Motion Vector Coding Competition Among Predictors

The candidate motion vector set is segmented into multiple subsets (first subset, second subset, third subset) based on spatial and temporal relationships. This segmentation allows the encoder to select from multiple specialized subsets rather than one large comprehensive set, improving prediction accuracy while managing signaling cost through selective usage.

The invention dynamically selects which subset of candidate motion vectors to use based on the characteristics of the current block and reference blocks. The encoder can adaptively choose between different subsets (e.g., spatial neighbors, temporal neighbors, or combined) to optimize prediction accuracy for each block while controlling overall signaling cost.

The invention merges multiple sources of motion vector candidates into unified subsets: spatial neighbors (left, above blocks), temporal neighbors (reference picture blocks), and derived candidates (median, zero vectors). This combination provides richer prediction options while maintaining coding efficiency through selective subset usage rather than encoding all candidates.

The invention changes the parameter of candidate selection from a fixed single set to multiple configurable subsets with different compositions. Each subset can be tailored to specific block characteristics (e.g., inter-coded blocks use temporal subsets, intra-coded blocks use spatial subsets), optimizing prediction performance adaptively while controlling signaling overhead.