Video Quantization Index Entropy Coding with TCQ Context Modeling

The patent merges multiple transform coefficients into groups and applies joint quantization to exploit correlations between coefficients within each group. Instead of processing each coefficient independently as in scalar quantization, the invention combines coefficients spatially or temporally adjacent to each other, allowing the quantization process to consider inter-coefficient relationships and preserve correlation information that would otherwise be lost.

The patent transitions from one-dimensional scalar quantization to multi-dimensional vector quantization by organizing coefficients into groups that capture spatial or temporal relationships. This dimensional expansion allows the quantization process to operate on vectors rather than scalars, enabling exploitation of correlations across multiple dimensions (e.g., horizontal, vertical, temporal) and improving entropy coding effectiveness.

The patent segments the quantization process into multiple stages: first dividing coefficients into groups based on spatial or temporal relationships, then applying quantization to each group rather than to individual coefficients. This segmentation allows the system to capture correlations within groups while maintaining manageable computational complexity through the structured grouping approach, rather than requiring complex global optimization.

The patent performs preliminary grouping and correlation analysis before the actual quantization operation. By pre-organizing coefficients into meaningful groups that exploit spatial or temporal correlations, the system prepares the data structure to maximize compression opportunities during subsequent entropy coding, without adding excessive complexity to the core quantization operation itself.