Video Encoder BDOF Processing for Block Size Constraints

The patent applies bi-directional optical flow (BDOF) processing selectively to specific picture block sizes (e.g., 16x16, 32x32, 64x64 blocks) rather than uniformly to all blocks. This local quality approach ensures high prediction accuracy for larger blocks where BDOF provides significant benefit, while avoiding unnecessary complexity for smaller blocks where simpler prediction methods suffice.

The patent changes the parameter of block size as a criterion for applying BDOF processing. By setting minimum block size thresholds (e.g., width≥16 and height≥16), the system dynamically adjusts the complexity of prediction processing based on the spatial dimensions of each picture block, thereby optimizing the balance between prediction accuracy and coding complexity.

The patent implements size-based selective processing where BDOF is applied only to picture blocks meeting minimum dimension thresholds (e.g., 16x16 or larger). This local quality strategy concentrates computational resources on blocks where BDOF provides the most significant accuracy improvement, while using faster, simpler prediction methods for smaller blocks, thereby reducing overall encoding time.

The patent applies BDOF processing partially rather than universally - specifically to blocks that meet size criteria. This partial action approach avoids the excessive computational burden of applying BDOF to all blocks, including small ones where the processing overhead would outweigh the accuracy benefits, thus optimizing encoding efficiency.