Motion Vector Encoding Using Differential Selection

The patent extracts only the essential differential motion vector information from the complete motion vector, encoding only what is necessary for reconstruction. By separating the predicted motion vector (which can be independently determined at the decoder) from the differential component (which carries the unique information), the method transmits only the minimal required data, reducing bit rate while maintaining reconstruction accuracy.

The patent performs preliminary determination of the predicted motion vector at the encoder side using prearranged criteria, so that the decoder can independently reconstruct the same predicted motion vector without receiving it explicitly. This preliminary action eliminates the need to transmit redundant predicted motion vector information, improving encoding efficiency.

The patent generates multiple predicted motion vector candidates but applies a selective evaluation approach, examining only the most promising candidates based on preliminary criteria rather than exhaustively evaluating all possibilities. This partial action approach achieves high prediction accuracy while avoiding the full computational burden of evaluating every candidate in detail.

The patent segments the predicted motion vector selection process into multiple stages: generating candidates from different sources (spatial neighbors, temporal references), evaluating them against prearranged criteria, and selecting the best match. This segmentation allows complex multi-candidate evaluation to be broken down into manageable steps, reducing overall processing complexity.