Motion Information Encoding Using Neighboring Block Reference Picture Selection

The patent applies parameter changes by introducing a change flag that indicates whether motion information should be predicted from temporal or spatial neighboring blocks. This binary parameter allows the system to adapt the prediction mode based on block characteristics, achieving accurate prediction while reducing bit usage by only signaling the change flag rather than transmitting full motion vectors for all blocks.

The patent segments the motion prediction process into distinct modes: temporal prediction for certain block sizes and spatial prediction for others. By dividing the prediction process into these segmented modes based on block size, the system achieves encoding efficiency through mode-specific optimization while keeping decoding complexity manageable through clear mode differentiation and straightforward selection logic.

The patent applies local quality by performing interpolation only when necessary, specifically when predicting motion information for blocks that require sub-pel accuracy. Rather than universally applying interpolation to all blocks, the system selectively applies the computationally intensive interpolation process only to those local regions where prediction accuracy demands it, thereby balancing precision improvements with computational complexity.