HEVC CABAC Lossless Coding With High-Throughput Binarization

The patent segments the context models into multiple categories (e.g., luma intra prediction, luma inter prediction, chroma prediction) and further divides each category into sub-categories based on prediction modes and block sizes. This segmentation allows the decoder to selectively apply only the necessary context models for each specific coding scenario, reducing the overall computational burden while maintaining compression effectiveness.

The patent implements a simplified CABAC mode where only a subset of context models is used instead of the full set. By applying partial action (using fewer context models than the standard), the patent reduces computational complexity while still achieving acceptable compression performance, particularly for high-bitrate scenarios where full CABAC may be overly complex.

The patent introduces dynamic selection of context model subsets based on coding conditions. The decoder dynamically determines which context models to apply based on factors such as prediction mode, block size, and bit-rate characteristics. This dynamic approach allows the system to adapt to different scenarios, using more sophisticated models when necessary and simpler models when sufficient, thereby optimizing processing efficiency while maintaining compression effectiveness.

The patent changes the parameter of context model selection by introducing a simplified mode that uses a reduced set of context models. By altering this parameter (the number and type of context models used), the patent achieves a balance between compression effectiveness and processing efficiency, particularly benefiting high-bitrate applications where the full set of context models provides diminishing returns.