Position-Dependent Intra Prediction Filtering for Video Blocks

The patent implements dynamic selection of prediction methods based on block characteristics. The PDPC method is conditionally applied to specific blocks with dimensions greater than 4x4 and specific prediction modes, while conventional methods are used for other blocks. This dynamic adaptation resolves the contradiction by using complex methods only where necessary, improving prediction accuracy for non-square blocks and wide-angle directions without universally increasing complexity.

The patent applies different prediction qualities to different spatial regions and block types. The PDPC method with position-dependent filtering is applied locally to blocks that benefit from it (non-square blocks, wide-angle directions), while conventional methods are used elsewhere. This local quality approach improves overall prediction accuracy by enhancing specific problematic regions without unnecessarily complicating the entire prediction process.

The patent implements partial application of the PDPC method by applying it only to blocks that meet specific criteria (dimensions greater than 4x4 and specific prediction modes), rather than to all blocks. This partial action approach improves prediction signal precision for blocks that need it while avoiding the computational overhead of applying the method universally, thus maintaining coding efficiency.

The patent changes the application parameters of the PDPC method based on block characteristics. The method is selectively enabled or disabled based on block dimensions, prediction mode, and other parameters. This parameter-based control allows the system to achieve high prediction precision when needed while maintaining high coding efficiency by avoiding unnecessary computations for blocks that don't require enhanced prediction.

The patent implements dynamic filtering where the position-dependent filtering process is conditionally applied based on block characteristics. For non-square blocks and blocks with specific prediction modes, the patent uses position-dependent filtering that adapts to the block geometry, ensuring prediction signal continuity. For other blocks, conventional filtering is used to maintain processing speed. This dynamic approach resolves the contradiction by applying precise filtering only where discontinuities occur.

The patent applies different filtering qualities to different blocks based on their characteristics. Position-dependent filtering with higher precision is applied locally to non-square blocks and blocks with wide-angle prediction directions where discontinuities are problematic, while conventional filtering is used for other blocks. This local quality approach ensures prediction signal continuity where needed without universally increasing processing time.