WO2022219888A1 - Image-decoding device, image-decoding method, and program - Google Patents

Abstract

The image-decoding device 200 according to the present invention comprises: an OBMC application possibility determination unit 241B2 configured so as to determine the possibility of applying OBMC prediction to each subject block; an OBMC processing unit 241B3 configured so as to process OBMC prediction for each subject block; and a decoding unit 210 that is configured so as to determine whether to decode an OBMC flag configured so as to control the possibility of applying OBMC prediction to each subject block, and so as to decode the OBMC flag.

Description

Image decoding device, image decoding method and program

The present invention relates to an image decoding device, an image decoding method and a program.

Non-Patent Document 1 and Non-Patent Document 2 disclose a technique called motion compensation (MC: Motion Compensation) prediction. Furthermore, Non-Patent Document 2 discloses a technique called Overlapped Block Motion Compensation (OBMC) prediction.

The interpolation filter length (the number of taps) for the luminance signal and the color difference signal used in MC in Non-Patent Document 1 is a maximum of 8 taps and 4 taps, respectively, whereas MC in Non-Patent Document 2 uses The interpolation filter lengths for luma and chrominance signals are up to 12 taps and 4 taps, respectively.

Also, the interpolation filter lengths for luminance and color difference signals used in OBMC in Non-Patent Document 2 are also the same maximum 12 taps and 4 taps.

Furthermore, in Non-Patent Document 2, there are MCs for each of the small size blocks of uni-prediction and bi-prediction that are not (restricted) in Non-Patent Document 1 (not restricted). On the other hand, in Non-Patent Document 2, the application of OBMC to such small size blocks is limited.

However, the technology disclosed in Non-Patent Document 2 has the following three extensions to the technology disclosed in Non-Patent Document 1, so that the encoding performance is improved. There is a problem that the reference pixel area (memory band) required for generation increases.
1. Addition of OBMC 2. Extension of interpolation filter length for MC and OBMC 3. Lifting of restrictions on MC for small size blocks in uni-prediction and bi-prediction Therefore, the present invention has been made in view of the above problems. It is an object of the present invention to provide an image decoding device, an image decoding method, and a program that can maintain the effect of improving coding performance while suppressing an increase in memory bandwidth.

A first feature of the present invention is an image decoding device to which OBMC prediction can be applied, the OBMC applicability determination unit configured to determine applicability of the OBMC prediction for each target block; Determining whether or not an OBMC processing unit configured to process the OBMC prediction for each target block and an OBMC flag configured to control applicability of the OBMC prediction for each target block are decoded. and a decoding unit configured to decode the OBMC flag.

A second feature of the present invention is an image decoding method to which OBMC prediction is applicable, comprising a step of determining applicability of the OBMC prediction for each target block, and processing the OBMC prediction for each target block. and determining whether or not an OBMC flag configured to control applicability of the OBMC prediction for each target block is decoded, and decoding the OBMC flag.

A third feature of the present invention is a program that causes a computer to function as an image decoding device to which OBMC prediction is applicable, wherein the image decoding device determines whether or not the OBMC prediction is applicable for each target block. an OBMC processing unit configured to process the OBMC prediction for each target block; and the applicability of the OBMC prediction for each target block. and a decoding unit configured to determine whether or not the OBMC flag configured as above is decoded, and decode the OBMC flag.

According to the present invention, it is possible to provide an image decoding device, an image decoding method, and a program that can maintain the effect of improving coding performance while suppressing an increase in memory bandwidth.

FIG. 1 is a diagram showing an example of the configuration of an image processing system 1 according to one embodiment. FIG. 2 is a diagram showing an example of functional blocks of the image encoding device 100 according to one embodiment. FIG. 3 is a diagram showing an example of functional blocks of the image decoding device 200 according to one embodiment. FIG. 4 is a diagram illustrating an example of functional blocks of the inter prediction unit 241 of the image decoding device 200 according to one embodiment. FIG. 5 is a diagram showing an example of functional blocks of the motion-compensated prediction signal generation unit 241B of the inter prediction unit 241 of the image decoding device 200 according to one embodiment. FIG. 6 is a diagram showing an example of a mechanism for generating a normal MC prediction signal and an OBMC prediction signal from a target MV and adjacent MVs. FIG. 7 is a diagram showing an example of a memory band expansion element common to normal MC and OBMC. FIG. 8 is a diagram showing an example of a memory band expansion element common to normal MC and OBMC. FIG. 9 is a diagram showing an example of a memory band expansion element common to normal MC and OBMC. FIG. 10A is a diagram showing an example of a memory bandwidth expansion element in OBMC. FIG. 10B is a diagram illustrating an example of a memory bandwidth expansion element in OBMC. FIG. 11 is an example of a comparison table of the interpolation filter types for normal MC and the interpolation filter types for OBMC according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment, respectively. FIG. 12 is a diagram showing an example of frequency response characteristics of the 8-tap and 12-tap interpolation filters disclosed in Non-Patent Document 1 and Non-Patent Document 2. In FIG. FIG. 13 is an example of a comparison table of minimum sizes of uni-prediction or bi-prediction target blocks or target sub-blocks to which normal MC can be applied according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment. is. FIG. 14 is an example of a comparison table of applicable block sizes of OBMC according to Non-Patent Document 2, the prior application, and the present embodiment. FIG. 15 is an example of a comparison table of the number of blending lines of OBMC according to Non-Patent Document 2, the prior application, and the present embodiment. FIG. 16 is a flowchart showing an example of processing for determining whether to decode the OBMC flag. FIG. 17 is an example of a comparison table of the OBMC flag decoding condition 1, the OBMC flag decoding condition 2, and the determination method of OBMC applicability according to Non-Patent Document 2, the prior application, and the present embodiment, respectively. FIG. 18 is an example of a comparison table of the OBMC flag decoding condition 1, the OBMC flag decoding condition 2, and the determination method of OBMC applicability according to Non-Patent Document 2, the prior application, and the present embodiment, respectively, in a modified example. is. FIG. 19 is a diagram for explaining an example of processing for determining transmission accuracy that can be controlled by AMVR in a modified example. FIG. 20 is a diagram for explaining an example of processing for determining MVD added by MMVD in a modified example.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. It should be noted that constituent elements in the following embodiments can be appropriately replaced with existing constituent elements and the like, and various variations including combinations with other existing constituent elements are possible. Therefore, the following description of the embodiments is not intended to limit the scope of the invention described in the claims.

<First Embodiment>
An image processing system 10 according to a first embodiment of the present invention will be described below with reference to FIGS. 1 to 20. FIG. FIG. 1 is a diagram showing an image processing system 10 according to this embodiment.

