Adaptive Search Region Setting for Motion Vector Prediction

The patent divides the search region into multiple sub-regions and performs motion vector prediction in a segmented manner. Instead of searching the entire large search region at once, the method segments it into smaller areas and processes them separately, reducing the computational complexity while maintaining prediction accuracy through hierarchical or progressive search strategies.

The patent dynamically adjusts the search region size and shape based on local characteristics of the video content, such as motion patterns, block type, and prediction mode. By making the search region adaptive rather than fixed, the method reduces unnecessary computations in areas with small motion while maintaining adequate search coverage in areas with large motion, thus improving computational efficiency without sacrificing accuracy.

The patent implements dynamic adjustment of search region parameters based on block characteristics. The search region size and shape are adaptively determined according to factors such as block type (intra/inter prediction), motion complexity, and local motion patterns. This allows the system to reduce search region size for blocks with small or simple motion, significantly reducing computation time while maintaining implementation feasibility through standardized adjustment rules.

The patent applies different search region configurations to different blocks based on their local characteristics. Instead of using a uniform search region for all blocks, the method tailors the search region size and shape to match the specific motion properties of each block, such as applying smaller search regions to stationary or slowly moving blocks and larger search regions to blocks with significant motion, thereby optimizing computation time across the entire video sequence.

The patent dynamically adapts the search region size based on local motion characteristics to maintain prediction accuracy while improving computational efficiency. By analyzing motion patterns and adjusting the search region accordingly, the method ensures that the search region is sufficiently large to capture the actual motion when needed, while being compact enough to reduce computations when motion is small, thus resolving the trade-off between accuracy and efficiency.

The patent changes key parameters of the search region (such as width, height, and center position) based on block characteristics and motion patterns. By dynamically adjusting these parameters rather than using fixed values, the method optimizes the balance between search region size and prediction accuracy, ensuring that the search region is appropriately sized for each specific case, thereby maintaining accuracy while improving computational efficiency.