Luma–Chroma Wavefront Processing for Parallel Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression systems, such as H.265/HEVC, face limitations in achieving high levels of parallelization during the encoding and decoding of luma and chroma components, particularly in intra slices where separate coding trees for luma and chroma introduce inefficiencies.
Innovation Solution
The proposed method involves independently subdividing luma and chroma components into coding units and propagating context-based adaptive binary coding context variables between rows to enable wavefront parallel processing, allowing parallel processing of luma and chroma components without shared CABAC contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If luma and chroma components are processed separately with independent coding trees, then parallelization capability is improved, but coding efficiency deteriorates due to loss of spatial and temporal redundancy exploitation
Solution Approach 1:
The patent divides the video picture into separate luma and chroma coding tree units, allowing independent parallel processing of each component. Each component is subdivided into coding tree units that can be processed independently while maintaining overall compression efficiency through selective context propagation.
Solution Approach 2:
The patent applies different context propagation strategies to different regions and components. CABAC context variables are propagated from previous rows for luma components but not for chroma components, allowing optimized parallelization at the component level while maintaining coding efficiency where needed.
2Productivity
If CABAC context variables are propagated from previous rows, then parallel processing efficiency is improved, but coding performance deteriorates due to context variable dependency
Solution Approach 1:
The patent selectively propagates CABAC context variables based on component type and region. Luma components receive context propagation from previous rows to enable parallel processing, while chroma components maintain independent context initialization to preserve coding performance. This localized application of context propagation resolves the contradiction between parallel efficiency and coding performance.
Solution Approach 2:
The patent segments the context propagation mechanism by component type, creating separate context management paths for luma and chroma. This segmentation allows parallel processing optimization for luma while maintaining coding performance for chroma, resolving the contradiction through differentiated treatment of components.
3Productivity
If separate coding trees are used for luma and chroma, then independence and parallelization are improved, but device complexity increases due to dual tree management
Solution Approach 1:
The patent merges the context management operations for separate luma and chroma coding trees by using a unified wavefront parallel processing framework. Both components follow the same row-by-row processing pattern and context propagation mechanism, reducing overall system complexity despite maintaining separate coding trees for independence.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
At least a method and an apparatus are presented for improving parallelization for wavefront parallel encoding and decoding of luma and chroma components of a vide picture. For example, a luma component and a chroma component of a video picture are independently subdivided into respectively a plurality of luma coding units and a plurality of chroma coding units. A context-based adaptive binary coding context variable is propagated from a previous row to a current row of the independently subdivided plurality of luma coding units, and a context-based adaptive binary coding context variable is propagated from a previous row to a current row of the independently subdivided plurality of chroma coding units. The video picture is encoded or decoded using the propagations to provide the wavefront parallel processing.