(Image processing system 10)
As shown in FIG. 1, an image processing system 10 according to this embodiment has an image encoding device 100 and an image decoding device 200 .

The image encoding device 100 is configured to generate encoded data by encoding an input image signal (picture). The image decoding device 200 is configured to generate an output image signal by decoding encoded data.

Here, such encoded data may be transmitted from the image encoding device 100 to the image decoding device 200 via a transmission path. Also, the encoded data may be stored in a storage medium and then provided from the image encoding device 100 to the image decoding device 200 .

(Image encoding device 100)
The image coding apparatus 100 according to this embodiment will be described below with reference to FIG. FIG. 2 is a diagram showing an example of functional blocks of the image encoding device 100 according to this embodiment.

As shown in FIG. 2 , the image coding apparatus 100 includes an inter prediction unit 111, an intra prediction unit 112, a subtractor 121, an adder 122, a transform/quantization unit 131, an inverse transform/inverse quantization It has a section 132 , an encoding section 140 , an in-loop filtering section 150 and a frame buffer 160 .

The inter prediction unit 111 is configured to generate a prediction signal by inter prediction (inter-frame prediction).

Specifically, the inter prediction unit 111 identifies a reference block included in the reference frame by comparing a frame to be encoded (target frame) with a reference frame stored in the frame buffer 160, and identifies a reference block included in the reference frame. It is configured to determine a motion vector (MV) for the block.

Also, the inter prediction unit 111 is configured to generate a prediction signal included in the encoding target block (hereinafter referred to as target block) for each target block based on the reference block and the motion vector. The inter prediction section 111 is configured to output the prediction signal to the subtractor 121 and the adder 122 . Here, the reference frame is a frame different from the target frame.

The intra prediction unit 112 is configured to generate a prediction signal by intra prediction (intra-frame prediction).

Specifically, the intra prediction unit 112 is configured to identify reference blocks included in the target frame and generate a prediction signal for each target block based on the identified reference blocks. Also, the intra prediction unit 112 is configured to output the prediction signal to the subtractor 121 and the adder 122 .

Here, the reference block is a block that is referenced for the target block. For example, the reference block is a block adjacent to the target block.

The subtractor 121 is configured to subtract the prediction signal from the input image signal and output the prediction residual signal to the transformation/quantization section 131 . Here, the subtractor 121 is configured to generate a prediction residual signal that is a difference between a prediction signal generated by intra prediction or inter prediction and an input image signal.

The adder 122 adds the prediction signal to the prediction residual signal output from the inverse transform/inverse quantization unit 132 to generate a pre-filtering decoded signal. It is configured to output to the loop filter processing unit 150 .

Here, the unfiltered decoded signal constitutes a reference block used by intra prediction section 112 .

The transform/quantization unit 131 is configured to perform transform processing on the prediction residual signal and acquire the coefficient level value. Further, the transform/quantization unit 131 may be configured to quantize the coefficient level values.

Here, the transform processing is processing for transforming the prediction residual signal into a frequency component signal. As such transformation processing, a base pattern (transformation matrix) corresponding to a discrete cosine transform (hereinafter referred to as DCT) may be used, and a discrete sine transform (hereinafter referred to as DST) may be used. A base pattern (transformation matrix) corresponding to may be used.

The inverse transform/inverse quantization unit 132 is configured to perform inverse transform processing on the coefficient level values output from the transform/quantization unit 131 . Here, the inverse transform/inverse quantization unit 132 may be configured to perform inverse quantization of the coefficient level values prior to the inverse transform processing.

Here, the inverse transform processing and inverse quantization are performed in a procedure opposite to the transform processing and quantization performed by the transform/quantization unit 131 .

The encoding unit 140 is configured to encode the coefficient level values output from the transform/quantization unit 131 and output encoded data.

Here, for example, the encoding is entropy encoding that assigns codes of different lengths based on the probability of occurrence of coefficient level values.

Also, the encoding unit 140 is configured to encode control data used in the decoding process in addition to the coefficient level values.

Here, the control data may include size data such as encoding block size, prediction block size, transform block size.

The in-loop filtering unit 150 is configured to perform filtering on the pre-filtering decoded signal output from the adder 122 and to output the post-filtering decoded signal to the frame buffer 160 .

Here, for example, the filter processing includes deblocking filter processing for reducing distortion occurring at the boundary portion of blocks (encoding blocks, prediction blocks or transform blocks), filter coefficients and filter selection information transmitted from the image encoding device 100. , adaptive loop filtering that switches the filter based on the local characteristics of the pattern of the image.

The frame buffer 160 is configured to accumulate reference frames used by the inter prediction section 111 .

Here, the decoded signal after filtering constitutes a reference frame used in inter prediction section 111 .

(Image decoding device 200)
The image decoding device 200 according to this embodiment will be described below with reference to FIG. FIG. 3 is a diagram showing an example of functional blocks of the image decoding device 200 according to this embodiment.

As shown in FIG. 3, the image decoding device 200 includes a decoding unit 210, an inverse transform/inverse quantization unit 220, an adder 230, an inter prediction unit 241, an intra prediction unit 242, and an in-loop filtering unit. 250 and a frame buffer 260 .

The decoding unit 210 is configured to decode the encoded data generated by the image encoding device 100 and decode the coefficient level values.

Here, the decoding is, for example, entropy decoding in a procedure opposite to the entropy encoding performed by the encoding unit 140.

Further, the decoding unit 210 may be configured to acquire the control data by decoding the encoded data. Note that, as described above, the control data may include size data, header information, and the like.

The inverse transform/inverse quantization unit 220 is configured to perform inverse transform processing on the coefficient level values output from the decoding unit 210 . Here, the inverse transform/inverse quantization unit 220 may be configured to perform inverse quantization of the coefficient level values prior to the inverse transform processing.

Here, the inverse transform processing and inverse quantization are performed in a procedure opposite to the transform processing and quantization performed by the transform/quantization unit 131 .

The adder 230 adds the prediction signal to the prediction residual signal output from the inverse transform/inverse quantization unit 220 to generate a pre-filtering decoded signal. It is configured to output to the filter processing unit 250 .

Here, the unfiltered decoded signal constitutes a reference block used by the intra prediction unit 242.

The inter prediction section 241, like the inter prediction section 111, is configured to generate a prediction signal by inter prediction (inter-frame prediction).

Specifically, the inter prediction unit 241 is configured to generate a prediction signal based on a motion vector decoded from encoded data and a reference signal included in a reference frame. The inter prediction section 241 is configured to output a prediction signal to the adder 230 .

The intra prediction unit 242, like the intra prediction unit 112, is configured to generate a prediction signal by intra prediction (intra-frame prediction).

