Screen Content Compression via Dynamic Context Probability Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression methods are inefficient for compressing synthetic data such as computer-generated graphics, text, and screenshots, as they fail to accurately estimate contextual probability distributions, leading to high computational complexity and suboptimal compression performance.

Innovation Solution

A method that dynamically generates a distribution model for each pixel and feeds it into a standard multi-level arithmetic coder, using a two-step processing approach with contextual pattern analysis to estimate and encode probabilities, reducing computational complexity while achieving excellent compression performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional predictive techniques or dictionary-based compression methods are used for screen content, then device complexity is reduced, but compression performance deteriorates due to inability to accurately estimate contextual probability distributions

Engineering Contradiction:
Improvecontextual probability estimation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-defining a limited set of context types (e.g., horizontal, vertical, diagonal neighborhoods) before processing the image. This allows the system to prepare context probability estimates in advance without computing all possible context configurations, thereby improving estimation accuracy while controlling computational complexity.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent changes parameters by adapting context probability distributions dynamically based on local image characteristics. Instead of using fixed probability models, the system adjusts context parameters (such as neighborhood size and pattern types) according to the actual data being compressed, achieving better compression performance without proportional increases in complexity.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If optimal estimation of contextual symbol probabilities is performed, then compression performance is improved, but computational effort increases enormously

Engineering Contradiction:
Improvecompression ratioVSAvoidcomputational effort
Core Design Contradiction:
ProductivityVSPower

Solution Approach 1:

The patent segments the complex probability estimation problem into smaller, manageable context types. Instead of computing probabilities for all possible pixel configurations, the image is divided into segments with similar local characteristics, and probability models are computed separately for each segment type. This reduces computational effort while maintaining compression performance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by computing probability estimates for only the most relevant context configurations rather than all possible contexts. The system identifies and processes the subset of context types that provide the most significant compression benefits, avoiding unnecessary computational overhead from less impactful context analyses.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If a dedicated compression system for synthetic data is implemented, then compression performance is improved, but adaptability to different image types decreases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidimage type flexibility
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamics by making the compression system adaptive through dynamic context probability estimation. The system can adjust its context models based on the actual characteristics of the input image, allowing it to handle different types of synthetic data (text, graphics, screenshots) effectively. This dynamic adaptation maintains high compression efficiency across diverse image types without requiring separate dedicated systems.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3319318B1Intra compression of screen content based on soft context formation
Publication Date: 2019.08.14 DEUTSCHE TELEKOM AG
  • EP3319318B1 patent drawingFigure 1
  • EP3319318B1 patent drawingFigure 2
  • EP3319318B1 patent drawingFigure 3

AI summary

The present invention relates to a method and a system for efficiently compressing images by estimating contextual probability distributions based on virtually all samples processed up to a current position in the image. The method comprises encoding a distribution symbol of the property vector X from a distribution of the property vectors found in the contextual patterns that are identical or similar to the contextual pattern of the respective pixel, when the property vector X of the respective pixel occurred in any of the contextual patterns that are identical or similar to the contextual pattern of the respective pixel; encoding a first stage exception handling symbol and a palette symbol of the respective property vector X of a dynamically generated palette of property vectors, when the respective property vector X of the respective pixel has not occurred in the contextual patterns that are identical or similar to the contextual pattern of the respective pixel; and encoding a first stage exception handling symbol, a second stage exception handling symbol, and a representation of all components of the property vector X, when the property vector X has not occurred in the contextual patterns that are identical or similar to the contextual pattern of the respective pixel and the property vector X is not found in the palette.