Method and apparatus of bilateral template MV refinement for video coding
Method and apparatus of using Bilateral Template MV Refinement are disclosed to improve coding efficiency or reducing complexity. According to one method, if a size associated with the current block is greater than a threshold, then Bilateral Template MV Refinement is applied to the current block. Otherwise, the Bilateral Template MV Refinement is not applied to the current block. In another method, the Bilateral Template MV Refinement is turned on or off implicitly based on the two reference blocks. According to yet another method, the Bilateral Template MV Refinement is performed on the sub-block level. According to yet another method, the Bilateral Template MV Refinement is performed motion vectors selected from candidates of AMVP (advance MV prediction) candidate list. According to yet another method, the Bilateral Template MV Refinement used a modified template to refine motion vector.
1. A method of video coding, the method comprising:
receiving input data associated with a current block in a current picture, wherein the current block is coded in a bi-prediction mode;
determining an initial MV0 (motion vector 0) pointing to a reference block 0 in a list0 reference picture and an initial MV1 (motion vector 1) pointing to a reference block 1 in a list1 reference picture, wherein the initial MV0 and the initial MV1 correspond to bi-prediction motion vectors of the current block;
determining whether a size of the current block is greater than a threshold;
in a case that the size of the current block is determined not to be greater than the threshold:
performing a bi-prediction of the current sub-block using the reference block 0 and the reference block 1; and
in a case that the size of the current block is determined to be greater than the threshold:
dividing the current block into current sub-blocks; and
for each sub-block of the current sub-blocks:
determining, according to the initial MV0 and the initial MV1, a refined sub-block motion vector 0 (sub-MV0′) pointing to a refined reference sub-block 0 in the list0 reference picture and a refined sub-block motion vector 1 (sub-MV1′) pointing to a refined reference sub-block 1 in the list reference picture, and
performing the bi-prediction of the sub-block of the current sub-blocks according to the refined reference sub-block 0 and the refined reference sub-block 1.
2. The method of claim 1 , wherein a flag is signalled in a sequence level, picture level, slice level, or PU level to indicate whether sub-block Bilateral Template MV Refinement is enabled.
3. The method of claim 1 , wherein
a same sub-block partition is used to divide each block of the current picture into corresponding sub-blocks having a sub-block width and a sub-block height, and
information associated with the sub-block width and the sub-block height is signalled in a sequence level, picture level, slice level, or PU level.
4. The method of claim 1 , wherein the determining the refined sub-block motion vector 0 (sub-MV0′) and the refined sub-block motion vector 1 (sub-MV1′) comprises:
deriving a template block according to the initial MV0 and the initial MV1; and
deriving the refined sub-block motion vector 0 and the refined sub-block motion vector 1 by performing motion estimation on the template block to identifying the refined reference sub-block 0 in the list 0 reference picture and identifying the refined reference sub-block 1 in the list 1 reference picture.
5. The method of claim 1 , wherein the determining the refined sub-block motion vector 0 (sub-MV0′) and the refined sub-block motion vector 1 (sub-MV1′) comprises:
performing a diamond search of motion vector pairs, starting with the initial MV0 and the initial MV1, based on matching a neighboring template of the corresponding current sub-block.
6. An apparatus of video coding, the apparatus comprising one or more electronic circuits or processors configured to:
receive input data associated with a current block in a current picture, wherein the current block is coded in a bi-prediction mode;
determine an initial MV0 (motion vector 0) pointing to a reference block 0 in a list0 reference picture and an initial MV1 (motion vector 1) pointing to a reference block 1 in a list1 reference picture, wherein the initial MV0 and the initial MV1 correspond to bi-prediction motion vectors of the current block;
determine whether a size of the current block is greater than a threshold;
in a case that the size of the current block is determined not to be greater than the threshold:
perform a bi-prediction of the current sub-block using the reference block 0 and the reference block 1; and
in a case that the size of the current block is determined to be greater than the threshold:
divide the current block into current sub-blocks; and
for each sub-block of the current sub-blocks:
determine, according to the initial MV0 and the initial MV1, a refined sub-block motion vector 0 (sub-MV0′) pointing to a refined reference sub-block 0 in the list0 reference picture and a refined sub-block motion vector 1 (sub-MV1′) pointing to a refined reference sub-block 1 in the list 1 reference picture, and
perform the bi-prediction of the sub-block of the current sub-blocks according to the refined reference sub-block 0 and the refined reference sub-block 1.
7. A method of video coding, the method comprising:
receiving input data associated with a current block in a current picture, wherein the current block is coded in a bi-prediction mode;
determining an initial MV0 (motion vector 0) pointing to a reference block 0 in a list0 reference picture and an initial MV1 (motion vector 1) pointing to a reference block 1 in a list1 reference picture, wherein the initial MV0 and the initial MV1 correspond to bi-prediction motion vectors of the current block;
determining a measurement value that represents a difference between the reference block 0 and the reference block 1;
determining whether the measurement value is greater than a threshold;
in a case that the measurement value is determined to be greater than the threshold:
performing a bi-prediction of the current block according to the reference block 0 and the reference block 1; and
in a case that that the measurement value is determined not to be greater than the threshold:
determining a refined motion vector 0 (MV0′) pointing to a refined reference block 0 in the list0 reference picture and a refined motion vector 1 (MV1′) pointing to a refined reference block 1 in the list reference picture that correspond to a reduced template cost than the initial MV0 and the initial MV1, and
performing the bi-prediction of the current block according to the refined reference block 0 and the refined reference block 1.