Block-Adaptive Intra Prediction Filtering for Video Coding

The patent applies preliminary action by pre-defining multiple filter sets with different filter lengths and characteristics before the actual video processing occurs. The encoder and decoder both pre-store these filter sets, and during processing, the appropriate filter set is selected based on block size and prediction mode, eliminating the need for real-time filter design and improving coding efficiency while maintaining manageable complexity.

The patent employs parameter changes by varying filter length and filter coefficients based on block size parameters and prediction mode parameters. Different block sizes (e.g., 4×4, 8×8, 16×16) and prediction modes trigger selection of different filter sets, allowing the system to adapt filtering strength to the specific processing context, thereby improving coding efficiency without requiring a single complex universal filter.

The patent applies partial action by selectively filtering only those reference samples that benefit from filtering based on block size and prediction mode characteristics. Not all reference samples undergo filtering - the system applies filtering partially where needed (e.g., larger blocks with certain prediction modes) and skips it where unnecessary (e.g., small blocks or specific modes), thus improving prediction accuracy without incurring full filtering overhead for all cases.

The patent segments the filtering operation by dividing reference samples into different groups based on their position and usage context. Different filter sets are applied to different segments of reference samples depending on block size and prediction mode, allowing the system to optimize prediction accuracy for each segment while minimizing overall processing time through selective application.

The patent applies local quality by using different filter characteristics for different local contexts (block sizes and prediction modes). Instead of a uniform filter applied everywhere, the system selects from multiple filter sets based on local properties - for example, using longer filters for larger blocks and specific prediction modes while using shorter or no filters for smaller blocks or other modes, thereby improving coding efficiency through context-adaptive filtering.

The patent implements dynamics by making the filter selection dynamic rather than static. The filter set is chosen dynamically based on runtime parameters such as block size and prediction mode, allowing the system to adapt to varying processing conditions. This dynamic selection mechanism improves coding efficiency by matching filter characteristics to actual processing needs while keeping the complexity manageable through predefined filter sets.