Affine Merge Candidate Scanning for More Accurate Video Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as VVC and AVS3, limit the derivation of affine merge candidates to adjacent neighboring blocks, which restricts the diversity and accuracy of motion prediction, particularly in affine motion compensation.
Innovation Solution
The method extends the derivation of affine merge candidates to include non-adjacent neighboring blocks, incorporating a scanning process to identify motion vector candidates and control point motion vectors, followed by a pruning process to remove redundant candidates, thereby enhancing the diversity and accuracy of affine motion prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If affine merge candidates are derived only from adjacent neighboring blocks, then the complexity of candidate derivation is reduced, but the diversity and accuracy of motion prediction deteriorates
Solution Approach 1:
The patent extends the derivation scope from adjacent neighboring blocks to non-adjacent neighboring blocks, adding a spatial dimension to the candidate pool. This allows the system to access motion information from blocks at different distances (e.g., 1-block distance, 2-block distance) without significantly increasing computational complexity, thereby improving motion prediction accuracy for affine transformation.
Solution Approach 2:
The patent segments the neighboring block derivation process into multiple distance levels. Instead of treating all neighbors uniformly, it divides them into adjacent (distance=0) and non-adjacent (distance>0) categories, allowing selective derivation from different spatial zones to enhance candidate diversity while managing complexity.
2Measurement precision
If affine merge candidates are derived from non-adjacent neighboring blocks, then the diversity and accuracy of motion prediction is improved, but the complexity of candidate derivation increases
Solution Approach 1:
The patent introduces a scanning distance parameter to control the extent of non-adjacent block derivation. By adjusting this parameter, the system can flexibly balance between candidate diversity and computational complexity, deriving candidates from blocks at varying distances based on specific coding conditions.
Solution Approach 2:
The patent applies partial action by selectively deriving candidates from non-adjacent blocks only when beneficial, rather than universally. It uses pruning processes to remove redundant candidates, ensuring that the increased derivation scope translates to actual performance improvement without unnecessary computational overhead.
3Adaptability or versatility
If multiple motion vector candidates are obtained from non-adjacent blocks, then the diversity of affine merge candidates is improved, but the amount of data to be processed increases
Solution Approach 1:
The patent implements a pruning process that discards redundant or low-quality motion vector candidates from the expanded pool. By removing duplicates and ineffective candidates, it recovers computational resources while preserving the beneficial diversity introduced by non-adjacent block derivation, maintaining an optimal candidate list size.
Data Source
AI summary
A method of video coding, an apparatus and a non-transitory computer-readable storage medium thereof are provided. The method includes obtaining one or more motion vector candidates from a plurality of non-adjacent neighbor blocks to a current block based on at least one scanning distance, where one of the at least one scanning distance may indicate a number of blocks away from one side of the current block, and the number may be a positive integer. Furthermore, the method may include obtaining one or more control point motion vectors (CPMVs) for the current block based on the one or more motion vector candidates.