Methods for encoding/decoding of video using common merging candidate set of asymmetric partitions
The present invention provides video encoding and decoding methods using block merging, which can be applied to a high resolution video of more than HD (High Definition) resolution. A video encoding method includes partitioning a current block into a first and a second prediction unit by using asymmetric partitioning, constructing a list of common merging candidate blocks including a predetermined number of common merging candidate blocks selected from adjacent blocks of the current block, selecting at least one candidate block from among the list of common merging candidate blocks and sending information of the selected candidate block to a decoder for each of the first and the second prediction unit.
1. A video decoding method, comprising:
constructing a list of common merging candidate blocks by including a predetermined number of the common merging candidate blocks selected from among adjacent blocks of a current coding unit, the current coding unit partitioned into a first prediction unit and a second prediction unit;
generating a motion vector by reconstructing a motion parameter of a block-merged block by using at least one common merging candidate block selected from the list of common merging candidate blocks for each of the first prediction unit and the second prediction unit; and
performing motion compensation by using the generated motion vector and a reference picture,
wherein the constructing of the list of common merging candidate blocks is performed only when a size of the current coding unit is 8-by-8.
2. The video decoding method of claim 1 , wherein the selected at least one common merging candidate block for the second prediction unit comprise adjacent blocks which are constructed even before reconstructing a motion parameter of the first prediction unit.
3. The video decoding method of claim 1 , wherein a largest coding unit (LCU) is divided into a plurality of non-overlapping motion estimation regions,
motion estimation is performed sequentially on the motion estimation regions within the largest coding unit (LCU), and
the motion estimation is performed in a parallel fashion for the first prediction unit and the second prediction unit belonging to a motion estimation region within the largest coding unit (LCU).
4. The video decoding method of claim 3 , wherein whether to allow a parallel merging for the first prediction unit and the second prediction unit within the same motion estimation region by using the common merging candidate blocks is determined according to a size of the motion estimation region.
5. The video decoding method of claim 4 , wherein the parallel merging for the first prediction unit and the second prediction unit within the same motion estimation region is allowed only when the size of the motion estimation region is larger than a predetermined size.
6. The video decoding method of claim 4 , wherein
in case the current prediction unit and another prediction unit adjacent to the current prediction unit belong to the same motion estimation region, said another prediction unit is denoted as non-available, and
in case the current prediction unit and said another prediction unit adjacent to the current prediction unit belong to motion estimation regions different from each other, said another prediction unit is denoted as available.
7. The video decoding method of claim 1 , wherein the list of common merging candidate blocks includes a spatial merging candidate block and a temporal merging candidate block.
8. The video decoding method of claim 1 , wherein the first prediction unit and the second prediction unit within the current coding unit utilize a reference picture index of a block, at a predetermined particular position, from among spatial common merging candidate blocks as a reference picture index for temporal motion vector prediction (MVP) of a temporal merging candidate block.
9. The video decoding method of claim 8 , wherein the block, at the predetermined particular position, uses a previously encoded adjacent block which is constructed before reconstructing a motion parameter of a first prediction unit and a second prediction unit from among blocks included in the list of common merging candidate blocks.