Pairwise Average Motion Vector Prediction for Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face inefficiencies in motion vector prediction, particularly in representing less likely intra prediction directions and combining motion vectors from various sources, which affects compression ratios and decoding efficiency.
Innovation Solution
The introduction of pairwise average motion vector prediction methods, which generate new motion vector predictors by averaging existing motion vector predictors and temporal or history-based predictors, and the use of a history-based motion vector prediction (HMVP) method to improve the construction of motion vector predictor candidate lists, enhancing the representation of motion information across reference pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional motion vector prediction methods are used, then the coding process is simple, but the compression ratio is insufficient and motion information representation is poor
Solution Approach 1:
The patent combines multiple motion vector predictors (spatial neighbors A0, A1, B0, B1, B2 and temporal predictor C0) into a unified candidate list, merging different prediction sources to improve motion information representation. The merge mode combines spatial and temporal predictors while the AMVP mode uses selected candidates from this merged list, resolving the contradiction between information completeness and processing complexity.
Solution Approach 2:
The patent segments the motion vector prediction process into distinct modes (merge mode and AMVP mode) with different candidate selection strategies. The merge mode uses a fixed set of spatial and temporal candidates, while AMVP mode selectively uses candidates from the same pool, allowing the system to segment the complexity based on prediction needs while maintaining comprehensive motion information representation.
2Productivity
If more motion vector predictors are added to the candidate list, then the coding efficiency improves, but the computational complexity increases
Solution Approach 1:
The patent creates a universal candidate list that serves both merge mode and AMVP mode prediction. The same set of spatial and temporal predictors (A0, A1, B0, B1, B2, C0) is constructed once and then used by both prediction modes, eliminating redundant computations. This universal candidate pool improves coding efficiency for both modes while avoiding the complexity of constructing separate candidate lists for each mode.
Data Source
AI summary
An apparatus for video decoding includes processing circuitry that decodes prediction information for a current block in a current coded picture. The prediction information indicates a prediction mode that uses a motion vector predictor candidate list for reconstructing the current block. The processing circuitry includes a new motion vector predictor into the motion vector predictor candidate list. The new motion vector predictor is one of a first motion vector predictor and a second motion vector predictor. The first motion vector predictor is a first average result of a pair of existing motion vector predictors in the motion vector predictor candidate list and a temporal motion vector predictor (TMVP). The second motion vector predictor is a second average result of a pair of history-based motion vector predictors (HMVPs) in an HMVP buffer. The processing circuitry reconstructs the current block based on the motion vector predictor candidate list.