Image Encoding Using Shared Probability Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image compression methods, such as MPEG-4 AVC/H.264, reduce encoding efficiency when applied to YUV 4:4:4 or RGB images, as they are optimized for YUV 4:2:0 formats, leading to color distortions and suboptimal processing of luminance and chrominance components.
Innovation Solution
Applying the same spatial and temporal prediction techniques to all image components of YUV 4:4:4 or RGB images without transforming them to the YUV domain, using a single encoding mode and context-based binary arithmetic coding with a shared probability model for each component.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If different encoding techniques are applied to Y and UV components in YUV 4:2:0 format, then compression efficiency is improved, but device complexity increases and color distortions occur in 4:4:4 or RGB formats
Solution Approach 1:
The patent applies a universal encoding approach where the same prediction and encoding techniques are used for all image components (Y, U, V, R, G, B) regardless of format. This simplifies the encoding process by eliminating the need for separate processing paths for different components, while still achieving good compression efficiency through context-based arithmetic coding that adapts to the statistical properties of each component.
Solution Approach 2:
The patent introduces context-based binary arithmetic coding with separate probability models for different image components. This allows the encoding to adapt locally to the specific statistical characteristics of each component (Y, U, V, R, G, B) while using a unified prediction framework. The context models are updated based on the actual data being encoded, providing component-specific optimization without requiring completely separate encoding paths.
2Productivity
If YUV 4:2:0 format with different sampling rates is used, then compression efficiency is improved, but image quality deteriorates due to color distortions
Solution Approach 1:
The patent changes the encoding parameters by using context-based arithmetic coding with component-specific probability models instead of uniform quantization. This allows for more precise representation of chrominance information in 4:4:4 and RGB formats, maintaining image quality while achieving compression. The probability models are adapted to the statistical properties of each component, providing better precision for color information.
3Measurement precision
If conventional spatial prediction with different block sizes is applied to Y and UV components, then encoding precision is improved, but device complexity increases
Solution Approach 1:
The patent employs a universal prediction framework that uses the same spatial prediction techniques and block sizes for all image components. This simplifies the processing by eliminating the need for separate prediction loops for different components. The context-based arithmetic coding provides the necessary precision adaptation without requiring separate prediction paths.
4Measurement precision
If separate probability models are used for each image component in arithmetic coding, then encoding precision is improved, but device complexity increases
Solution Approach 1:
The patent implements context-based binary arithmetic coding with separate probability models for different image components. Each component (Y, U, V, R, G, B) has its own context model that captures the local statistical properties. The models are updated dynamically based on the data being encoded, providing high precision while managing complexity through shared coding infrastructure and context model reuse across components.
Data Source
AI summary
A method, medium, and apparatus encoding and/or decoding an image in order to increase encoding and decoding efficiency by performing binary-arithmetic coding/decoding on a binary value of a syntax element using a probability model having the same syntax element probability value for respective context index information of each of at least two image components.