Intra Prediction Filtering by Block Size for Video Coding

The patent applies filtering selectively based on block size parameters. Different filter kernels (e.g., 5-tap vs. 7-tap) are chosen based on the block dimensions, and filtering is applied only when the block size exceeds certain thresholds. This parameter-based selection optimizes the balance between prediction accuracy and computational complexity by avoiding unnecessary filtering operations on small blocks while applying appropriate filtering to larger blocks that benefit from it.

The patent applies filtering selectively to specific regions and blocks based on their characteristics. Rather than uniformly filtering all reference samples, the method identifies blocks that benefit from filtering (e.g., larger blocks, blocks with specific prediction modes) and applies filtering only to those cases. This local quality approach ensures high prediction sample quality where needed while minimizing unnecessary encoding complexity.

The patent uses block size parameters to control filter application decisions. By establishing clear thresholds and rules based on block dimensions, the system can efficiently determine whether filtering should be applied without requiring additional signaling. The block size itself serves as the control parameter, eliminating the need for separate filter selection flags and reducing signaling overhead while maintaining coding efficiency.