Specifically, the intra prediction unit 242 is configured to identify reference blocks included in the target frame and generate a prediction signal for each prediction block based on the identified reference blocks. The intra prediction section 242 is configured to output the prediction signal to the adder 230 .

Similar to in-loop filtering section 150 , in-loop filtering section 250 performs filtering on the unfiltered decoded signal output from adder 230 and outputs the filtered decoded signal to frame buffer 260 . is configured to

Here, for example, the filter processing includes deblocking filter processing for reducing distortion occurring at boundaries of blocks (encoding blocks, prediction blocks, transform blocks, or sub-blocks obtained by dividing them), transmission from the image encoding device 100 This is adaptive loop filtering that switches filters based on the filter coefficients, filter selection information, and local characteristics of the pattern of the image.

The frame buffer 260, like the frame buffer 160, is configured to accumulate reference frames used in the inter prediction section 241.

Here, the decoded signal after filtering constitutes a reference frame used by the inter prediction unit 241 .

(Inter prediction unit 241)
The inter prediction unit 241 according to this embodiment will be described below with reference to FIG. FIG. 4 is a diagram showing an example of functional blocks of the inter prediction unit 241 according to this embodiment.

As shown in FIG. 4, the inter prediction unit 241 has a motion vector (MV) derivation unit 241A and a motion compensation (MC) prediction signal generation unit 241B.

The motion vector (MV) derivation unit 241A acquires the motion vector of the target block by decoding the target frame and the reference frame input from the frame buffer 260 and the control data received from the image encoding device 100, It is configured to output a motion vector to a motion compensation (MC) prediction signal generator 241B.

Here, as a method of deriving the motion vector of the target block, the techniques of AMVP (Adaptive Motion Vector Prediction) mode and merge mode disclosed in Non-Patent Document 1 are used.

Furthermore, AMVP mode and merge mode include Affine mode and SbTMVP (Sub-block Temporal Motion Vector) in which a target block is divided into sub-block units and a motion vector is derived for each sub-block (hereinafter referred to as target sub-block). Prediction) mode.

Since the technique disclosed in Non-Patent Document 1 or Non-Patent Document 2 can be applied to the present embodiment as it is, the mode for deriving the motion vector of the target block will be omitted.

The motion compensation (MC) prediction signal generation unit 241B generates prediction signals based on the motion vector (hereinafter referred to as target MV) of the target block or target sub-block input from the motion vector (MV) derivation unit 241A. is configured as The details of the method of generating the prediction signal from the motion vector will be described later.

(Motion Compensation (MC) Prediction Signal Generation Unit 241B)
The motion compensation (MC) prediction signal generator 241B according to this embodiment will be described below with reference to FIG. FIG. 5 is a diagram showing an example of functional blocks of the motion compensation (MC) prediction unit 241B according to this embodiment.

As shown in FIG. 5, the motion compensation (MC) prediction unit 241B has a normal MC processing unit 241B1, an OBMC applicability determination unit 241B2, and an OBMC processing unit 241B3.

The normal MC processing unit 241B1 is configured to generate a normal MC prediction signal for the target block or target sub-block based on the target MV.

Here, if the target MV indicates a sub-pixel precision position, an interpolation filter is used to generate a normal MC prediction signal for the target block or target sub-block.

When an interpolation filter is used, the type of interpolation filter is selected based on the control data decoded by the decoding unit 210 and the mode information when the motion vector is derived by the motion vector (MV) deriving unit 241A. be. Here, the details of the interpolation filter type for normal MC will be described later.

The OBMC applicability determination unit 241B2 is configured to determine the applicability of OBMC to the target block or target sub-block.

Here, the process of determining whether or not OBMC is applicable may be performed in units of 4×4 pixel blocks facing block boundaries or sub-block boundaries, as in Non-Patent Document 2.

When it is determined that OBMC is applicable, the OBMC applicability determining unit 241B2 is configured to notify the subsequent OBMC processing 241B3 of that effect.

On the other hand, when it is determined that OBMC is not applicable, the OBMC applicability determination unit 241B2 is configured to output the normal MC prediction signal generated by the normal MC processing unit 241B1 as an inter prediction signal.

Here, the details of the method for determining whether or not OBMC is applicable will be described later.

The OBMC processing unit 241B3 calculates motion vectors (hereinafter, adjacent MV ) and a reference frame to generate a prediction signal (hereinafter referred to as an OBMC prediction signal) for an arbitrary number of lines (hereinafter referred to as the number of blending lines) from the block boundary or sub-block boundary for the target block or target sub-block. is configured as

Here, as in Non-Patent Document 2, the OBMC prediction signal generation process may be performed in units of 4×4 pixel blocks facing block boundaries or sub-block boundaries.

The OBMC processing unit 241B3 is configured to blend the OBMC prediction signal generated in this way with the normal MC prediction signal of the target block or target sub-block, and output as an inter prediction signal.

Details of the method of generating the OBMC prediction signal will be described later.

(Enlargement of reference pixel area (memory band) by normal MC and OBMC)
6 to 9, enlargement of the reference pixel region by normal MC and OBMC according to Non-Patent Document 1, Non-Patent Document 2, Japanese Patent Application No. 2018-246858 (hereinafter referred to as the prior application), and the present embodiment, respectively, with reference to FIGS. will be explained. Note that the content similar to that of the earlier application was announced at IEEE ICIP2020.

FIG. 6 is a diagram showing an example of a mechanism for generating a normal MC prediction signal and an OBMC prediction signal from a target MV and adjacent MVs.

As shown in FIG. 6, normal MC obtains reference pixels for generating a normal MC prediction signal from a reference frame using the target MV. At this time, when the target MV refers to sub-pixel precision positions, an interpolation filter is used to generate a normal MC prediction signal, so the reference pixel area is expanded with respect to the area of the reference block corresponding to the target block.

On the other hand, OBMC uses adjacent MVs to obtain reference pixels for OBMC prediction signals from reference frames. At this time, the area for obtaining reference pixels is the number of blending lines across the boundary of the target block or target sub-block.

Furthermore, when the adjacent MV refers to the sub-pixel precision position, the reference pixel region is Expanding.

In this embodiment, the reference pixel area obtained from the reference frame to generate the normal MC prediction signal and OBMC prediction signal is defined as the memory band.

7 to 9 are diagrams showing examples of memory band expansion elements common to normal MC and OBMC.

FIG. 7 shows an example in which the longer the interpolation filter length (the number of taps), the larger the reference pixel area. Specifically, in the example of FIG. 7, the reference pixel areas are compared when the interpolation filter length is 2 taps, 4 taps, and 8 taps.

