Overlapped decoder side motion refinement

A method for encoding or decoding a block within a picture in a video sequence, the method comprising:

• for a subblock of the block, determining an initial motion vector (MV) and MV candidates;
• for each of the MV candidates:
• using the MV candidate to derive a first reference block and a second reference block, wherein:
• a size of the first and second reference blocks is not equal to a size of the subblock of the block;
• using the MV candidate to derive the first and second reference blocks comprises, for each of the first and second reference blocks, deriving first and second groups of reference samples based on the MV candidate;
• the first group of reference samples is a block of reference samples having a width and a height equal to a width and a height, respectively, of the subblock of the block;
• and
• the second group of reference samples includes the reference samples of the reference block that are not in the first group of reference samples;
• and
• deriving a cost value based on samples from the first and second reference blocks, wherein the cost value is derived as a sum of difference values between reference samples of the first and second reference blocks, difference values between reference samples of the first group of reference samples of the first reference block and reference samples of the first group of reference samples of the second reference block are weighted with a first weighting factor, difference values between reference samples of the second group of reference samples of the first reference block and reference samples of the second group of reference samples of the second reference block are weighted with a second weighting factor, and the second weighting factor is larger than the first weighting factor;
• using the cost values from the MV candidates to derive an MV candidate having a smallest cost value among the MV candidates;
• using the MV candidate having the smallest cost value to derive prediction samples for the subblock of the block;
• and
• splitting the block into n subblocks, wherein n is greater than or equal to 1, and an accumulated number of samples of the n subblocks is equal to the number of samples of the block.

An apparatus comprising processing circuitry and a storage unit, wherein the apparatus is configured to encode or decode a block within a picture in a video sequence, and the apparatus is configured to, in encoding or decoding the block:

• for a subblock of the block, determine an initial motion vector (MV) and MV candidates;
• for each of the MV candidates:
• use the MV candidate to derive a first reference block and a second reference block, wherein a size of the first and second reference blocks is not equal to a size of the subblock of the block, and the first reference block for the subblock of the block touches but does not overlap a first reference block for an adjacent second subblock of the block;
• and
• derive a cost value based on samples from the first and second reference blocks;
• use the cost values from the MV candidates to derive an MV candidate having a smallest cost value among the MV candidates;
• use the MV candidate having the smallest cost value to derive prediction samples for the subblock of the block;
• and
• split the block into n subblocks, wherein n is greater than or equal to 1, and an accumulated number of samples of the n subblocks is equal to the number of samples of the block.