CABAC Context Model Grouping for Parallel VVC Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in context model initialization and inheritance for CABAC, particularly in Versatile Video Coding (VVC, which can disrupt frame-level parallelism and result in suboptimal initial context models.
Innovation Solution
Grouping context models based on coding parameter values such as temporal layer, slice type, and quantization parameter (QP) to initialize and update context states independently within each group, allowing for improved CABAC efficiency while maintaining frame-level parallelism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If context models are initialized globally for CABAC in VVC, then coding efficiency is improved, but frame-level parallelism is disrupted
Solution Approach 1:
The patent segments the context models into multiple groups based on coding parameter values (temporal layer, slice type, QP). Each group is initialized independently, allowing frame-level parallelism to be maintained while still achieving coding efficiency improvements through selective context model inheritance within each group.
2Productivity
If context models are inherited from previous pictures, then CABAC efficiency is improved, but initialization complexity increases
Solution Approach 1:
The patent applies local quality by allowing context model inheritance only for specific groups of context models that benefit from it, rather than globally. The initialization complexity is reduced by selectively applying inheritance based on coding parameter groups, maintaining CABAC efficiency where needed while simplifying initialization elsewhere.
3Adaptability or versatility
If multiple context model groups are maintained, then coding flexibility is improved, but memory requirements increase
Solution Approach 1:
The patent merges context model groups that share the same initialization characteristics. By grouping context models based on coding parameters (temporal layer, slice type, QP) and initializing them in groups rather than individually, memory requirements are reduced while maintaining the flexibility to adapt to different coding conditions.
Data Source
AI summary
A video bitstream comprising a current block in a current picture is received. A plurality of groups of context models associated with coding parameter values used to code the current picture is determined. Each group of the plurality of groups of context models is grouped based on one or more of the coding parameter values. A group of context models for the current block is determined from the plurality of groups of context models according to at least one of the coding parameter values associated with the current block. Initial context states are determined for context models belonging to the determined group of context models. The current block is reconstructed based on the determined initial context states of the group of context models.