Enhanced Motion Vector Prediction With Adaptive Candidate Ordering

The patent implements dynamic reference frame prioritization where the processing order of reference frames is adaptively adjusted based on temporal distance and quality metrics. The system dynamically determines which reference frames to process first and how many to use, optimizing motion vector prediction accuracy for each coding unit based on current video content characteristics rather than using a fixed processing order

The patent segments the motion vector prediction process into distinct stages: identifying multiple reference frames, calculating temporal distances, evaluating quality metrics, prioritizing reference frames, and generating motion vector predictors. This segmentation allows the system to selectively apply complex processing only where needed while maintaining simplicity in other areas, resolving the contradiction between coding efficiency and hardware complexity

The system dynamically selects the number of reference frames (e.g., 1, 2, or 3) to use for motion vector prediction based on video content characteristics. The processing adapts its complexity level: for simple scenes, fewer reference frames are processed; for complex scenes with significant motion, more reference frames are utilized. This dynamic adaptation improves motion vector accuracy when needed while avoiding unnecessary processing complexity for simple content

The patent changes key parameters such as temporal distance thresholds and quality metric weights adaptively based on video content. By adjusting these parameters dynamically, the system optimizes the balance between utilizing forward and backward motion information for accuracy while controlling processing complexity through content-aware parameter adaptation

The patent applies partial action by selectively processing only the most relevant reference frames based on prioritization criteria. Instead of processing all available reference frames equally, the system identifies and processes only the top N prioritized reference frames (where N is dynamically determined), achieving good prediction accuracy with reduced computational overhead by doing just enough processing rather than exhaustive processing of all reference frames

The system changes computational parameters such as the number of reference frames to process and the complexity of quality metric calculations based on content characteristics. For low-complexity modes, simplified quality metrics are used; for high-accuracy modes, more sophisticated metrics are applied. This parameter adaptation resolves the contradiction by adjusting computational overhead to match the actual prediction accuracy requirements of each scene