FIG. 8 shows an example in which the reference pixel area expands as the number of target MVs or adjacent MVs increases. Specifically, in the example of FIG. 8, the reference pixel regions for prediction with one motion vector (uni-prediction) and prediction with two motion vectors (bi-prediction) are compared.

FIG. 9 shows an example in which the smaller the target block or target sub-block is, the larger the reference pixel area is. Specifically, in the example of FIG. 9, reference pixel areas of a 16×16 pixel block and a 4×4 pixel block are compared.

FIG. 10 is a diagram showing an example of memory bandwidth expansion elements in OBMC.

As shown in FIG. 10A, the reference pixel area expands as the number of OBMC application locations increases with respect to block boundaries or sub-block boundaries having a total of four surfaces. Also, as shown in FIG. 10B, the larger the number of OBMC blending lines, the larger the reference pixel area.

(Interpolation filter type for normal MC)
Hereinafter, interpolation filter types for normal MC according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment will be described with reference to FIGS. 11 and 12 .

FIG. 11 is an example of a comparison table of normal MC interpolation filter types and OBMC interpolation filter types according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment. Note that the types of interpolation filters for OBMC will be described later.

FIG. 12 is a diagram showing an example of frequency response characteristics of the 8-tap and 12-tap interpolation filters disclosed in Non-Patent Document 1 and Non-Patent Document 2. FIG.

As shown in FIG. 11, the interpolation filter type for normal MC in Non-Patent Document 1 branches depending on which of the following three conditions the target block or target sub-block meets.
A1. In the case of a luminance signal and a mode other than Affine A2. In case of luminance signal and Affine mode A3. Color Difference Signal If the target block or target sub-block is a luminance signal and is in a mode other than Affine (case A1. above), the following two conditions are applied.

First, if the target block's hpelIfIdx is valid (hpelIfIdx=1) and the target MV references half-pixel precision locations, a 6-tap Gaussian filter is selected. Here, hpelIfIdx is an internal parameter indicating whether or not the SIF (Switchable Interpolation Filter) disclosed in Non-Patent Document 1 is applicable.

Second, in cases other than the above, the 8-tap interpolation filter disclosed in Non-Patent Document 1 is selected.

When the target block or target sub-block is a luminance signal and is in Affine mode (case A2. above), the 6-tap interpolation filter disclosed in Non-Patent Document 1 is selected.

When the target block or target sub-block is a chrominance signal (however, the chrominance signal down-sampled with respect to the luminance signal) (case A3 above), the 4-tap interpolation disclosed in Non-Patent Document 1 A filter is selected.

As shown in FIG. 11, the interpolation filter type for normal MC in Non-Patent Document 2 branches depending on which of the following three conditions the target block or target sub-block meets.
B1. In the case of a luminance signal and a mode other than Affine B2. Luminance signal and Affine mode B3. Color-difference signal If the target block or target sub-block is a luminance signal and is in a mode other than Affine (case B1 above), the following two conditions are applied.

First, if the target block's hpelIfIdx is valid (hpelIfIdx=1) and the target MV references half-pixel precision locations, a 6-tap Gaussian filter is selected.

Second, in cases other than the above, the 12-tap interpolation filter disclosed in Non-Patent Document 2 is selected.

When the target block or target sub-block is a luminance signal and is in Affine mode (case B2 above), the 12-tap interpolation filter disclosed in Non-Patent Document 2 is selected.

When the target block or target sub-block is a chrominance signal (however, the chrominance signal down-sampled with respect to the luminance signal) (case B3 above), the 4-tap interpolation disclosed in Non-Patent Document 2 A filter is selected.

As described above, Non-Patent Document 2 has an extended interpolation filter length (number of taps) compared to Non-Patent Document 1. As a result, compared to Non-Patent Document 1, Non-Patent Document 2 has a steeper frequency response in the transition region as shown in FIG. Better performance. On the other hand, Non-Patent Document 2 has an increased memory bandwidth compared to Non-Patent Document 1.

As shown in FIG. 11, the interpolation filter type for normal MC of the prior application has the same configuration as Non-Patent Document 1.

As shown in FIG. 11, the interpolation filter type for normal MC according to the present embodiment may have the same configuration as in Non-Patent Document 1 or Non-Patent Document 2. However, when the interpolation filter type for normal MC according to the present embodiment has the same configuration as Non-Patent Document 2, the memory band increases as compared to Non-Patent Document 2 as compared to Non-Patent Document 2, but encoding Better performance.

(Block size to which MC is normally applicable)
Hereinafter, interpolation filter types for normal MC according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment will be described with reference to FIG.

FIG. 13 is an example of a comparison table of minimum sizes of uni-prediction or bi-prediction target blocks or target sub-blocks to which normal MC can be applied according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment. is. Hereinafter, the minimum size of a target block or target sub-block for uni-prediction is denoted as S _Uni-Min , and the minimum size of a target block or target sub-block for bi-prediction is denoted as S _Bi-Min .

S _Uni-Min and S _Bi-Min in Non-Patent Document 1 are 4×8/8×4 pixels and 8×8 pixels, respectively.

S _Uni-Min and S _Bi-Min in Non-Patent Document 2 are 4×4 pixels and 4×4 pixels, respectively. As a result, motion compensation prediction can be performed with a finer block size according to the image characteristics. increases.

S _Uni-Min and S _Bi-Min in the prior application are the same as in Non-Patent Document 1.

As shown in FIG. 13, S _Uni-Min and S _Bi-Min in this embodiment may have the same configuration as in Non-Patent Document 1 or Non-Patent Document 2. FIG. However, when S _Uni-Min and S _Bi-Min in the present embodiment have the same configuration as Non-Patent Document 2, the coding performance is improved in the same manner as Non-Patent Document 2 with respect to Non-Patent Document 1. , the memory bandwidth increases.

(Applicable block size of OBMC)
Hereinafter, with reference to FIG. 14, block sizes applicable to OBMC according to Non-Patent Document 2, the prior application, and the present embodiment will be described.

FIG. 14 is an example of a comparison table of block sizes applicable to OBMC according to Non-Patent Document 2, the prior application, and the present embodiment.

Non-Patent Document 2 makes it possible to apply OBMC to blocks of 64 pixels or more for both uni-prediction and bi-prediction. That is, Non-Patent Document 2 has a configuration in which OBMC is not applied to blocks of 32 pixels or less (4×4 pixels or 4×8/8×4 pixels).

The prior application states that OBMC can be applied to uni-prediction blocks. Furthermore, the applicable block size is 4×8/8×4 pixels or more, which is the minimum size of Non-Patent Document 1.

