Motion Vector Predictor Precision for Lower-Bitrate Video Coding

The patent changes the parameter representation from explicit precision indication to implicit precision through differential precision coding. The motion vector difference components are coded with different precisions based on their magnitude, allowing the system to achieve high prediction accuracy for small differences while using fewer bits for large differences, thus resolving the contradiction between precision and data requirements

The patent applies partial precision to different components of the motion vector difference. Instead of using uniform precision for all components, the system uses higher precision only when needed (for smaller difference values) and lower precision when sufficient (for larger difference values), optimizing the balance between accuracy and data efficiency

The patent applies local quality by assigning different precision levels to different motion vector difference components based on their specific characteristics. Each component is coded with the appropriate precision tailored to its magnitude and importance, rather than using a uniform precision level, thereby improving compression efficiency without requiring complex explicit precision signaling

The system determines the appropriate precision for each motion vector difference component automatically based on its magnitude, without requiring external precision indication signals. The coding process self-adjusts the precision level, simplifying the implementation while optimizing compression efficiency through adaptive differential precision