Constrained Block Partitioning for Video Decoding Efficiency

The patent segments the block partitioning process into distinct phases: first applying quadtree partitioning to divide the current block into smaller regions, then applying binary tree partitioning only to specific sub-regions. This segmentation allows the encoder to achieve fine-grained partitioning where needed while maintaining coarser structures elsewhere, improving processing efficiency without uniformly increasing complexity across the entire partitioning structure.

The patent introduces a size parameter constraint (maximum block size) that controls when ternary-tree splits can be applied. By changing the parameter from 'unrestricted size' to 'maximum size threshold', the system maintains partitioning flexibility for blocks that benefit from it while automatically disabling complex splits for smaller blocks where they would increase computational complexity without providing benefit.

The patent performs preliminary partitioning actions by establishing a constrained partitioning structure upfront using quadtree and binary tree splits with maximum size limits. This preliminary structuring ensures that all necessary partitions are determined before processing begins, allowing the decoder to process each block exactly once without needing to revisit previously processed regions, thus maintaining both efficiency and completeness.