As a result, in the prior application, the increase in memory bandwidth due to OBMC can be limited to uni-prediction normal MC. In addition, by controlling the interpolation filter length (number of taps) according to the block size of the target block or target sub-block or the number of adjacent MVs, which will be described later, OBMC can be applied from small-sized blocks where the memory bandwidth tends to increase. The improvement in coding performance can be maintained while suppressing the increase.

In contrast to the conventional technology described above, the present embodiment enables OBMC to be applied to uni-prediction or bi-prediction blocks with a minimum size of 4×4 pixels or more, which is the minimum size of Non-Patent Document 2, respectively.

As a result, in the present embodiment, the number of blocks to which OBMC can be applied is increased, and the potential performance improvement effect of OBMC can be exploited. A new increase in flag decoding amount (transmission amount from the encoder's point of view) occurs. The details of how to solve these two problems will be described later.

As a modification, depending on the tap length of the interpolation filter prepared in advance in the image decoding device 200 (for example, in the case of an 8-tap filter), OBMC may be set to 32 pixels or more for uni-prediction or bi-prediction blocks. good.

(Interpolation filter type for OBMC)
OBMC interpolation filter types according to Non-Patent Document 2, the prior application, and the present embodiment will be described below with reference to FIG.

FIG. 11 is an example of a comparison table of interpolation filter types for normal MC and interpolation filter types for OBMC according to Non-Patent Document 1, Non-Patent Document 2, the prior application, and the present embodiment, respectively. Note that the types of interpolation filters for normal MC are as described above.

As shown in FIG. 11, the OBMC interpolation filter type of Non-Patent Document 2 branches depending on which of the following two conditions the target block or target sub-block meets.
C1. If it is a luminance signal C2. Color difference signal When the target block or target sub-block is a luminance signal (C1. above), the following two conditions are branched.

First, if the target block's hpelIfIdx is valid (hpelIfIdx=1) and the neighboring MVs refer to half-pixel precision locations, a 6-tap Gaussian filter is selected.

Second, in cases other than the above, the 12-tap interpolation filter disclosed in Non-Patent Document 1 is selected.

When the target block or target sub-block is a chrominance signal (however, the chrominance signal down-sampled with respect to the luminance signal) (case C2 above), the 4-tap interpolation disclosed in Non-Patent Document 1 A filter is selected.

As shown in FIG. 11, the OBMC interpolation filter type of the prior application branches depending on whether the target block or target sub-block satisfies the following two conditions.
D1. If it is a luminance signal D2. When the target block or target sub-block is a luminance signal (in the case of D1. above), 2-tap to 8-tap disclosed in Non-Patent Document 1 can be selected depending on the size of the block. An interpolation filter is selected.

Specifically, the smaller the block size, the shorter the interpolation filter length (short tap) interpolation filter (eg, 2-tap filter) is selected, and the larger the block size, the longer the interpolation filter length (long tap) interpolation filter. (eg, an 8-tap filter) is selected. As a result, the effect of improving the coding performance can be maintained while suppressing an increase in memory bandwidth to a certain degree.

When the target block or target sub-block is a chrominance signal (however, the chrominance signal down-sampled with respect to the luminance signal) (case D2. above), 2 taps disclosed in Non-Patent Document 1 to 4 An interpolation filter of taps is selected. Specifically, a short-tap interpolation filter is selected for a smaller block size, and a long-tap interpolation filter is selected for a larger block size.

Which interpolation filter length (number of taps) of the interpolation filter can be selected for which block size may be designed to be below the upper limit of the memory bandwidth intended by the designer in advance. It is shown that.

In addition, in the prior application, as a modification of the interpolation filter type selection method, a method of varying the selectable interpolation filter length in consideration of not only the block size of the target block or target sub-block but also the number of adjacent MVs is shown. there is

According to the above, by using the interpolation filter type selection method for OBMC of the prior application, it is possible to maintain the effect of improving the coding performance while suppressing the increase in the memory band to a certain extent.

On the other hand, shortening the interpolation filter length of OBMC for small-sized blocks means that the potential effect of improving the coding performance of OBMC cannot be exploited. This is because, as described above, the longer the interpolation filter length, the smaller the prediction error for sharp edges and the like, but the shorter the interpolation filter length, the larger the prediction error for steep edges and the like.

Since small-sized blocks are more likely to be selected in areas with complex image characteristics such as sharp edges and a lot of motion, the longer the interpolation filter length is, the more the prediction error is reduced.

In addition, since small-sized blocks tend to have large differences in MV between adjacent blocks due to their nature, OBMC is likely to reduce prediction errors at block boundaries or sub-block boundaries.

Considering such a problem of the conventional technology, the present embodiment controls the interpolation filter length as follows according to the block size of the target block or target sub-block, contrary to the prior application.

If the block size of the target block or target sub-block is small, lengthen the interpolation filter length (use a long-tap interpolation filter for OBMC)
On the other hand, if the block size of the target block or target sub-block is large, shorten the interpolation filter length (use a short-tap interpolation filter for OBMC).
Here, a plurality of block size thresholds for switching the interpolation filter length may be provided. For example, a long-tap (12-tap or 8-tap) interpolation filter is used for a small size block of 32 pixels or less (4×4 pixels or 4×8/8×4 pixels).

On the other hand, for large-sized blocks such as 256 pixels or more, 512 pixels or more, or 1024 pixels or more, a short-tap (4-tap or 2-tap) interpolation filter is used.

As a modification, a 6-tap Gaussian filter with a high smoothing effect may be preferentially used for large-sized blocks.

The increase in memory bandwidth due to the application of long-tap OBMC to small-sized blocks is controlled by adjusting the following parameters, which will be described later.
1. Number of adjacent MVs2. Number of Blending Lines If there are two or more adjacent MVs, select one adjacent MV and apply OBMC.

The details of the method for controlling the number of blending lines according to this embodiment will be described later.

(Number of OBMC blending lines)
Hereinafter, the number of blending lines of OBMC according to Non-Patent Document 2, the prior application, and the present embodiment will be described with reference to FIG.

FIG. 15 is an example of a comparison table of OBMC blending line numbers according to Non-Patent Document 2, the prior application, and the present embodiment.

Here, L luma_B represents the number of blending lines for the target block boundary in the luminance signal, L _{chroma_B} represents the number of blending lines for the target block boundary in the chrominance signal, and L _{luma_SbB} represents the number of blending lines for the target sub-block boundary in the luminance signal _. , and the number of blending lines for the target sub-block boundary in the chrominance signal is indicated as L _{chroma_SbB} .

In Non-Patent Document 2, fixed values of 4 lines, 1 line, 3 lines, and 1 line are used for L _{luma_B} , L _{chroma_B} , L _{luma_SbB} , and L _{chroma_SbB} , respectively.

