Intra Prediction Block Partitioning for Better Video Compression

The current block is partitioned into multiple partial blocks (e.g., two partial blocks), and different intra-prediction modes are applied to each partial block. This segmentation allows the encoder to capture local variations in the block more effectively, improving compression efficiency while managing complexity through structured partitioning.

Different intra-prediction modes are selected for different partial blocks based on local characteristics. The encoder determines the most suitable prediction mode for each partial block independently, allowing local optimization of prediction accuracy and compression efficiency without requiring complex global optimization.

The block is divided into a limited number of partial blocks (e.g., 2), which balances prediction accuracy improvement with encoding time constraints. This segmentation level provides sufficient local variation capture without excessive computational overhead.

The encoder changes the prediction mode parameter for each partial block independently, allowing adaptation to local characteristics. This parameter variation improves prediction accuracy while the limited number of partial blocks keeps the encoding time increase manageable.

Inflection points and slope information are calculated in advance during the partitioning determination phase, before the actual prediction and encoding steps. This preliminary calculation of geometric characteristics enables more efficient subsequent encoding operations by pre-establishing the optimal partitioning structure.

The partitioning shape determination process uses self-contained geometric analysis of the block boundaries and pixel gradients. The inflection point and slope calculations are performed autonomously on the block data itself, without requiring external reference information, which simplifies the overall computational framework.