Intra Prediction Reference Block Selection for Better Coding Performance

The patent divides the reference data blocks into multiple distinct sets (first reference data block set, second reference data block set, third reference data block set) based on their spatial relationships with the current data block. This segmentation allows the system to manage different types of reference blocks separately, improving coding performance by selecting from diverse reference sources while maintaining organized complexity through structured categorization.

The patent performs preliminary organization of reference data blocks into multiple sets before the actual predictive coding process. By pre-categorizing reference blocks based on their spatial relationships and characteristics, the system prepares the reference data structure in advance, allowing for faster selection during encoding/decoding operations and reducing processing time while maintaining improved coding performance.

The patent applies different selection criteria to different spatial locations relative to the current data block. Neighbor relationship data blocks are included in specific sets (second and third reference data block sets) where they can provide locally relevant reference information for improving coding accuracy. The system selectively uses these locally correlated blocks only where appropriate, rather than uniformly applying correlation-based selection across all reference blocks, thus balancing accuracy improvement with reliability.