process on the orthogonal transform coefficient, and generates an orthogonal transform coefficient after quantization.

  In step S908, the motion vector information storage unit 704 stores the motion vector information including the reference index information and the motion vector of L0 and L1 output from the motion vector detection unit 701. These pieces of information are used for encoding the next block.

  In addition, each process of step S902-S903, step S904-S907, and S908 is performed in parallel and is in no particular order.

  In step S909, the inverse quantization unit 711 performs an inverse quantization process on the quantized orthogonal transform coefficient to generate an orthogonal transform coefficient. Next, the inverse orthogonal transform unit 712 performs an inverse orthogonal transform process on the orthogonal transform coefficient to generate a prediction error signal.

  In step S910, the decoded pixel generation unit 713 adds the prediction error signal and the prediction pixel signal to generate a decoded pixel.

  In step S911, the decoded image storage unit 703 stores a decoded image including decoded pixels. This decoded image is used for subsequent block encoding processing.

  In step S912, the entropy encoding unit 714 entropy-encodes the reference index of L0 and L1, the difference vector, the prediction candidate index, and the quantized orthogonal transform coefficient, and outputs the result as a stream.

  As described above, according to the fifth embodiment, it is possible to provide a moving picture encoding apparatus that can improve the accuracy of a prediction vector and increase the encoding efficiency. Note that it is obvious for the prediction vector generation unit 705 of the moving picture encoding apparatus 700 that any one of the first to fourth embodiments can be used by those skilled in the art.

  In addition, the inventors' experiment has shown that the prediction vector candidate is calculated by the prediction vector generation unit described in each embodiment, thereby improving the coding efficiency by about 1% to 2% with respect to the test sequence of the standardization technique. Confirmed by

[Modification]
FIG. 26 is a block diagram illustrating an example of the configuration of the image processing apparatus 800. The image processing device 800 is an example of the moving image encoding device or the moving image decoding device described in the embodiments. As illustrated in FIG. 26, the image processing apparatus 800 includes a control unit 801, a main storage unit 802, an auxiliary storage unit 803, a drive device 804, a network I / F unit 806, an input unit 807, and a display unit 808. These components are connected to each other via a bus so as to be able to transmit and receive data.

  The control unit 801 is a CPU that controls each device, calculates data, and processes in a computer. The control unit 801 is an arithmetic device that executes a program stored in the main storage unit 802 or the auxiliary storage unit 803. The control unit 801 receives data from the input unit 807 or the storage device, calculates and processes the data, and then displays the display unit 808. Or output to a storage device.

  The main storage unit 802 is a ROM (Read Only Memory), a RAM (Random Access Memory), or the like, and a storage device that stores or temporarily stores programs and data such as an OS and application software that are basic software executed by the control unit 801. It is.

  The auxiliary storage unit 803 is an HDD (Hard Disk Drive) or the like, and is a storage device that stores data related to application software and the like.

  The drive device 804 reads the program from the recording medium 805, for example, a flexible disk, and installs it in the storage device.

  The recording medium 805 stores a predetermined program. The program stored in the recording medium 805 is installed in the image processing apparatus 800 via the drive apparatus 804. The installed predetermined program can be executed by the image processing apparatus 800.

  The network I / F unit 806 has a communication function connected via a network such as a LAN (Local Area Network) or a WAN (Wide Area Network) constructed by a data transmission path such as a wired and / or wireless line. This is an interface between the device and the image processing apparatus 800.

  The input unit 807 includes a keyboard having cursor keys, numeric input, various function keys, and the like, a mouse and a slice pad for performing key selection on the display screen of the display unit 808, and the like. An input unit 807 is a user interface for a user to give an operation instruction to the control unit 801 or input data.

  The display unit 808 has an LCD (Liquid Crystal Display) or the like, and performs display according to display data input from the control unit 801. Note that the display unit 808 may be provided outside, and in that case, the image processing apparatus 800 includes a display control unit.

  As described above, the moving image encoding process or the moving image decoding process described in the above-described embodiment may be realized as a program for causing a computer to execute. By installing this program from a server or the like and causing the computer to execute it, the above-described image encoding process or image decoding process can be realized.

  In addition, the moving image encoding program or the moving image decoding program is recorded on the recording medium 805, and the recording medium 805 on which the program is recorded is read by a computer or a portable terminal, so that the above-described moving image encoding process or moving image is performed. Decoding processing can also be realized.

  The recording medium 805 is a recording medium that records information optically, electrically, or magnetically, such as a CD-ROM, a flexible disk, or a magneto-optical disk, and information is electrically stored such as a ROM or flash memory. Various types of recording media such as a semiconductor memory for recording can be used. Further, the moving picture encoding process or the moving picture decoding process described in each of the above embodiments may be implemented in one or a plurality of integrated circuits.

  Although each embodiment has been described in detail above, it is not limited to a specific embodiment, and various modifications and changes can be made within the scope described in the claims. It is also possible to combine all or a plurality of the components of the above-described embodiments.

