Context Modeling for Variable Block Entropy Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video and image coding systems face inefficiencies in entropy coding and context modeling, particularly in determining contexts for block partitioning structures like quadtree, binary tree, and combined quadtree plus binary tree, which affect coding efficiency and complexity.

Innovation Solution

The method involves context-based entropy coding for symbols associated with blocks partitioned using quadtree, binary tree, or combined quadtree plus binary tree structures, determining contexts based on neighboring blocks' information and block shape or depth, and applying adaptive binary arithmetic coding to improve coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If context determination is based on neighboring blocks and block characteristics, then coding efficiency is improved, but computational complexity increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies local quality by determining contexts based on local neighboring block information and specific block characteristics (depth, shape) rather than using a uniform context model. This allows the encoding to adapt to local statistical properties, improving coding efficiency while keeping the complexity localized to relevant regions

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes parameters by using multiple context indices (ctxLnc, ctxRnc, ctxSnc) that are dynamically selected based on block depth, shape, and neighboring block properties. This parameter adaptation allows the system to optimize for different block types and positions, improving overall coding efficiency

Inventive Principle:
Principle #35Parameter changes

2Productivity

If multiple context indices are used for different block types and positions, then entropy coding performance is optimized, but the complexity of context selection increases

Engineering Contradiction:
Improveentropy coding performanceVSAvoidcontext selection complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the context modeling space by creating different context indices for different block types (luma/chroma), different positions (left/right/neighboring), and different depths. This segmentation allows each segment to be optimized independently, improving overall entropy coding performance while organizing the complexity into manageable segments

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic context selection where the appropriate context index is chosen based on runtime conditions such as block depth, block shape, and neighboring block characteristics. This dynamic adaptation optimizes coding performance for each specific case while the systematic selection rules keep the complexity manageable

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3363199B1Method and apparatus of entropy coding and context modelling for video and image coding
Publication Date: 2021.05.19 HFI INNOVATION INC
  • EP3363199B1 patent drawingFigure 1~2
  • EP3363199B1 patent drawingFigure 3~4
  • EP3363199B1 patent drawingFigure 5

AI summary

A method and apparatus for applying entropy coding to a symbol associated with a block are disclosed. According to the present invention, context-based entropy coding is applied to source symbols associated with blocks having variable block sizes generated by partitioning an initial block using a quadtree structure, a binary-tree structure or a combined quadtree plus binary-tree structure. Contexts according to the present invention are based on some information derived from neighbouring blocks and also based on at least one of the shape, the size and the depth of the current block since the statistics of the symbols associated with the current block may be correlated with how the current block has been partitioned through a tree structure. A current symbol to be encoded or decoded may correspond to split flags and modes associated with the tree structure, skip flag or prediction mode flag.