Super-Block Video Coding With Quadtree CTU Partitioning

The patent divides the video frame into a hierarchical structure: super coding blocks (SCBs) are first segmented, then each SCB is further partitioned into coding tree units (CTUs) using recursive quadtree partitioning. This multi-level segmentation allows the system to adapt to different video content characteristics while managing complexity through structured organization. The decoder receives explicit SCB partitioning information and uses it to guide the recursive partitioning process, making the adaptation more efficient.

The patent implements dynamic partitioning where the coding structure adapts based on content characteristics. The super coding blocks can have variable sizes and shapes, and the recursive quadtree partitioning dynamically adjusts the CTU boundaries based on local video content properties. This dynamic behavior allows the system to optimize for different scene types (e.g., smooth areas vs. detailed areas) while maintaining a manageable decoding process through systematic rules.

The patent extracts the super coding block partitioning information as a separate, explicit component in the bitstream. Instead of encoding partitioning details implicitly through complex syntax, the SCB boundaries and sizes are directly signaled. This extraction approach reduces the information overhead by separating the high-level SCB structure from the detailed CTU partitioning, allowing more efficient encoding of the partitioning data.

The patent performs preliminary partitioning of the video frame into super coding blocks before the actual CTU-level encoding. This preliminary action establishes a coarse-grained structure that simplifies subsequent detailed partitioning. By pre-defining SCB boundaries and characteristics, the system reduces the amount of information that needs to be encoded at the finer CTU level, thereby reducing overall partitioning information overhead while maintaining coding efficiency.