Auto-Relocated Block Vectors for DIMD and SGPM Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in efficiently selecting block vectors for decoder side intra mode derivation and spatial geometric partitioning mode, leading to suboptimal compression and decoding performance.
Innovation Solution
A device utilizing the auto-relocated block vector (ARBV) process to enhance block vector selection by cascading block vectors and incorporating them into a merge list, followed by template-based ranking to select optimal block vectors for coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If block vectors are selected using conventional methods for decoder side intra mode derivation and spatial geometric partitioning mode, then the coding process is simple, but the compression and decoding performance is suboptimal
Solution Approach 1:
The patent performs preliminary actions by obtaining block vectors from neighboring merge blocks and creating a merged block vector list before the actual coding process. This pre-computation of candidate block vectors enables the decoder to select optimal vectors efficiently during decoding, improving performance without adding significant complexity to the main coding workflow.
Solution Approach 2:
The patent introduces an intermediary structure (the merged block vector list containing ARBV block vectors) that mediates between the available neighboring block vectors and the final selection for coding. This intermediary list allows systematic evaluation and selection of optimal block vectors, bridging the gap between simple conventional methods and optimal performance requirements.
2Reliability
If more block vectors are obtained from neighboring merge blocks and cascaded, then the choice of block vectors is improved, but the processing steps increase
Solution Approach 1:
The patent merges multiple block vector sources (neighboring merge blocks) into a single consolidated merged block vector list. This combining approach allows the system to consider a broader range of candidate block vectors for optimal selection while managing processing efficiency through structured consolidation rather than handling vectors as separate unorganized elements.
Solution Approach 2:
The patent segments the block vector selection process into distinct stages: obtaining block vectors from neighboring merge blocks, cascading them to create the merged list, and then selecting optimal vectors for coding. This segmentation allows each stage to be optimized independently, improving overall productivity by preventing any single stage from becoming a bottleneck.
3Reliability
If template based ranking is applied to the merge list, then the optimal block vectors are selected, but the ranking process adds computational steps
Solution Approach 1:
The patent implements a self-service ranking mechanism where the template-based ranking process automatically evaluates and orders block vectors in the merged list based on their suitability for coding. This self-organizing approach selects optimal block vectors without requiring external intervention or complex manual optimization, achieving high selection accuracy through an automated yet manageable ranking algorithm.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
A device (e.g., a video decoder) may utilize the auto-relocated block vector (ARBV) process to improve the choice of block vectors for DIMD and SGPM. The device may obtain a plurality of block vectors from neighboring merge blocks. The device based on a determination that a first reference block associated with a first merge block is block vector coded, may cascade a first block vector and a second block vector to create a third block vector. The first block vector is associated with the first merge block and the first block vector points to the first reference block. The second block vector is associated with the first reference block and points to a second reference block. Based on a determination that the second reference block associated with the first reference block is not coded with a block vector, add the third block vector to a merge list.