In contrast, the prior application shows a method of decreasing the number of blending lines when the block size of OBMC is small, and increasing the number of blending lines when the block size is large.

Specifically, in the prior application, variable values of 2 lines and 4 lines according to the block size are used for L _{luma_B} and L _{chroma_B} , and a fixed value of 1 line is used for L _{luma_SbB} and L _{chroma_SbB} . It is configured.

More specifically, in the prior application, L _{luma_B} and L _{chroma_B} have two blending lines in the parallel direction to the boundary where the horizontal or vertical width of the block is 4 pixels or less. In the case of , the number of blending lines in the direction parallel to the boundary of 4 pixels or less is 4 lines.

As a result, according to the prior application, it is possible to suppress an increase in memory bandwidth due to the application of OBMC to small-sized blocks.

In contrast to the above-described prior art, according to the present embodiment, the following two OBMC blending line number control methods are used as techniques for suppressing the memory bandwidth and further improving the encoding performance.
・ Further reduce the number of OBMC blending lines for small size blocks (4 lines to 1 line)
・ Further expand the number of blending lines of OBMC for large size blocks (4 lines to 8 lines)
As means for realizing this, for example, a plurality of threshold values may be provided for each block size. For example, in the case of 32 pixels or less (4×4 pixels or 4×8/8×4 pixels), 1 line or 2 lines are used, and in the case of 64 pixels or more and less than 256 pixels, 4 lines or 3 lines are used. In the case of pixels or more, 8 lines may be used.

Also, the number of lines for the width and length of the block may be controlled separately. For example, for a block of 4×16 pixels/16×4 pixels, etc., the number of blending lines in the direction parallel to the short side (4 pixels) is 1 or 2, and the blending line in the direction parallel to the long side (16 pixels) is The number of in-lines may be 4 lines or 3 lines. The increase and decrease in the number of lines in the parallel direction (short side and long side) may be reversed.

As a result of the above, the following effects can be expected.

For small-sized blocks, by further reducing the number of lines compared to the conventional technology, a long-tap interpolation filter can be applied to OBMC. As a result, while suppressing an increase in the OBMC memory bandwidth due to the extension of the interpolation filter length, it is possible to enjoy the benefit of reduction in prediction error due to the application of the long-tap interpolation filter.

On the other hand, for large-sized blocks, by further increasing the number of lines compared to the conventional technology, it becomes easier to enjoy the effect of smoothing discontinuities at block boundaries, which tend to occur in large-sized blocks, over a wide range from the boundaries.

(Method for judging applicability of OBMC)
A method for determining whether OBMC is applicable or not according to Non-Patent Document 2, the prior application, and the present embodiment will be described below with reference to FIGS. 16 and 17. FIG.

In Non-Patent Document 2 and the prior application, as a method for determining whether or not OBMC is applicable, a method of decoding in target block units (transmission from the image encoding apparatus 100) and control data for controlling whether or not OBMC is applicable A method of making a determination using a flag (OBMC flag) is disclosed.

Here, the decoding unit 210 determines whether or not to decode the OBMC flag. Specifically, the flowchart shown in FIG. 16 shows such processing.

As shown in FIG. 16, in step S16-1, the decoding unit 210 determines decoding condition 1 of the OBMC flag.

Here, if the decoding condition 1 of the OBMC flag is not satisfied, the process proceeds to step S16-2. On the other hand, if the decoding condition 1 of the OBMC flag is satisfied, the process proceeds to step S16-3. The details of the decoding condition 1 of the OBMC flag will be described later.

At step S16-2, the decoding unit 210 determines the decoding condition 2 of the OBMC flag.

Here, if the decoding condition 2 of the OBMC flag is satisfied, the process proceeds to step S16-4. On the other hand, if the decoding condition 2 of the OBMC flag is not satisfied, the process proceeds to step S16-5. The details of the decoding condition 2 of the OBMC flag will be described later.

In step S16-3, the decoding unit 210 does not decode the OBMC flag, estimates the value of the OBMC flag to be 0, and terminates this process.

At step S16-4, the decoding unit 210 decodes the OBMC flag, and ends this process.

In step S16-5, the decoding unit 210 does not decode the OBMC flag, estimates the value of the OBMC flag to be 1, and terminates this process.

　The value of the OBMC flag consists of 0 or 1. An OBMC flag value of 1 indicates that OBMC is applicable, and an OBMC flag value of 0 indicates that OBMC is not applicable.

The OBMC applicability determination unit 241B2 uses the value of the OBMC flag, the block size and prediction mode of the target block, the prediction mode of the adjacent block, the target MV and the adjacent MV to finally determine whether or not OBMC is applicable. is configured to Details of such a determination method will be described later.

FIG. 17 is an example of a comparison table of the OBMC flag decoding condition 1, the OBMC flag decoding condition 2, and the determination method for OBMC applicability according to Non-Patent Document 2, the prior application, and the present embodiment, respectively.

First, each decoding condition and determination method according to Non-Patent Document 2 will be described.

The decoding condition 1 of the OBMC flag in Non-Patent Document 2 is composed of the following conditions.
E1. The OBMC flag for each slice is 0.
E2. The target block is in intra prediction mode.
E3. The target block is in IBC mode.
E4. The target block is in LIC mode.

Here, the OBMC flag for each slice is a flag that controls whether or not OBMC decoding for each target slice is applicable. Such an OBMC flag may be decoded on a picture-by-picture basis instead of being decoded on a slice-by-slice basis. The value of the OBMC flag consists of 0 and 1, and has the same meaning as the object block unit.

The combination of conditions E1. to E4. above may be changed according to the designer's intention.

Decoding condition 2 of the OBMC flag in Non-Patent Document 2 consists only of the condition that the target block is in merge mode.

The method of determining whether or not OBMC can be applied in Non-Patent Document 2 consists of the following conditions F1. to F6. If any one of the conditions F1.-F6. is satisfied, it is determined that OBMC is not applicable. If none of the conditions F1.-F6. are satisfied, it is determined that OBMC is applicable.
F1. OBMC flag=0.
F2. The target block is in intra prediction mode.
F3. The target block is in IBC mode.
F4. The target block is in LIC mode.
F5. The neighboring block is in intra prediction mode.
F6. The target MV and adjacent MV are the same.

The combination of the above conditions F1. to F6. may be changed according to the designer's intention.

Secondly, each decoding condition and determination method related to the prior application will be explained.

The decoding condition 1 of the OBMC flag of the prior application consists of the following conditions.
G1. The OBMC flag for each slice is 0.
G2. The target block is in intra prediction mode.
G3. The target block is in IBC mode.

