Sub-block based decoder side motion vector derivation
Devices, systems and methods for digital video coding, which include decoder side motion vector derivation (DMVD) tools, are described. An exemplary method for video processing includes making a decision, based on a determination that a current block of a video is coded using a multi-hypothesis prediction mode, regarding a selective enablement of a DMVD tool for the current block, wherein the DMVD tool derives a refinement of motion information signaled in a bitstream representation of the video; and performing, based on the decision, a conversion between the current block and the bitstream representation.
1. A method of processing video data, comprising:
dividing a coding tree unit (CTU) into a plurality of coding units (CUs);
constructing a motion candidate list for a current coding unit (CU);
deriving motion information for the current CU based on the motion candidate list;
dividing the current CU into a plurality of sub-CUs, if at least one condition of a size of the current CU is met;
making a decision regarding a selective enablement of a decoder-side motion vector derivation (DMVD) tool at a sub-CU level, wherein the DMVD tool is used to refine motion information; and
performing, based on the decision, a conversion between the current CU and the bitstream,
wherein the DMVD tool is a bi-directional optical flow tool which is used to refine a motion vector based on at least one gradient value corresponding to a sample in a reference CU of the current CU,
wherein responsive to a height (H) or a width (W) of the current CU being greater than a threshold (L), the current CU is divided into multiple sub-CUs, and wherein L is a positive integer, and wherein L is predetermined.
2. The method of claim 1 , wherein the width of the sub-CU is min(L, W).
3. The method of claim 1 , wherein the height of the sub-CU is min(L, H).
4. The method of claim 1 , wherein the DMVD tool is enabled for each of the plurality of sub-CUs of the current CU.
5. The method of claim 1 , wherein the sub-CU is treated as a block and all operations required in the DMVD tool are performed for the sub-CU.
6. The method of claim 1 , wherein L=64, wherein a size of the current CU is 64×128, 128×64 or 128×128, and wherein a size of each of plurality of sub-CUs is 64×64.
7. The method of claim 1 , wherein L=16.
8. The method of claim 1 , wherein L=64, wherein a size of the current CU is N×128 or 128×N, and wherein a size of both of the plurality of sub-CUs is N×64 or 64×N, respectively.
9. The method of claim 1 , wherein L is signaled in the bitstream in a sequence parameter set (SPS), a picture parameter set (PPS), a picture header, a slice header, a tile group header, or a tile header.
10. The method of claim 1 , wherein L is based on at least one of a size or a coding mode of the current CU, a picture type, or a temporal layer index.
11. The method of claim 1 , wherein the conversion comprises encoding the bitstream from the current CU.
12. The method of claim 1 , wherein the conversion comprises decoding the current CU from the bitstream.
13. An apparatus for processing video data comprising a processor and a non-transitory memory with instructions thereon, wherein the instructions upon execution by the processor, cause the processor to:
divide a coding tree unit (CTU) into a plurality of coding units (CUs);
construct a motion candidate list for a current coding unit (CU);
derive motion information for the current CU based on the motion candidate list;
divide the current CU into a plurality of sub-CUs, if at least one condition of a size of the current CU is met;
make a decision regarding a selective enablement of a decoder-side motion vector derivation (DMVD) tool at a sub-CU level, wherein the DMVD tool is used to refine motion information; and
perform, based on the decision, a conversion between the current CU and the bitstream,
wherein the DMVD tool is a bi-directional optical flow tool which is used to refine a motion vector based on at least one gradient value corresponding to a sample in a reference CU of the current CU,
wherein responsive to a height (H) or a width (W) of the current CU being greater than a threshold (L), the current CU is divided into multiple sub-CUs, and wherein L is a positive integer, and wherein L is predetermined.
14. The apparatus of claim 13 , wherein the DMVD tool is one of a decoder-side motion vector refinement (DMVR) tool which is used to refine a signaled motion vector based on at least one motion vector with an offset to the signaled motion vector, or a bi-directional optical flow tool which is used to refine a motion vector based on at least one gradient value corresponding to a sample in a reference block of the current CU.
15. The apparatus of claim 13 , wherein the width of the sub-CU is min(L, W).
16. The apparatus of claim 13 , wherein the height of the sub-CU is min(L, H).
17. A non-transitory computer-readable storage medium storing instructions that cause a processor to:
divide a coding tree unit (CTU) into a plurality of coding units (CUs);
construct a motion candidate list for a current coding unit (CU);
derive motion information for the current CU based on the motion candidate list;
divide the current CU into a plurality of sub-CUs, if at least one condition of a size of the current CU is met;
make a decision regarding a selective enablement of a decoder-side motion vector derivation (DMVD) tool at a sub-CU level, wherein the DMVD tool is used to refine motion information; and
perform, based on the decision, a conversion between the current CU and the bitstream,
wherein the DMVD tool is a bi-directional optical flow tool which is used to refine a motion vector based on at least one gradient value corresponding to a sample in a reference CU of the current CU,
wherein responsive to a height (H) or a width (W) of the current CU being greater than a threshold (L), the current CU is divided into multiple sub-CUs, and wherein L is a positive integer, and wherein L is predetermined.
18. The non-transitory computer-readable storage medium of claim 17 , wherein the width of the sub-CU is min(L, W).
19. The non-transitory computer-readable storage medium of claim 17 , wherein the height of the sub-CU is min(L, H).
20. A non-transitory computer-readable recording medium storing a bitstream of a video which is generated by a method performed by a video processing apparatus, wherein the method comprises:
divide a coding tree unit (CTU) into a plurality of coding units (CUs);
construct a motion candidate list for a current coding unit (CU);
derive motion information for the current CU based on the motion candidate list;
dividing the current CU into a plurality of sub-CUs, if at least one condition of a size of the current CU is met;
making a decision regarding a selective enablement of a decoder-side motion vector derivation (DMVD) tool at a sub-CU level, wherein the DMVD tool is used to refine motion information; and
generating, based on the decision, the bitstream from the current CU,
wherein the DMVD tool is a bi-directional optical flow tool which is used to refine a motion vector based on at least one gradient value corresponding to a sample in a reference CU of the current CU,
wherein responsive to a height (H) or a width (W) of the current CU being greater than a threshold (L), the current CU is divided into multiple sub-CUs, and wherein L is a positive integer, and wherein L is predetermined.