Arithmetic Decoding of Last Position Bits With Segmented Contexts

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional context adaptive binary arithmetic coding and decoding techniques face challenges in accurately predicting symbol probabilities, leading to decreased coding efficiency due to inappropriate switching among contexts, especially when different bit positions have varying probabilities of symbol occurrence.

Innovation Solution

The method involves binarizing last position information to generate binary signals with varying lengths, allowing for context-switched arithmetic coding of prefix parts and fixed probability decoding of suffix parts, with the last bit position using an exclusive context to enhance coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If context adaptive binary arithmetic coding is used for last position information, then coding efficiency can be improved through probability prediction, but inappropriate context switching decreases accuracy and coding efficiency

Engineering Contradiction:
Improvecoding efficiencyVSAvoidprobability prediction accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the last position information into multiple bins based on block size (e.g., 4 bins for 4x4 blocks, 9 bins for 8x8 blocks, 16 bins for 16x16 blocks). Each bin corresponds to a specific range of last position values and uses a dedicated context model. This segmentation allows the coding system to adapt to the varying probability distributions at different position ranges without requiring complex dynamic context switching, thereby resolving the contradiction between coding efficiency and prediction accuracy.

Inventive Principle:
Principle #1Segmentation

2Productivity

If multiple contexts are used for different bit positions, then coding efficiency improves, but device complexity and memory requirements increase

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcontext switching complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies local quality by assigning different context models to different bins based on their local characteristics (block size and position range). Each bin uses a context model optimized for its specific local probability distribution rather than using a single global context model. This approach improves coding efficiency for each local region while avoiding the complexity of dynamic context switching by making the context selection static and deterministic based on bin assignment.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3849199B1Image decoding method and image decoding device
Publication Date: 2022.12.28 TAGIVAN II
  • EP3849199B1 patent drawingFigure 1
  • EP3849199B1 patent drawingFigure 2
  • EP3849199B1 patent drawingFigure 3A~3B

AI summary

An image coding method including: binarizing (S401) last position information to generate (i) a binary signal which includes a first signal having a length smaller than or equal to a predetermined maximum length and does not include a second signal or (ii) a binary signal which includes the first signal having the predetermined maximum length and the second signal; first coding (S402) for arithmetically coding each of binary symbols included in the first signal using a context switched among a plurality of contexts according to a bit position of the binary symbol; and second coding (S404) for arithmetically coding the second signal using a fixed probability when the binary signal includes the second signal, wherein in the first coding, a binary symbol at a last bit position of the first signal is arithmetically coded using a context exclusive to the last bit position, when the first signal has the predetermined maximum length.