Here, the OBMC flag for each slice is a flag that controls whether or not OBMC decoding for each target slice is applicable. Such an OBMC flag may be decoded on a picture-by-picture basis instead of being decoded on a slice-by-slice basis. The value of the OBMC flag consists of 0 and 1, and has the same meaning as the object block unit.

The combination of conditions G1. to G3. above may be changed according to the designer's intention.

The decoding condition 2 of the OBMC flag of the prior application consists of the following two conditions.
H1. The target block is in merge mode.
H2. The target block is in AMVP mode and the target block size is 256 pixels or more.

The method for judging the applicability of OBMC in the prior application consists of the following conditions I1. to I5. If any one of the conditions I1.-I5. is satisfied, it is determined that OBMC is not applicable. If none of the conditions I1.-I5. are satisfied, it is determined that OBMC is applicable.
I1. OBMC flag=0.
I2. The target block is in intra prediction mode.
I3. The target block is in IBC mode.
I4. The neighboring block is in intra prediction mode.
I5. The error between the target MV and the adjacent MV is less than the threshold (1.5 pixels for uni-prediction, 1 pixel for bi-prediction)
Combinations of the above conditions I1. to I6. may be changed according to the designer's intention.

Third, each decoding condition and determination method according to this embodiment will be described.

The decoding condition 1 of the OBMC flag in this embodiment may be the same as in Non-Patent Document 2 or the prior application.

The decoding condition 2 of the OBMC flag of this embodiment is composed of the following three conditions.
J1. The target block is in merge mode.
J2. The target block is in AMVP mode and the target block size is 256 pixels or more.
J3. The target block is in AMVP mode and the target block size is 32 pixels or less.

By adding the above condition J3., the following effects can be expected.

As described above, in OBMC according to the present embodiment, a long-tap interpolation filter can be applied to small-sized blocks. As a result, the effect of reducing the OBMC prediction error in small-sized blocks can be expected, but if the OBMC flag is decoded for each small-sized block, the amount of code increases accordingly.

As a means of avoiding such a situation, when a long-tap interpolation filter is prepared for a small-sized block of OBMC, a small-sized block (specifically, the target block is AMVP mode and the target block size is 32 pixels below), by omitting the decoding process of the OBMC flag, an increase in the code amount of the OBMC flag in the small size block can be reduced.

As another modification, the block size threshold for omitting the OBMC flag decoding process for the current block in AMVP mode may be controlled according to the block size applicable to OBMC and the interpolation filter length that can be used in OBMC.

For example, if a long-tap interpolation filter can be applied to both small-sized blocks and large-sized blocks, the above threshold is set small. One method is to set the lower limit of the block size applicable to OBMC as such a threshold. Judgment condition J4. is configured as follows.
J4. The target block is in AMVP mode and the target block size is equal to or larger than the lower limit of the block size to which OBMC is applicable.

The lower limit of the block size to which the above-mentioned OBMC can be applied is, for example, in Non-Patent Document 2 and in the present embodiment, 64 pixels or 32 pixels, which allows long-tap interpolation filters to be used from small size blocks.

This J4. to J2. and J3. can further simplify the decoding condition regarding whether or not the OBMC flag is decoded.

On the other hand, if a short-tap interpolation filter is applicable to small-sized blocks and a long-tap interpolation filter is applicable to large-sized blocks, the above threshold may be set large. One method is to set such a threshold to the lower limit of the block size at which the long-tap interpolation filter can be applied. For example, if a short-tap interpolation filter can be applied to a target block of less than 256 pixels, and a long-tap interpolation filter can be applied to a target block of 256 pixels or more, even if the threshold is set to 256 pixels, good.

As a result, the effect of reducing the memory bandwidth by controlling the interpolation filter length according to the block size (applying a short-tap interpolation filter to small-sized blocks) and the effect of reducing the code amount by omitting the decoding process of the OBMC flag are obtained. I can expect it.

The method of determining whether or not OBMC is applicable in this embodiment may be the same as in Non-Patent Document 2 or the prior application.

(Example of modification of OBMC applicability determination method)
Hereinafter, with reference to FIGS. 18 to 20, modified examples of the method of determining the applicability of OBMC according to Non-Patent Document 2, the prior application, and the present embodiment will be described.

In this modification, the following two conditions are added to the OBMC flag decoding condition 1 determined by the decoding unit 210 and the OBMC applicability determination condition determined by the OBMC applicability determination unit 241B.
K1. AMVR is valid and MVD is decimal pixel precision in the target block.
K2. The target block DMMVD is valid and MVD is decimal pixel precision.

Here, AMVR (Adaptive Motion Vector Resolution) and MMVD (Merge with Motion Vector Difference) are technologies disclosed in Non-Patent Document 1, respectively.

AMVR is a technology that controls the transmission accuracy of the motion vector difference (MVD) added to the motion vector prediction (MVP) derived by AMVP, and the final target MV is rounded to the transmission accuracy of such MVD.

The transmission precision that can be controlled by AMVR is defined for each of Affine AMVP, normal AMVP, and IBC, as shown in FIG. ), an Affine mode non-applicability flag (inter_affine_flag), and an internal parameter (MODE_IBC) indicating that the target block is in the IBC mode.

MVD (MmvdDistance) added by MMVD is ph_mmvd_fullpel_only_flag decoded in picture units and mmvd_distance_idx decoded in target block units, as shown in FIG.
is determined by

In this modified example, when AMVR is valid in the target block, for example, when the transmission precision of MVD becomes decimal pixel precision for Affine AMVP and normal AMVP, the decoding unit 210 decodes the OBMC flag. and assumes the value of the OBMC flag to be 0.

Further, in this modified example, when the MMVD is valid in the target block and the MVD added by the MMVD has decimal pixel precision, the decoding unit 210 sets the value of the OBMC flag to 0 without decoding the OBMC flag. presume.

The effects of such control are as follows.

When AMVR or MMVD is valid for the target block, it is possible for the decoding unit 210 (in the control data decoding stage) to determine whether the target MV has decimal pixel precision or integer pixel precision.

If the target MV has sub-pixel accuracy even by AMVR or MMVD, it can be regarded as fine motion compensation prediction, so discontinuity (prediction error) at block boundaries is less likely to occur. Therefore, OBMC may not be applicable.

On the other hand, if the target MV is rounded to integer pixel precision by AMVR or MMVD, it can be regarded as coarse motion compensation prediction, so discontinuity (prediction error) at block boundaries is likely to occur. In this case, by making it possible to apply always-on OBMC that reduces prediction errors, an improvement in coding performance can be expected.

