Delta motion vector in prediction refinement with optical flow process
A video processing method is disclosed to include: generating one or more initial predictions for a sub-block of a current video block of the video; refining the one or more initial predictions to generate a final prediction block using a prediction refinement with optical flow (PROF) procedure by calculating motion vector differences based on variables having a bit width not exceeding a predetermined maximum bit width; and performing a conversion between the current video block and a coded representation of the video using the variables having the bit width not exceeding the predetermined maximum bit width.
1. A video processing method, comprising:
determining, for a current video block, an applicability of an optical flow operation with an affine mode according to a first rule that specifies a condition for applying the optical flow operation;
generating initial prediction samples of a sub-block of the current video block coded with the affine mode;
applying the optical flow operation to generate final prediction samples for the sub-block by deriving a prediction refinement based on motion vector differences dMvH and/or dMvV, wherein dMvH and dMvV indicate motion vector differences along a horizontal direction and a vertical direction; and
performing a conversion between the current video block and a bitstream of a video based on the final prediction samples,
wherein a clipping operation is performed on the dMvH and/or dMvV before the deriving of the prediction refinement according to a second rule,
wherein the second rule specifies to perform the clipping operation before the prediction refinement being acquired and wherein the clipping operation is performed based on a precision of the dMvH and/or dMvV, and
wherein the dMvH and/or dMvV is clipped as follows:
d Mv H [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv H [pos X ][pos Y ]);
d Mv V [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv V [pos X ][pos Y ]).
2. The method of claim 1 , wherein K=6.
3. The method of claim 1 , wherein the initial prediction samples are derived based on an initial motion vector, and wherein the initial motion vector is derived based on a dimension of the current video block and a position of the sub-block.
4. The method of claim 1 , wherein the motion vector differences are calculated based on variables having a bit width not exceeding a predetermined maximum bit width.
5. The method of claim 1 , wherein the prediction refinement is derived based on the motion vector differences and gradient values.
6. The method of claim 5 , wherein the gradient values are calculated based on a difference between multiple initial prediction samples along the horizontal direction and the vertical direction.
7. The method of claim 1 , wherein the performing of the conversion includes decoding the current video block from the bitstream.
8. The method of claim 1 , wherein the performing of the conversion includes encoding the current video block into the bitstream.
9. 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:
determine, for a current video block, an applicability of an optical flow operation with an affine mode according to a first rule that specifies a condition for applying the optical flow operation;
generate initial prediction samples of a sub-block of the current video block coded with the affine mode;
apply the optical flow operation to generate final prediction samples for the sub-block by deriving a prediction refinement based on motion vector differences dMvH and/or dMvV, wherein dMvH and dMvV indicate motion vector differences along a horizontal direction and a vertical direction; and
perform a conversion between the current video block and a bitstream of a video based on the final prediction samples,
wherein a clipping operation is performed on the dMvH and/or dMvV before the deriving of the prediction refinement according to a second rule,
wherein the second rule specifies to perform the clipping operation before the prediction refinement being acquired and wherein the clipping operation is performed based on a precision of the dMvH and/or dMvV, and
wherein the dMvH and/or dMvV is clipped as follows:
d Mv H [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv H [pos X ][pos Y ]);
d Mv V [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv V [pos X ][pos Y ]).
10. The apparatus of claim 9 , wherein K=6.
11. The apparatus of claim 9 , wherein the initial prediction samples are derived based on an initial motion vector, and wherein the initial motion vector is derived based on a dimension of the current video block and a position of the sub-block.
12. The apparatus of claim 9 , wherein the motion vector differences are calculated based on variables having a bit width not exceeding a predetermined maximum bit width.
13. The apparatus of claim 9 , wherein the prediction refinement is derived based on the motion vector differences and gradient values.
14. The apparatus of claim 13 , wherein the gradient values are calculated based on a difference between multiple initial prediction samples along the horizontal direction and the vertical direction.
15. A non-transitory computer-readable storage medium storing instructions that cause a processor to:
determine, for a current video block, an applicability of an optical flow operation with an affine mode according to a first rule that specifies a condition for applying the optical flow operation;
generate initial prediction samples of a sub-block of the current video block coded with the affine mode;
apply the optical flow operation to generate final prediction samples for the sub-block by deriving a prediction refinement based on motion vector differences dMvH and/or dMvV, wherein dMvH and dMvV indicate motion vector differences along a horizontal direction and a vertical direction; and
perform a conversion between the current video block and a bitstream of a video based on the final prediction samples,
wherein a clipping operation is performed on the dMvH and/or dMvV before the deriving of the prediction refinement according to a second rule,
wherein the second rule specifies to perform the clipping operation before the prediction refinement being acquired and wherein the clipping operation is performed based on a precision of the dMvH and/or dMvV, and
wherein the dMvH and/or dMvV is clipped as follows:
d Mv H [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv H [pos X ][pos Y ]);
d Mv V [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv V [pos X ][pos Y ]).
16. 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:
determining, for a current video block, an applicability of an optical flow operation with an affine mode according to a first rule that specifies a condition for applying the optical flow operation;
generating initial prediction samples of a sub-block of the current video block coded with the affine mode;
applying the optical flow operation to generate final prediction samples for the sub-block by deriving a prediction refinement based on motion vector differences dMvH and/or dMvV, wherein dMvH and dMvV indicate motion vector differences along a horizontal direction and a vertical direction; and
generating the bitstream of the video based on the final prediction samples,
wherein a clipping operation is performed on the dMvH and/or dMvV before the deriving of the prediction refinement according to a second rule,
wherein the second rule specifies to perform the clipping operation before the prediction refinement being acquired and wherein the clipping operation is performed based on a precision of the dMvH and/or dMvV, and
wherein the dMvH and/or dMvV is clipped as follows:
d Mv H [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv H [pos X ][pos Y ]);
d Mv V [pos X ][pos Y ]=Clip3(−2 K−1 ,2 K−1 −1 ,d Mv V [pos X ][pos Y ]).