Intra Prediction Mode Selection for Video Compression

The patent segments the prediction mode selection process into two distinct phases: a rapid initial selection phase that narrows down candidate modes, and a refined optimization phase that evaluates only the selected candidates. This segmentation avoids exhaustive evaluation of all prediction modes while maintaining accurate mode selection, thereby reducing encoding process complexity without sacrificing prediction accuracy.

Instead of performing complete block predictions for all prediction modes, the patent applies partial action by conducting simplified preliminary evaluations to identify promising candidate modes, then performing full predictions only on these selected candidates. This partial approach significantly reduces the number of expensive block prediction operations while still achieving optimal or near-optimal mode selection.

The patent extracts and encodes only the essential prediction mode information for selected candidate modes rather than encoding complete mode selection data for all possible modes. By taking out only the necessary mode identifiers for the reduced set of candidate modes, the patent maintains sufficient information for accurate decoding while significantly reducing the bit quantity required.

The patent applies local quality by encoding prediction mode information selectively based on the actual candidates selected for each block, rather than uniformly encoding all possible modes. This localized approach encodes information only where needed (for selected candidates) and omits redundant information, thereby maintaining decoding accuracy while reducing overall bit consumption.