Image decoding method and device therefor
According to the disclosure of the present document, weight index information for sub-block merge candidates of a current block can be derived and coding efficiency can be increased.
1. An image decoding method performed by a decoding apparatus, the method comprising:
obtaining image information comprising inter prediction mode information and residual information through a bitstream;
generating residual samples based on the residual information;
generating a sub-block merge candidate list for a current block based on the inter prediction mode information;
selecting one candidate from among candidates included in the sub-block merge candidate list;
generating L0 prediction samples and L1 prediction samples of the current block based on the selected candidate, wherein the weight index is for the weighted averaging;
generating prediction samples by weighted averaging the L0 prediction samples and the L1 prediction samples based on a weight index for the selected candidate, wherein the weight index is for the weighted averaging; and
generating reconstructed samples based on the prediction samples and the residual samples,
wherein the sub-block merge candidate list comprises a constructed affine merge candidate,
wherein the constructed affine merge candidate comprises at least one of a control point motion vector CPMV0 for a control point CP0, a control point motion vector CPMV1 for a control point CP1, a control point motion vector CPMV2 for a control point CP2 or a control point motion vector CPMV3 for a control point CP3,
wherein the CP0 is related to a top-left corner of the current block, the CP1 is related to a top-right corner of the current block, the CP2 is related to a bottom-left corner of the current block and the CP3 is related to a bottom-right corner of the current block,
wherein based on a case that the constructed affine merge candidate is generated based on {CP0, CP1}, the constructed affine merge candidate comprises CPMV0 and the weight index for the constructed affine merge candidate is equal to the weight index for the CP0, and
wherein based on a case that the constructed affine merge candidate is generated based on {CP1, CP2, CP3}, the constructed affine merge candidate comprises CPMV1 and the weight index for the constructed affine merge candidate is equal to the weight index for the CP1.
2. The image decoding method of claim 1 , wherein the weight index for the CP0 corresponds to a weight index of a block that is used for deriving the CPMV0 among the top-left neighboring block, the first left neighboring block, and the first top neighboring block of the current block.
3. The image decoding method of claim 1 , wherein the weight index for the CP1 corresponds to a weight index of a block that is used for deriving the CPMV1 among the top-right neighboring block and the second top neighboring block of the current block.
4. The image decoding method of claim 1 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein a weight index for the pair-wise candidate is derived based on a weight index of the first candidate.
5. The image decoding method of claim 1 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein based on a case that a weight index for the first candidate and a weight index for the second candidate are identical to each other, a weight index for the pair-wise candidate is derived based on a weight index of the first candidate,
wherein based on a case that the weight index for the first candidate and the weight index for the second candidate are not same, the weight index for the pair-wise candidate is derived based on a default weight index, and
wherein the default weight index corresponds to a weight index for giving a same weight to each of the L0 prediction samples and the L1 prediction samples.
6. The image decoding method of claim 1 , wherein the candidates comprise a subblock-based temporal motion vector prediction (SbTMVP) candidate, and
wherein a weight index for the SbTMVP candidate is derived as 0.
7. The image decoding method of claim 1 , wherein the candidates comprise a subblock-based temporal motion vector prediction (SbTMVP) candidate,
wherein a weight index for the SbTMVP candidate is derived based on weight index information for a center block in a col block,
wherein the col block comprises a block which is located in a reference picture different from a current picture in which the current block is located, and which is derived based on a motion vector of a left neighboring block of the current block, and
wherein the center block comprises a bottom-right subblock among four subblocks located in a center of the col block.
8. An image encoding method performed by an encoding apparatus, the method comprising:
determining an inter prediction mode of a current block and generating inter prediction mode information indicating the inter prediction mode;
generating a sub-block merge candidate list for the current block based on the inter prediction mode;
generating selection information indicating one candidate among candidates included in the sub-block merge candidate list;
generating residual information based on residual samples of the current block; and
encoding image information comprising the inter prediction mode information, the selection information, and the residual information,
wherein the sub-block merge candidate list comprises a constructed affine merge candidate,
wherein the constructed affine merge candidate comprises at least one of a control point motion vector CPMV0 for a control point CP0, a control point motion vector CPMV1 for a control point CP1, a control point motion vector CPMV2 for a control point CP2 or a control point motion vector CPMV3 for a control point CP3,
wherein the CP0 is related to a top-left corner of the current block, the CP1 is related to a top-right corner of the current block, the CP2 is related to a bottom-left corner of the current block and the CP3 is related to a bottom-right corner of the current block,
wherein based on a case that the constructed affine merge candidate is generated based on {CP0, CP1}, the constructed affine merge candidate comprises CPMV0 and a weight index for the constructed affine merge candidate is equal to the weight index for the CP0,
wherein based on a case that the constructed affine merge candidate is generated based on {CP1, CP2, CP3}, the constructed affine merge candidate comprises CPMV1 and the weight index for the constructed affine merge candidate is equal to the weight index for the CP1, and
wherein the weight index for the constructed affine merge candidate is for a weighted averaging based on L0 prediction samples and L1 prediction samples.
