Joint Motion Vector Difference Coding With Adaptive Resolution

The patent combines the coding of multiple motion vector differences into a joint coding process. Instead of encoding each motion vector difference separately, the patent merges them into a unified coding structure where shared information is extracted and encoded once, eliminating redundancy while maintaining decoding simplicity through standardized syntax elements.

The patent applies adaptive resolution to motion vector difference coding by selecting different precision levels for different regions or contexts. The resolution of motion vector differences is adapted based on local characteristics such as block size, motion complexity, or reference picture distance, allowing higher precision where needed and lower precision where acceptable, thus optimizing the balance between compression efficiency and complexity.

The patent introduces dynamic adaptation of motion vector difference resolution through context-dependent signaling. The precision level is not fixed but dynamically selected based on coding conditions such as block size, motion characteristics, and reference picture parameters. This dynamic approach allows the system to adapt resolution to local requirements, improving compression efficiency while managing complexity through conditional logic rather than universal high complexity.

The patent changes the precision parameter of motion vector differences adaptively. Instead of using a fixed resolution for all motion vector differences, the patent modifies the precision parameter (e.g., quarter-pixel, eighth-pixel, or pixel-level precision) based on coding conditions. This parameter adaptation allows the system to allocate bits more efficiently, using higher precision only when necessary and lower precision when sufficient, thereby improving overall compression efficiency.

The patent segments the joint coding process into distinct functional components: motion vector difference extraction, joint encoding of shared information, and separate reconstruction of individual motion vectors. This segmentation allows the complex joint coding to be broken down into manageable steps with well-defined interfaces, reducing decoding complexity while maintaining compression efficiency. Each segment can be independently optimized and verified.

The patent introduces intermediate representations in the joint coding process. Instead of directly encoding final motion vector values, the patent uses intermediate syntax elements that capture shared information in a compact form. These intermediates serve as a bridge between the compressed bitstream and the final decoded motion vectors, allowing efficient transmission while preserving the necessary precision for accurate reconstruction through standardized decoding procedures.