IBC Merge Mode Using BVD Lists for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently utilizing block vector differences for improved compression efficiency, particularly in intra prediction modes, leading to suboptimal performance in video encoding and decoding processes.
Innovation Solution
The implementation of a method for video encoding and decoding that utilizes a block vector difference (BVD) list, where initial interval sizes are powers of 2, and BVD candidates are determined based on offset and direction, with a template-matching cost analysis to select optimal BVD candidates, enhancing the reconstruction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional intra prediction methods are used, then encoding simplicity is maintained, but compression efficiency is suboptimal
Solution Approach 1:
The patent segments the block vector difference search space into multiple candidate lists (first BVD candidate list and second BVD candidate list) with different generation methods. The first list uses simple geometric relationships with the current block, while the second list uses template matching. This segmentation allows the encoder to try simpler methods first and fall back to more complex methods only when necessary, improving compression efficiency while controlling encoding complexity.
Solution Approach 2:
The patent performs preliminary generation of BVD candidates using simple geometric relationships before performing template matching. The first BVD candidate list is generated using only basic block relationships, and only if these candidates are insufficient does the system perform the more computationally intensive template matching to generate the second list. This preliminary action approach ensures that simpler methods are exhausted first, maintaining encoding simplicity while improving compression efficiency.
2Measurement precision
If more BVD candidates are generated using template matching, then reconstruction accuracy is improved, but encoding time increases
Solution Approach 1:
The patent generates BVD candidates partially through template matching by limiting the search to specific candidate positions derived from geometric relationships rather than performing exhaustive template matching across the entire reference region. The first BVD candidate list uses partial template matching at predetermined positions, and the second list uses full template matching only when the first list is insufficient. This partial action approach improves reconstruction accuracy while significantly reducing encoding time compared to complete template matching.
3Measurement precision
If uniform interval sizes are used for BVD offsets, then implementation simplicity is maintained, but candidate selection accuracy is reduced
Solution Approach 1:
The patent applies different interval sizes for different candidate positions in the BVD offset search. Instead of using a uniform interval for all candidates, the system uses smaller intervals (e.g., 1, 2, or 4) for candidates closer to the current block and larger intervals (e.g., 8, 16, or 32) for candidates farther away. This local quality approach improves candidate selection accuracy by providing finer granularity where needed while maintaining coarser granularity where less precision is required, without significantly increasing implementation complexity.
Data Source
AI summary
Aspects of the disclosure include methods and an apparatus for video coding. The apparatus includes processing circuitry that receives coded information of a current block in a current picture. The current block is predicted based on a reference block in the current picture indicated by a block vector (BV) to be determined based on a BV predictor (BVP) and a BV difference (BVD). The processing circuitry determines a BVD list including BVD candidates based at least on BVD offsets from the BVP, determines the BVD from the BVD candidates in the BVD list, and reconstructs the current block using the BVD. For each adjacent pair of the BVD offsets, an initial interval size indicates a difference between the adjacent pair of the BVD offsets. Each of the initial interval sizes is different from other initial interval sizes corresponding to other adjacent pairs of the BVD offsets.