Gradient calculation in different motion vector refinements

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 first block of a video which is coded with a non-affine mode, a first final prediction block for the first video block by refining one or more first initial predictions for the first video block using a first optical flow calculation procedure, wherein each first initial prediction is refined based on at least one gradient of the each first initial prediction;
• perform a first conversion between the first video block and a bitstream of the video using the first final prediction block;
• determine, for a second block of the video which is coded with an affine mode, a second final prediction block for the second video block by refining one or more second initial predictions for the second video block using a second optical flow calculation procedure, wherein each second initial prediction is refined based on at least one gradient of the each second initial prediction;
• and
• perform a second conversion between the second video block and the bitstream of the video using the second final prediction block,
• wherein a same padding operation is used for a reference block for calculating the gradient for the first optical flow calculation procedure and a reference block for calculating the gradient for the second optical flow calculation procedure, and
• wherein the reference block for calculating the gradient for the first optical flow calculation procedure and the reference block for calculating the gradient for the second optical flow calculation procedure are padded with integer samples.

A non-transitory computer-readable storage medium storing instructions that cause a processor to:

• determine, for a first block of a video which is coded with a non-affine mode, a first final prediction block for the first video block by refining one or more first initial predictions for the first video block using a first optical flow calculation procedure, wherein each first initial prediction is refined based on at least one gradient of the each first initial prediction;
• perform a first conversion between the first video block and a bitstream of the video using the first final prediction block;
• determine, for a second block of the video which is coded with an affine mode, a second final prediction block for the second video block by refining one or more second initial predictions for the second video block using a second optical flow calculation procedure, wherein each second initial prediction is refined based on at least one gradient of the each second initial prediction;
• and
• perform a second conversion between the second video block and the bitstream of the video using the second final prediction block,
• wherein a same padding operation is used for a reference block for calculating the gradient for the first optical flow calculation procedure and a reference block for calculating the gradient for the second optical flow calculation procedure, and
• wherein the reference block for calculating the gradient for the first optical flow calculation procedure and the reference block for calculating the gradient for the second optical flow calculation procedure are padded with integer samples.

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 first block of a video which is coded with a non-affine mode, a first final prediction block for the first video block by refining one or more first initial predictions for the first video block using a first optical flow calculation procedure, wherein each first initial prediction is refined with at least one gradient of the each first initial prediction;
• generating the bitstream based on the first final prediction block;
• determining, for a second block of the video which is coded with an affine mode, a second final prediction block for the second video block by refining one or more second initial predictions for the second video block using a second optical flow calculation procedure, wherein each second initial prediction is refined with at least one gradient of the each second initial prediction;
• and
• generating the bitstream based on the second final prediction block,
• wherein a same padding operation is used for a reference block for calculating the gradient for the first optical flow calculation procedure and a reference block for calculating the gradient for the second optical flow calculation procedure, and
• wherein the reference block for calculating the gradient for the first optical flow calculation procedure and the reference block for calculating the gradient for the second optical flow calculation procedure are padded with integer samples.