Scalable Video Coding Motion Prediction Using Reference Layer Blocks

The patent extends motion prediction from two-dimensional spatial neighbors to three-dimensional spatio-temporal prediction by incorporating motion information from the reference layer (base layer) at the corresponding block position. This adds a temporal dimension to the prediction approach, allowing the system to utilize motion data from different layers to improve both accuracy and reliability.

The patent implements a dynamic motion prediction mechanism that selects between intra-layer neighboring block prediction and inter-layer corresponding block prediction based on availability and effectiveness. The system adaptively switches between prediction modes depending on whether neighboring blocks are available and which mode provides better prediction results, thereby managing complexity through adaptive selection rather than always using the most complex approach.

The patent applies partial action by selectively applying motion prediction only when necessary. The system determines whether to perform motion prediction based on whether motion information is actually needed for the current block (e.g., when inter prediction is enabled and motion vectors are available). This avoids unnecessary processing time while maintaining prediction completeness for cases where motion information is required.