Multiple hypothesis for sub-block prediction blocks

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 current video block into multiple sub-blocks;
• generate a merge candidate list for at least one sub-block of the multiple sub-blocks;
• and
• perform a conversion between the video block and a bitstream of the video block based on the merge candidate list,
• wherein the merge candidate list comprises at least one merge candidate with multi-hypothesis mode;
• the at least one merge candidate with multi-hypothesis mode includes a merge candidate calculated by jointly utilizing two uni-prediction merge candidates in the merge candidate list,
• wherein the two uni-prediction merge candidate are selected from the merge candidate list by considering candidate types of candidates in the merge candidate list;
• wherein one of the two uni-prediction merge candidates is a sub-block based temporal motion vector prediction candidate, and the other is an affine merge candidate, or the two uni-prediction merge candidates are both affine merge candidates,
• wherein the bitstream includes an indication that multiple hypothesis mode is used for the at least one sub-block, and
• wherein the indication is based on a flag that is signaled dependent on a height or a width of the video block as one of the following:
• i. the flag is not signaled if w>32 and h>32;
• ii. the flag is not signaled if w>32 or h>32;
• iii. the flag is not signaled if w*h>1024;
• iv. the flag is not signaled if w<16 and h<16;
• v. the flag is not signaled if w<16 or h<16;
• vi. The flag is not signaled if w*h<256,
• where w and h are integers, w indicates the width of the video block, and h indicates the height of the video block.

A method of storing a bitstream of a video to a computer-readable recording medium, comprising:

• dividing a current video block of the video into multiple sub-blocks;
• generating a merge candidate list for at least one sub-block of the multiple sub-blocks;
• and
• performing a conversion between the video block and a bitstream of the video block based on the merge candidate list, and
• storing the bitstream to the computer-readable medium;
• wherein the merge candidate list comprises at least one merge candidate with multi-hypothesis mode;
• the at least one merge candidate with multi-hypothesis mode includes a merge candidate calculated by jointly utilizing two uni-prediction merge candidates in the merge candidate list,
• wherein the two uni-prediction merge candidate are selected from the merge candidate list by considering candidate types of candidates in the merge candidate list;
• wherein one of the two uni-prediction merge candidates is a sub-block based temporal motion vector prediction candidate, and the other is an affine merge candidate, or
• the two uni-prediction merge candidates are both affine merge candidates,
• wherein the bitstream includes an indication that multiple hypothesis mode is used for the at least one sub-block, and
• wherein the indication is based on a flag that is signaled dependent on a height or a width of the video block as one of the following:
• i. the flag is not signaled if w>32 and h>32;
• ii. the flag is not signaled if w>32 or h>32;
• iii. the flag is not signaled if w*h>1024;
• iv. the flag is not signaled if w<16 and h<16;
• v. the flag is not signaled if w<16 or h<16;
• vi. The flag is not signaled if w*h<256,
• where w and h are integers, w indicates the width of the video block, and h indicates the height of the video block.