Adaptive Motion Vector Resolution in Video Coding

The patent merges the coding of x- and y-components of motion vector difference values into a single unified coding process. Instead of separately coding each component, the invention codes both components simultaneously using a single syntax element that indicates whether each component is zero, followed by joint coding of non-zero components. This merging reduces the total number of bits required while maintaining the ability to represent both components accurately.

The invention combines the entropy coding of x- and y-components into a single operation. The syntax element structure allows the entropy coder to process both components in one pass, improving throughput by eliminating the need for separate coding operations. This unified approach maintains implementation clarity while significantly improving coding speed and efficiency.

The patent applies local quality by differentiating the coding approach based on whether motion vector components are zero or non-zero. The syntax element first indicates whether each component is zero, then only codes non-zero components with their precise values. This selective coding maintains sub-pixel precision for necessary components while avoiding unnecessary complexity for zero components, achieving precision where needed without overall bitstream bloat.