8-bit Line Graph Transform Cores for Video Coding

The patent segments the transform operation into multiple 4-point line graph transform stages to implement an 8-point transform. By dividing the 8-point transform into two sequential 4-point transforms, the system achieves efficient 8-point transformation capability using only well-optimized 4-point transform cores, thereby improving coding efficiency without requiring separate 8-point transform implementations.

The patent creates a universal transform core design where a single 4-point line graph transform core serves multiple purposes. The same core is used both as a standalone 4-point transform and as a building block for constructing 8-point transforms through concatenation, eliminating the need for separate dedicated 8-point transform cores and improving adaptability.

The patent segments larger transforms into multiple 4-point transform operations. An 8-point transform is divided into two 4-point transforms, a 16-point transform into four 4-point transforms, and a 64-point transform into six 4-point transforms. This segmentation maintains coding accuracy by preserving the mathematical equivalence of the transforms while simplifying implementation through reuse of a single optimized 4-point core.

The patent implements a nested structure where 4-point transform cores are nested within larger transform operations. The same 4-point core is recursively applied multiple times in sequence to achieve 8-point, 16-point, and 64-point transforms, creating a hierarchical implementation that reduces complexity by eliminating the need for separate core implementations for each transform size.