Context-Adaptive CABAC for Large CTU Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding technologies face inefficiencies in encoding and decoding large coding tree units (CTUs) due to limitations in context modeling, particularly when CTU sizes exceed 64x64, leading to reduced coding efficiency and increased complexity.

Innovation Solution

The implementation of context-adaptive binary arithmetic coding (CABAC) that selects contexts for syntax elements based on the CU depth of neighboring CUs, adjusting the number of contexts used and applying constraints on CU sizes to enhance coding efficiency for both intra- and inter-coded CUs, especially for CTUs larger than 64x64.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the number of contexts in CABAC is increased to improve modeling accuracy for large CTUs, then coding efficiency improves, but device complexity and computational load increase

Engineering Contradiction:
Improvecontext modeling accuracyVSAvoidcoding device complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies different context modeling strategies to different regions of the video data. Specifically, it uses different numbers of contexts (e.g., 3 contexts for one syntax element, 6 contexts for another) depending on the local characteristics of the coding units, rather than uniformly applying the same number of contexts throughout. This allows improved modeling accuracy where needed while reducing complexity in other areas.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the context modeling process into multiple independent context models. Instead of using a single large number of contexts for all syntax elements, it divides the modeling into separate context sets (e.g., separate contexts for split_cu_flag, separate contexts for other syntax elements), each with its own optimized number of contexts. This segmentation reduces overall computational complexity while maintaining modeling accuracy.

Inventive Principle:
Principle #1Segmentation

2Productivity

If CTU size is increased to reduce the number of coding units, then processing speed improves, but coding efficiency and context modeling accuracy deteriorate

Engineering Contradiction:
Improveprocessing speedVSAvoidcoding efficiency
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements dynamic context modeling that adapts to different CTU sizes. The context modeling parameters and number of contexts are adjusted based on the actual CTU size being processed. This allows the system to maintain high coding efficiency whether processing large CTUs (for speed) or smaller CTUs (for accuracy), as the context model dynamically adapts to the current processing scale.

Inventive Principle:
Principle #15Dynamics

3Device complexity

If the number of contexts is reduced to decrease computational complexity, then device complexity decreases, but context probability modeling accuracy deteriorates

Engineering Contradiction:
Improvecomputational complexityVSAvoidcontext probability modeling accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent changes key parameters of the context modeling process, specifically the number of contexts used for different syntax elements. By carefully selecting and adjusting these parameters (e.g., using 3 contexts for certain elements, 6 for others), the patent achieves an optimal balance between computational complexity and modeling accuracy, reducing complexity where possible while maintaining necessary precision.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3251354B1Contexts for large coding tree units
Publication Date: 2022.04.20 QUALCOMM INC
  • EP3251354B1 patent drawingFigure 1
  • EP3251354B1 patent drawingFigure 2
  • EP3251354B1 patent drawingFigure 3

AI summary

In an example, a method of coding video data may include selecting, as part of an entropy coding process, a context for a syntax element associated with a current coding unit (CU) based on information relating to CU depth of one or more neighboring CUs relative to the current CU. The syntax element may represent whether the current CU is split into sub-CUs. The method may include entropy coding the syntax element based on the selected context.