Entropy Coding Context Model Selection Based on Transformation Depth

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image compression methods, such as MPEG-1, MPEG-2, and H.264/AVC, face complexity in entropy encoding and decoding due to the need for selecting appropriate context models for transformation units, which complicates the process and reduces efficiency.

Innovation Solution

The method determines a context model for entropy encoding and decoding based on the transformation depth, which indicates the hierarchical splitting relationship between coding and transformation units, simplifying the selection process and operation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple context models are used for different transformation units, then encoding accuracy is improved, but device complexity and operation difficulty increase

Engineering Contradiction:
Improveencoding accuracyVSAvoidcontext model selection complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the transformation unit into different types (first type when transformation depth equals coding depth, second type when transformation depth is less than coding depth) and applies different context model selection rules to each type. This segmentation allows accurate encoding for each specific case while maintaining clear, manageable selection criteria without requiring complex unified logic.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If complex context model selection criteria are used, then encoding precision is improved, but ease of operation deteriorates

Engineering Contradiction:
Improvesyntax element decoding precisionVSAvoidcontext model selection ease
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent performs preliminary classification of transformation units based on the relationship between transformation depth and coding depth before selecting context models. By pre-establishing the two-type classification system, the decoder can quickly determine which context model to use without performing complex real-time analysis, thereby improving both precision and operational ease.

Inventive Principle:
Principle #10Preliminary action

3Ease of operation

If transformation depth-based context model selection is implemented, then operation complexity is reduced, but adaptability to different coding scenarios may be limited

Engineering Contradiction:
Improveentropy encoding operation simplicityVSAvoidcontext model adaptability
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent applies different context model selection strategies locally to different transformation unit types. First-type transformation units (where transformation depth equals coding depth) use one selection rule, while second-type transformation units (where transformation depth is less than coding depth) use another rule. This local differentiation maintains operational simplicity while adapting to the specific characteristics of each transformation unit type.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3361735B1Entropy coding of a video and entropy decoding of a video
Publication Date: 2020.01.08 SAMSUNG ELECTRONICS CO LTD
  • EP3361735B1 patent drawingFigure 1~2
  • EP3361735B1 patent drawingFigure 3
  • EP3361735B1 patent drawingFigure 4

AI summary

Provided are entropy decoding and encoding methods of a video. The entropy decoding method includes obtaining a transformation unit significant coefficient flag indicating whether a non-zero transformation coefficient exists in the transformation unit, from a bitstream, determining a context model for arithmetically decoding the transformation unit significant coefficient flag, based on the transformation depth of the transformation unit and arithmetically decoding the transformation unit significant coefficient flag based on the determined context model.