Rounding in pairwise average candidate calculations
Devices, systems, and methods for unified rounding in sub-block based prediction are described. In a representative aspect, a method of video processing includes generating, for a processing of a current block of video, a pairwise merge candidate based on a pair of motion candidates, and performing, based on the pairwise merge candidate, a conversion between the current block and a bitstream representation of the video. In another representative aspect, a method of video processing includes generating, for a current block of video coded using a geometry partition mode, a uni-prediction motion candidate based on a scaled motion vector and a List0 motion vector, and performing, based on the uni-prediction motion candidate, a conversion between the current block and a bitstream representation of the video.
1. A method of coding video data, comprising:
constructing, for a first block of a video, a merge motion candidate list which comprises at least two merge motion candidates;
deriving, based on a position index to the merge motion candidate list and a first unified rounding process, a pairwise average merging candidate from a pair of merge motion candidates included in the merge motion candidate list;
performing, based on the pairwise average merging candidate, a first conversion between the first block and a bitstream of the video;
determining, for a second conversion between a second block of the video and the bitstream of the video, that whether or not a first affine mode is allowed for the second block and whether or not a second affine mode is allowed for the second block are based on a same dimension constrain; and
performing, based on the determining, the second conversion between the second block and the bitstream,
wherein the first rounding process is defined as:
SignShift
(
m
,
s
)
=
{
(
m
+
off
)
s
m
≥
0
-
(
(
-
m
+
off
)
s
)
m
<
0
,
wherein off and s are integers, and wherein >> represents a right shift operation.
2. The method of claim 1 , wherein the pair of merge motion candidates comprises a first motion vector MV 0 =(MV 0 x , MV 0 y ) of a reference picture list L in a first merge motion candidate and a second motion vector MV 1 =(MV 1 x , MV 1 y ) of the reference picture list L in a second merge motion candidate, and wherein MV*=(MV*x, MV*y) is the pairwise merge candidate such that:
MV* x =SignShift(MV0 x +MV1 x,s ) and MV* y =SignShift(MV0 y +MV1 y,s ).
3. The method of claim 2 , wherein position indices of the pair of merge motion candidates are fixed with 0 and 1.
4. The method of claim 2 , wherein the reference picture list L comprises reference picture list 0 (L0) or reference picture list 1 (L1).
5. The method of claim 2 , further comprising inserting the pairwise average merging candidate into the merge motion candidate list.
6. The method of claim 2 , wherein off=0.
7. The method of claim 2 , wherein s=1.
8. The method of claim 1 , wherein off=0 upon a determination that s=0.
9. The method of claim 1 , wherein off=(1<<(s−1))−1.
10. The method of claim 1 , wherein s=4.
11. The method of claim 1 , wherein the first conversion comprises encoding the first block into the bitstream.
12. The method of claim 1 , wherein the first conversion comprises decoding the first block from the bitstream.
13. 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:
construct, for a first block of a video, a merge motion candidate list which comprises at least two merge motion candidates;
derive, based on a position index to the merge motion candidate list and a first unified rounding process, a pairwise average merging candidate from a pair of merge motion candidates included in the merge motion candidate list;
perform, based on the pairwise average merging candidate, a first conversion between the first block and a bitstream of the video;
determine, for a second conversion between a second block of the video and the bitstream of the video, that whether or not a first affine mode is allowed for the second block and whether or not a second affine mode is allowed for the second block are based on a same dimension constrain; and
perform, based on the determination, the second conversion between the second block and the bitstream,
wherein the first rounding process is defined as:
SignShift
(
m
,
s
)
=
{
(
m
+
off
)
s
m
≥
0
-
(
(
-
m
+
off
)
s
)
m
<
0
,
wherein off and s are integers, and wherein >> represents a right shift operation.
14. The apparatus of claim 13 , wherein the pair of merge motion candidates comprises a first motion vector MV 0 =(MV 0 x , MV 0 y ) of a reference picture list L in a first merge motion candidate and a second motion vector MV 1 =(MV 1 x , MV 1 y ) of the reference picture list L in a second merge motion candidate, and wherein MV*=(MV*x, MV*y) is the pairwise merge candidate such that:
MV* x =SignShift(MV0 x +MV1 x,s ) and MV* y =SignShift(MV0 y +MV1 y,s ).