In addition, in the decoding condition 2 of the OBMC flag, if AMVR or MMVD is valid in the target block and the precision of the MVD can be identified as integer precision, the decoding unit 210 decodes the OBMC flag without decoding the OBMC flag. is assumed to be 1, OBMC can always be applied when the target MV is rounded to integer pixel accuracy by AMVR and MMVD described above, and unnecessary code amount can be reduced.

According to the present embodiment described above, by controlling the OBMC interpolation filter length according to the target block size or the number of OBMC blending lines, it is possible to maintain the effect of improving the coding performance while suppressing an increase in the memory bandwidth. .

Further, according to the present embodiment, by controlling the number of blending lines of OBMC according to the target block size or the interpolation filter lengths of MC and OBMC, an increase in memory bandwidth can be suppressed while improving the coding performance. can be maintained.

Further, according to the present embodiment, by controlling whether or not OBMC can be applied according to the target block size or the interpolation filter lengths of MC and OBMC, excessive decoding processing ( bit amount) can be reduced.

Furthermore, according to the present embodiment, by controlling whether or not OBMC can be applied according to the MV accuracy of the target block or adjacent blocks, excessive decoding processing (bit amount) of the OBMC flag transmitted in target block units can be reduced.

(Matching interpolation filter types used for normal MC and OBMC prediction)
An embodiment for matching the types of interpolation filters used for normal MC and OBMC prediction will be described below. In Non-Patent Document 2, a case may occur where interpolation filter lengths with different properties are used for normal MC prediction and OBMC prediction. Specifically, this is a case where a normal interpolation filter is applied to one side and the SIF (6-tap Gaussian filter) is applied to the other side.

The above case can be avoided by matching the interpolation filter used for OBMC prediction with the interpolation filter used for normal MC prediction.

As one method, in OBMC prediction, the hpelIfIdx value of the adjacent block is ignored, and the hpelIfIdx value of the target block is prioritized to determine the type of interpolation filter.

For example, when the hpelIfIdx of the target block is invalid and the SIF is not selected, and the hpelIfIdx of the adjacent block is valid and the SIF is selected, the hpelIfIdx value of the target block is prioritized to perform interpolation of both the target block and the adjacent block. The filter type can be matched with a normal interpolation filter.

On the other hand, when the hpelIfIdx of the target block is valid and the SIF is not selected, and the hpelIfIdx of the adjacent block is invalid and the SIF is not selected, by prioritizing the value of the hpelIfIdx of the target block, the interpolation filter of both the target block and the adjacent block is selected. The type can be matched with SIF.

However, if the adjacent MV of the adjacent block is not a 1/2 pixel precision position, the SIF may be invalid. It may be matched with an interpolation filter.

As described above, it is possible to avoid blending of reference pixels to which interpolation filters having different properties are applied by OBMC prediction, improve prediction performance, and as a result improve coding performance.

The image encoding device 100 and the image decoding device 200 described above may be implemented as a program that causes a computer to execute each function (each process).

In each of the above-described embodiments, the present invention is applied to the image encoding device 100 and the image decoding device 200 as examples, but the present invention is not limited to this. The same can be applied to an image encoding system and an image decoding system having the functions of the device 100 and the image decoding device 200. FIG.

DESCRIPTION OF SYMBOLS 10... Image processing system 100... Image encoding apparatus 111, 241... Inter prediction part 112, 242... Intra prediction part 121... Subtractor 122, 230... Adder 131... Transform/ quantization part 132, 220... Inverse transform/inverse Quantization unit 140 Encoding units 150, 250 In- loop filtering units 160, 260 Frame buffer 200 Image decoding device 210 Decoding unit 241A Motion vector (MV) deriving unit 241B Motion compensation (MC) prediction signal Generation unit 241B1: Normal MC processing unit 241B2: OBMC applicability determination unit 241B3: OBMC processing unit

Claims (10)

1. An image decoding device to which OBMC prediction is applicable,
   an OBMC applicability determination unit configured to determine applicability of the OBMC prediction for each target block;
   an OBMC processing unit configured to process the OBMC prediction for each target block;
   a decoding unit configured to determine whether or not an OBMC flag configured to control applicability of the OBMC prediction for each target block is decoded, and to decode the OBMC flag. An image decoding device characterized by:
2. 2. The image decoding apparatus according to claim 1, wherein the OBMC processing unit is configured to vary an interpolation filter length for OBMC according to the block size of the target block or the number of OBMC blending lines. .
3. The OBMC processing unit uses a long-tap interpolation filter in the OBMC processing of small-sized blocks and uses a short-tap interpolation filter in the OBMC processing of large-sized blocks when the number of OBMC blending lines falls within a certain range. 3. An image decoding device according to claim 2, characterized in that it is adapted to use.
4. 2. The image decoding device according to claim 1, wherein the OBMC processing unit is configured to vary the number of OBMC blending lines according to the block size of the target block or an interpolation filter length for OBMC. .
5. When the interpolation filter length for OBMC falls within a certain range, the OBMC processing unit reduces the number of OBMC blending lines for small-sized blocks, and reduces the number of OBMC blending lines for large-sized blocks. 5. The image decoding device according to claim 4, wherein the image decoding device is configured to extend the .
6. The decoding unit is configured to estimate the value of the OBMC flag to be 1 without decoding the OBMC flag when the OBMC prediction using a long-tap interpolation filter is applicable to small-sized blocks. 2. The image decoding device according to claim 1, wherein:
7. The decoding unit specifies the accuracy of MVD from control data when AMVR or MMVD is valid in the target block, and the OBMC without decoding the OBMC flag when the target MV is specified as decimal pixel precision. 2. The image decoding device according to claim 1, wherein the value of the flag is estimated to be 0.
8. The decoding unit specifies the accuracy of MVD from control data when AMVR or MMVD is valid in the target block, and when the target MV is specified as integer pixel precision, the OBMC without decoding the OBMC flag. 8. The image decoding device according to claim 1, which is configured to estimate a flag value of 1.
9. An image decoding method to which OBMC prediction is applicable,
   determining applicability of the OBMC prediction for each target block;
   processing the OBMC prediction for each target block;
   determining whether or not an OBMC flag configured to control applicability of the OBMC prediction for each target block is decoded, and decoding the OBMC flag.
10. A program that causes a computer to function as an image decoding device to which OBMC prediction can be applied,
    The image decoding device is
    an OBMC applicability determination unit configured to determine applicability of the OBMC prediction for each target block;
    an OBMC processing unit configured to process the OBMC prediction for each target block;
    a decoding unit configured to determine whether or not an OBMC flag configured to control applicability of the OBMC prediction for each target block is decoded, and to decode the OBMC flag. Program characterized.

Images