9. The image encoding method of claim 8 , wherein the weight index for the CP0 corresponds to a weight index of a block that is used for deriving the CPMV0 among the top-left neighboring block, the first left neighboring block, and the first top neighboring block of the current block.
10. The image encoding method of claim 8 , wherein the weight index for the CP1 corresponds to a weight index of a block that is used for deriving the CPMV1 among the top-right neighboring block and the second top neighboring block of the current block.
11. The image encoding method of claim 8 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein a weight index for the pair-wise candidate is derived based on a weight index of the first candidate.
12. The image encoding method of claim 8 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein based on a case that a weight index for the first candidate and a weight index for the second candidate are identical to each other, a weight index for the pair-wise candidate is derived based on a weight index of the first candidate,
wherein based on a case that the weight index for the first candidate and the weight index for the second candidate are not same, the weight index for the pair-wise candidate is derived based on a default weight index, and
wherein the default weight index corresponds to a weight index for giving a same weight to each of the L0 prediction samples and the L1 prediction samples.
13. The image encoding method of claim 8 , wherein the candidates comprise a subblock-based temporal motion vector prediction (SbTMVP) candidate, and
wherein a weight index for the SbTMVP candidate is derived as 0.
14. The image encoding method of claim 8 , wherein the candidates comprise a subblock-based temporal motion vector prediction (SbTMVP) candidate,
wherein a weight index for the SbTMVP candidate is derived based on weight index information for a center block in a col block,
wherein the col block comprises a block which is located in a reference picture different from a current picture in which the current block is located, and which is derived based on a motion vector of a left neighboring block of the current block, and
wherein the center block comprises a bottom-right subblock among four subblocks located in a center of the col block.
15. A non-transitory computer readable storage medium storing a bitstream generated by an image encoding method, the method comprising:
determining an inter prediction mode of a current block and generating inter prediction mode information indicating the inter prediction mode;
generating a sub-block merge candidate list for the current block based on the inter prediction mode;
generating selection information indicating one candidate among candidates included in the sub-block merge candidate list;
generating residual information based on residual samples of the current block; and
encoding image information to generate the bitstream, wherein the image information comprises the inter prediction mode information, the selection information, and the residual information,
wherein the sub-block merge candidate list comprises a constructed affine merge candidate,
wherein the constructed affine merge candidate comprises at least one of a control point motion vector CPMV0 for a control point CP0, a control point motion vector CPMV1 for a control point CP1, a control point motion vector CPMV2 for a control point CP2 or a control point motion vector CPMV3 for a control point CP3,
wherein the CP0 is related to a top-left corner of the current block, the CP1 is related to a top-right corner of the current block, the CP2 is related to a bottom-left corner of the current block and the CP3 is related to a bottom-right corner of the current block,
wherein based on a case that the constructed affine merge candidate is generated based on {CP0, CP1}, the constructed affine merge candidate comprises CPMV0 and a weight index for the constructed affine merge candidate is equal to the weight index for the CP0,
wherein based on a case that the constructed affine merge candidate is generated based on {CP1, CP2, CP3}, the constructed affine merge candidate comprises CPMV1 and the weight index for the constructed affine merge candidate is equal to the weight index for the CP1, and
wherein the weight index for the constructed affine merge candidate is for a weighted averaging based on L0 prediction samples and L1 prediction samples.
16. The non-transitory computer readable storage medium of claim 15 , wherein the weight index for the CP0 corresponds to a weight index of a block that is used for deriving the CPMV0 among the top-left neighboring block, the first left neighboring block, and the first top neighboring block of the current block.
17. The non-transitory computer readable storage medium of claim 15 , wherein the weight index for the CP1 corresponds to a weight index of a block that is used for deriving the CPMV1 among the top-right neighboring block and the second top neighboring block of the current block.
18. The non-transitory computer readable storage medium of claim 15 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein a weight index for the pair-wise candidate is derived based on a weight index of the first candidate.
19. The non-transitory computer readable storage medium of claim 15 , wherein the candidates comprise a pair-wise candidate, and the pair-wise candidate is derived based on a first candidate and a second candidate in the merge candidate list,
wherein based on a case that a weight index for the first candidate and a weight index for the second candidate are identical to each other, a weight index for the pair-wise candidate is derived based on a weight index of the first candidate,
wherein based on a case that the weight index for the first candidate and the weight index for the second candidate are not same, the weight index for the pair-wise candidate is derived based on a default weight index, and
wherein the default weight index corresponds to a weight index for giving a same weight to each of the L0 prediction samples and the L1 prediction samples.
20. The non-transitory computer readable storage medium of claim 15 , wherein the candidates comprise a subblock-based temporal motion vector prediction (SbTMVP) candidate, and
wherein a weight index for the SbTMVP candidate is derived as 0.