15. The apparatus of claim 14 , wherein position indices of the pair of merge motion candidates are fixed with 0 and 1,
wherein the reference picture list L comprises reference picture list 0 (L0) or reference picture list 1 (L1),
wherein the instructions further cause the processor to insert the pairwise average merging candidate into the merge motion candidate list,
wherein off=0, and
wherein s=1.
16. The apparatus of claim 13 , wherein off=0 upon a determination that s=0, or off=(1<<(s−1))−1, and wherein s=4.
17. A non-transitory computer-readable storage medium storing instructions that cause a processor to:
construct, for a first block of a video, a merge motion candidate list which comprises at least two merge motion candidates;
derive, based on a position index to the merge motion candidate list and a first unified rounding process, a pairwise average merging candidate from a pair of merge motion candidates included in the merge motion candidate list;
perform, based on the pairwise average merging candidate, a first conversion between the first block and a bitstream of the video;
determine, for a second conversion between a second block of the video and the bitstream of the video, that whether or not a first affine mode is allowed for the second block and whether or not a second affine mode is allowed for the second block are based on a same dimension constrain; and perform, based on the determining, the second conversion between the second block and the bitstream,
wherein the first rounding process is defined as:
SignShift
(
m
,
s
)
=
{
(
m
+
off
)
s
m
≥
0
-
(
(
-
m
+
off
)
s
)
m
<
0
,
wherein off and s are integers, and wherein >> represents a right shift operation.
18. The non-transitory computer-readable storage medium of claim 17 , wherein the pair of merge motion candidates comprises a first motion vector MV 0 =(MV 0 x , MV 0 y ) of a reference picture list L in a first merge motion candidate and a second motion vector MV 1 =(MV 1 x , MV 1 y ) of the reference picture list L in a second merge motion candidate, and wherein MV*=(MV*x, MV*y) is the pairwise merge candidate such that: MV*x=SignShift (MV 0 x +MV 1 x, s ) and MV*y=SignShift (MV 0 y +MV 1 y,s ),
wherein the position indices of the pair of merge motion candidates are fixed with 0 and 1,
wherein the reference picture list L comprises reference picture list 0 (L0) or reference picture list 1 (L1),
wherein the instructions further cause the processor to insert the pairwise average merging candidate into the merge motion candidate list,
wherein off=0, or off=0 upon a determination that s=0, or off=(1<<(s−1))−1, and
wherein s=1 or 4.
19. 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:
constructing, for a first block of the video, a merge motion candidate list which comprises at least two merge motion candidates;
deriving, based on a position index to the merge motion candidate list and a first rounding process, a pairwise average merging candidate from a pair of merge motion candidates included in the merge motion candidate list;
determining, for a second block of the video, that whether or not a first affine mode is allowed for the second block and whether or not a second affine mode is allowed for the second block are based on a same dimension constrain;
generating, based on the pairwise average merging candidate and the determining, the bitstream of the video;
wherein the first rounding process is defined as:
SignShift
(
m
,
s
)
=
{
(
m
+
off
)
s
m
≥
0
-
(
(
-
m
+
off
)
s
)
m
<
0
,
wherein off and s are integers, and wherein >> represents a right shift operation.
20. The non-transitory computer-readable recording medium of claim 19 , wherein the pair of merge motion candidates comprises a first motion vector MV 0 =(MV 0 x , MV 0 y ) of a reference picture list L in a first merge motion candidate and a second motion vector MV 1 =(MV 1 x , MV 1 y ) of the reference picture list L in a second merge motion candidate, and wherein MV*=(MV*x, MV*y) is the pairwise merge candidate such that: MV*x=SignShift (MV 0 x +MV 1 x, s ) and MV*y=SignShift (MV 0 y +MV 1 y,s ),
wherein the position indices of the pair of merge motion candidates are fixed with 0 and 1,
wherein the reference picture list L comprises reference picture list 0 (L0) or reference picture list 1 (L1),
wherein the method further includes inserting the pairwise average merging candidate into the merge motion candidate list,
wherein off=0, or off=0 upon a determination that s=0, or off=(1<<(s−1))−1, and
wherein s=1 or 4.