In addition, the following additional remarks are disclosed regarding the above Example.
(Appendix 1)
A video decoding device that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
A reference picture list storage unit that stores picture information of pictures that can be referred to by the processing target block;
A motion vector information storage unit that stores motion vector information including a motion vector of a block spatially and temporally adjacent to the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers;
A prediction vector generation unit that corrects the prediction vector candidate by a predetermined amount in the 0 direction when performing scaling using the picture information and the motion vector information on the prediction vector candidate for the motion vector of the processing target block;
A video decoding device comprising:
(Appendix 2)
The prediction vector generation unit
A scaling factor calculator that calculates a scaling factor using the picture information and the motion vector information;
A scaling calculator that calculates the predetermined amount based on the magnitude of the scaling coefficient and corrects the prediction vector candidate using the calculated predetermined amount;
The moving picture decoding apparatus according to Supplementary Note 1, comprising:
(Appendix 3)
The scaling factor has a predetermined precision as the precision after the decimal point,
The scaling calculator is
A scaling unit that multiplies the scaling vector by a prediction vector candidate and expands it to a precision below a predetermined decimal point to obtain a scaled prediction vector candidate;
A correction unit that corrects the prediction vector candidate scaled by the scaling unit by a predetermined amount in the 0 direction;
An adjustment unit that converts the predicted vector candidate scaled by a scaling coefficient having a precision below a predetermined decimal point in the scaling unit and corrected by the correction unit, to the nearest integer;
The moving picture decoding apparatus according to Supplementary Note 1 or 2, further comprising:
(Appendix 4)
4. The moving picture decoding apparatus according to appendix 3, wherein when the precision below the predetermined decimal point is N bits, the adjustment unit shifts right by N bits after adding 2 ^N−1 .
(Appendix 5)
The prediction vector generation unit
The moving picture decoding apparatus according to any one of supplementary notes 1 to 4, wherein when the prediction vector candidate belongs to a block adjacent in the time direction, correction processing of the prediction vector candidate is performed.
(Appendix 6)
The prediction vector generation unit
The moving picture decoding apparatus according to any one of supplementary notes 1 to 4, wherein when the size of the prediction vector candidate is equal to or smaller than a first predetermined value, correction processing for the prediction vector candidate is performed.
(Appendix 7)
The prediction vector generation unit
The moving picture decoding apparatus according to any one of Supplementary notes 1 to 4, wherein when the interval between a picture to which a prediction vector belongs and a picture referred to by the prediction vector is equal to or smaller than a second predetermined value, correction processing for a prediction vector candidate is performed. .
(Appendix 8)
The moving image decoding according to any one of appendices 1 to 7, wherein when the precision after the decimal point of the scaling coefficient is N bits, the predetermined value to be corrected by the correction unit is 1 or more and 2 ^N−2 or less. apparatus.
(Appendix 9)
A video encoding device that performs an encoding process using a motion vector of a processing target block and a prediction vector candidate of the motion vector for each processing target block into which an input image is divided,
A reference picture list storage unit that stores picture information of pictures that can be referred to by the processing target block;
A motion vector information storage unit for storing motion vector information including a motion vector of a block adjacent to the processing target block spatially and temporally, a reference picture identifier indicating a picture to which the motion vector refers;
A prediction vector generation unit that corrects the prediction vector candidate by a predetermined amount in the 0 direction when performing scaling using the picture information and the motion vector information on the prediction vector candidate for the motion vector of the processing target block;
A video encoding device comprising:
(Appendix 10)
A video decoding method executed by a video decoding device that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
Motion including picture information of a picture that can be referred to by the processing target block, a motion vector of a block spatially and temporally adjacent to the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers When performing scaling on a prediction vector candidate for the motion vector of the processing target block using vector information, the prediction vector candidate is corrected by a predetermined amount in the 0 direction and scaled,
A moving picture decoding method for calculating a difference from a motion vector of the processing target block using a scaled prediction vector candidate.
(Appendix 11)
A moving picture coding method executed by a moving picture coding apparatus that performs coding processing using a motion vector of a processing target block and a prediction vector candidate of the motion vector for each processing target block into which an input image is divided There,
Motion including picture information of a picture that can be referred to by the processing target block, a motion vector of a block spatially and temporally adjacent to the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers When performing scaling on a prediction vector candidate for the motion vector of the processing target block using vector information, the prediction vector candidate is corrected by a predetermined amount in the 0 direction and scaled,
A moving picture coding method for calculating a difference from a motion vector of the processing target block using a scaled prediction vector candidate.
(Appendix 12)
A moving picture decoding program to be executed by a moving picture decoding apparatus that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
Motion including picture information of a picture that can be referred to by the processing target block, a motion vector of a block spatially and temporally adjacent to the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers When performing scaling on a prediction vector candidate for the motion vector of the processing target block using vector information, the prediction vector candidate is corrected by a predetermined amount in the 0 direction and scaled,
Calculating a difference from the motion vector of the processing target block using the scaled prediction vector candidate;
A moving picture decoding program for causing a moving picture decoding apparatus to execute processing.
(Appendix 13)
A moving picture coding program to be executed by a moving picture coding apparatus that performs coding processing using a motion vector of a processing target block and a prediction vector candidate of the motion vector for each processing target block into which an input image is divided There,
Motion including picture information of a picture that can be referred to by the processing target block, a motion vector of a block spatially and temporally adjacent to the processing target block, and a reference picture identifier indicating a picture to which the motion vector refers When performing scaling on a prediction vector candidate for the motion vector of the processing target block using vector information, the prediction vector candidate is corrected by a predetermined amount in the 0 direction and scaled,
Calculating a difference from the motion vector of the processing target block using the scaled prediction vector candidate;
A moving picture coding program that causes a moving picture coding apparatus to execute processing.

