Video Bitstream Decoding With Partition-Aware Merge Candidate Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video signal processing methods require inefficient transmission of prediction information, lack accurate prediction modes, and struggle with selecting appropriate merging candidate units for enhanced coding efficiency.
Innovation Solution
The method involves recursively partitioning coding units into smaller units, using a merging mode to share prediction information between adjacent units, and adaptively determining units to be merged based on position, size, and motion information, reducing the need for transmitting redundant data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion information is transmitted for each coding unit individually, then prediction accuracy is maintained, but transmission complexity and bit rate increase
Solution Approach 1:
The patent merges motion information from neighboring coding units to form prediction candidates for the current coding unit. Instead of transmitting independent motion information for each unit, the system combines available motion vectors from adjacent blocks (left, upper, upper-right neighbors) to create a set of prediction candidates, thereby reducing transmission complexity while maintaining prediction accuracy through selective candidate evaluation.
2Productivity
If motion information is shared between adjacent units, then transmission efficiency improves, but prediction accuracy may deteriorate
Solution Approach 1:
The patent applies partial action by selectively using only the most relevant motion information from neighboring units. Instead of fully sharing all motion data, the system evaluates multiple prediction candidates derived from neighbors and selects the best match for the current coding unit. This partial utilization of neighboring motion information maintains transmission efficiency while preserving prediction accuracy through intelligent selection.
3Measurement precision
If multiple prediction candidates are evaluated, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the motion information acquisition process into distinct stages: first collecting motion vectors from specific neighboring units (left, upper, upper-right), then forming prediction candidates from these segmented sources, and finally evaluating only the necessary candidates. This segmentation of the prediction process reduces computational complexity by avoiding exhaustive search while maintaining accuracy through structured candidate evaluation.
4Reliability
If redundant prediction information is transmitted, then decoding accuracy is ensured, but bit rate increases
Solution Approach 1:
The patent uses copying by deriving motion information for the current coding unit from copies of motion vectors already available in neighboring units. Instead of transmitting redundant motion information for each block, the system copies and reuses motion vectors from adjacent blocks to form prediction candidates, thereby ensuring decoding accuracy through reliable motion compensation while significantly reducing bit rate by eliminating redundant transmissions.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method for decoding a bitstream for a video signal that is performed by a decoding apparatus comprises obtaining a plurality of prediction blocks from a coding block, obtaining merge flag information of a current prediction block of the plurality of prediction blocks from the bitstream, the merge flag information indicating whether the current prediction block is coded in a merge mode, obtaining merge index information of the current prediction block from the bitstream, the merge index information specifying a merging candidate to be merged with the current prediction block from among merging candidates including spatial and temporal neighboring blocks of the current prediction block, deriving prediction information of the current prediction block based on the merging candidates and the merge index information, the prediction information including motion vector information and reference index information and performing an inter prediction for the current prediction block based on the prediction information of the current prediction block. Based on the coding block 2Nx2N being partitioned into a left Nx2N prediction block and a right Nx2N prediction block and the current prediction block being the right Nx2N prediction block of the coding block, the left Nx2N prediction block of the coding block is excluded from the merging candidates. Based on the coding block 2Nx2N being partitioned into an upper 2NxN prediction block and a lower 2NxN prediction block and the current prediction block being the lower 2NxN prediction block of the coding block, the upper 2NxN prediction block of the coding block is excluded from the merging candidates.