Video Decoding Motion Vector Prediction via Candidate List Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently reducing redundancy in video data, particularly in motion vectors, which affects compression ratios and bandwidth requirements, especially in applications requiring high compression without significant distortion.
Innovation Solution
The method involves constructing a candidate list of motion vector predictors for video decoding, including corner and side neighbors, and segmenting blocks based on size requirements to enhance motion vector prediction, thereby improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If motion vectors are coded directly without prediction, then motion vector accuracy is maintained, but data requirements and bandwidth increase
Solution Approach 1:
The patent applies preliminary action by constructing a candidate list of motion vector predictors before the actual motion vector coding process. The decoder receives prediction information indicating a merge submode, then constructs a candidate list including corner and side neighbors. This preliminary construction enables efficient motion vector prediction without transmitting full motion vector data, thereby reducing data requirements while maintaining accuracy.
2Productivity
If block segmentation is performed to increase prediction candidates, then compression efficiency improves, but processing complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the current block into sub-blocks and constructing candidate lists for each sub-block independently. The method segments blocks based on size requirements, including corner neighbors and side neighbors for each segment. This segmentation enables more precise motion vector prediction tailored to local motion characteristics, improving compression efficiency while managing processing complexity through systematic subdivision.
Solution Approach 2:
The patent applies local quality by constructing different candidate lists for different sub-blocks based on their specific spatial positions and motion characteristics. Each sub-block receives customized prediction candidates from its local neighborhood (corner and side neighbors), rather than applying a uniform prediction approach across the entire block. This local adaptation improves prediction accuracy for regions with different motion patterns.
Data Source
AI summary
Aspects of the disclosure provide methods and apparatuses for video decoding. In some embodiments, an apparatus for video decoding includes processing circuitry. The processing circuitry decodes prediction information for a block in a current coded picture that is a part of a coded video sequence. The prediction information is indicative of a merge submode. Then, the processing circuitry constructs, in response to the merge submode, a candidate list of candidate motion vector predictors for the block. The candidate list includes one or more first candidates that are corner neighbors of the block and at least one second candidate that is a side neighbor of the block. Then, the processing circuitry reconstructs the block according to motion information associated with the second candidate.