100 moving picture decoding apparatus 101 entropy decoding unit 102 reference picture list storage unit 103 motion vector information storage unit 104 prediction vector generation unit 105 motion vector restoration unit 106 prediction pixel generation unit 107 inverse quantization unit 108 inverse orthogonal transform unit 109 decoded pixel generation Unit 110 decoded image storage unit 201 scaling coefficient calculation unit 202 vector information acquisition unit 203 scaling operation unit 301 scaling unit 302, 304, 306 correction unit 303, 305 adjustment unit 361 offset calculation unit 401 block determination unit 402, 502, 602 scaling operation Part A
403, 503, 603 Scaling operation part B
501 Motion vector determination unit 601 Time difference determination unit 700 Video encoding device 701 Motion detection unit 702 Reference picture list storage unit 703 Decoded image storage unit 704 Motion vector information storage unit 705 Prediction vector generation unit 706 Difference vector calculation unit 707 Prediction pixel generation Unit 708 prediction error generation unit 709 orthogonal transform unit 710 quantization unit 711 inverse quantization unit 712 inverse orthogonal transform unit 713 decoded pixel generation unit 714 entropy coding unit

Claims (3)

A video decoding method executed by a video decoding device that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
Reference is made to the motion vector mvc having picture information of a picture that can be referred to by the processing target block, a horizontal component mvcx and a vertical component mvcy of a block spatially or temporally adjacent to the processing target block, and the motion vector mvc A prediction vector candidate mvc ′ for the motion vector mv of the block to be processed is calculated by scaling the motion vector mvc of the adjacent block using motion vector information including a reference picture identifier indicating a picture to be The Scale coefficient has N bits as the precision after the decimal point, and when N = 8 bits,
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 128) >> 8}
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 128) >> 8}
abs (·): function that returns an absolute value sign (·): a function that returns a sign (1 or −1), and a predetermined amount a is set in the 0 direction within a range of 1 ≦ a ≦ 2 ^N−2. To scale to
A moving picture decoding method for calculating a difference from a motion vector of the processing target block using a scaled prediction vector candidate mvc ′ = (mvcx ′, mvcy ′). A video decoding device that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
A reference picture list storage unit that stores picture information of pictures that can be referred to by the processing target block;
Motion vector information including a motion vector mvc having a horizontal component mvcx and a vertical component mvcy of a block spatially or temporally adjacent to the processing target block, and a reference picture identifier indicating a picture referred to by the motion vector mvc. A motion vector information storage unit for storing;
Using the picture information and the motion vector information, a predicted vector candidate mvc ′ for the motion vector mv of the processing target block is calculated by scaling the motion vector mvc of the adjacent block. Has N bits as precision and N = 8 bits,
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 128) >> 8}
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 128) >> 8}
abs (·): function that returns an absolute value sign (·): a function that returns a sign (1 or −1), and a predetermined amount a is set in the 0 direction within a range of 1 ≦ a ≦ 2 ^N−2. A predictive vector generation unit that corrects and scales to
A video decoding device comprising: A moving picture decoding program to be executed by a moving picture decoding apparatus that performs a decoding process using a motion vector of the processing target block and a prediction vector candidate of the motion vector for each processing target block,
Reference is made to the motion vector mvc having picture information of a picture that can be referred to by the processing target block, a horizontal component mvcx and a vertical component mvcy of a block spatially or temporally adjacent to the processing target block, and the motion vector mvc A prediction vector candidate mvc ′ for the motion vector mv of the block to be processed is calculated by scaling the motion vector mvc of the adjacent block using motion vector information including a reference picture identifier indicating a picture to be The Scale coefficient has N bits as the precision after the decimal point, and when N = 8 bits,
mvcx ′ = sign (Scale × mvcx) × {(abs (Scale × mvcx) −a + 128) >> 8}
mvcy ′ = sign (Scale × mvcy) × {(abs (Scale × mvcy) −a + 128) >> 8}
abs (·): function that returns an absolute value sign (·): a function that returns a sign (1 or −1), and a predetermined amount a is set in the 0 direction within a range of 1 ≦ a ≦ 2 ^N−2. To scale to
Using the scaled prediction vector candidate mvc ′ = (mvcx ′, mvcy ′) to calculate a difference from the motion vector of the processing target block;
A moving picture decoding program for causing a moving picture decoding apparatus to execute processing.

Images