Bilateral Matching for Chained Motion Vector Prediction in Video Coding

The patent segments the CMVP process by dividing candidate selection into multiple stages: initial candidate generation from neighboring blocks, bilateral matching condition evaluation, and selective insertion into candidate lists. This segmentation allows the system to process only relevant candidates through complex bilateral matching operations, reducing overall processing complexity while maintaining prediction accuracy.

The patent applies local quality by evaluating bilateral matching conditions specifically for certain candidate positions and reference picture configurations rather than uniformly processing all candidates. This selective approach focuses computational resources on critical prediction paths, improving accuracy where it matters most while reducing unnecessary processing elsewhere.

The patent extracts and evaluates only the essential bilateral matching conditions for candidate selection, separating the critical accuracy-determining operations from less important processing steps. By taking out only the necessary bilateral matching evaluations, the system achieves improved prediction accuracy while minimizing the power consumption associated with complex operations.

The patent performs preliminary evaluation of bilateral matching conditions before final candidate insertion, pre-identifying promising candidates that meet the conditions. This preliminary action allows the system to avoid costly bilateral matching operations for candidates that would not be selected anyway, thereby reducing power consumption while maintaining accuracy.

The patent applies partial action by evaluating bilateral matching conditions for only a subset of CMVP candidates rather than all candidates. Specifically, it focuses on candidates from certain neighboring block positions and reference picture lists that are most likely to provide accurate predictions, achieving good coding efficiency without the time cost of exhaustive evaluation.

The patent performs preliminary filtering of CMVP candidates based on basic availability and position criteria before applying the time-consuming bilateral matching condition evaluation. This preliminary action reduces the number of candidates requiring full bilateral matching evaluation, thereby reducing processing time while maintaining coding efficiency through selective evaluation of the most promising candidates.