Joint spatial and temporal block merge mode for HEVC
View Patent ↗In one embodiment, a spatial merge mode or a temporal merge mode for a block of video content may be used in merging motion parameters. Both spatial and temporal merge parameters are considered concurrently and do not require utilization of bits or flags or indexing to signal a decoder. If the spatial merge mode is determined, the method merges the block of video content with a spatially-located block, where merging shares motion parameters between the spatially-located block and the block of video content. If the temporal merge mode is determined, the method merges the block of video content with a temporally-located block, where merging shares motion parameters between the temporally-located block and the block of video content.
1. A method comprising:
determining a merge mode for a current block of video content, wherein the merge mode is selected from a candidate list, wherein the candidate list comprises neighboring blocks that are spatially and temporally located with the current block;
determining the merge mode by analyzing motion vector differences between the neighboring spatial blocks and temporal blocks;
selecting one or more blocks for encoding the current block;
encoding the current block using motion parameters of the selected one or more selected blocks; and
transmitting information to a decoder to indicate which motion parameters from the one or more selected blocks to use in decoding the current block;
wherein the one or more spatially-located neighboring blocks reside in the same picture as the current block;
wherein the one or more temporally-located neighboring blocks reside in a different picture from the current block;
wherein no bits, flags or indexes are used to indicate the merge mode;
wherein the candidate list comprises:
MV A , MV B , MV C , MV D assigned as motion vectors (MVs) of spatial neighbors A, B, C, and D, for a region; and
MV A′ , MV B′ , MV C′ , MV D′ and MV X′ assigned as MVs of the temporal neighbors A′, B′, C′, D′ and X′, for a region;
wherein determining the merge mode additionally comprises:
scaling the motion vectors so that all the scaled motion vectors span the same temporal distance; and
calculating spatial MV differences between MV A and MV D and between MV B and MV C as well as temporal MV difference between MV A and MV A′ and between MV B and MV B′ as
{
Δ
MV
AD
=
MV
A
-
MV
D
Δ
MV
BC
=
MV
B
-
MV
C
Δ
MV
AA
′
=
MV
A
-
MV
A
′
Δ
MV
BB
′
=
MV
B
-
MV
B
′
;
and
wherein determining the merge mode additionally comprises:
defining the average norm of temporal MV differences as ∥ΔMV Temporal ∥=½(∥ΔMV AA′ ∥+∥ΔMV BB′ ∥); and
comparing the average norm of temporal MV difference, ∥ΔMV Temporal ∥, with the norms of the two spatial MV differences, ΔMV AD and ΔMV BC .
2. The method of claim 1 , wherein the candidate list comprises one or more of spatially-located neighboring blocks, one or more of temporally-located neighboring blocks, or a combination thereof.
3. The method of claim 1 wherein the candidate list additionally comprises:
refIdx A , refIdx B , refIdx C , refIdx D , refIdx A′ , refIdx B′ , refIdx C′ , refIdx D′ , and refIdx X′ , assigned as reference pictures pointed by MV A , MV B , MV C , MV D , MV A′ , MV B′ , MV C′ , MV D′ , and MV X′ , individually.