Graph-Based Prediction Using Optimization Function

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding standards face challenges in accurately predicting signal models for complex signals like audio, images, and video, as model parameters change significantly with different sources, and existing prediction methods rely on experimental statistics and limited geometrical models.

Innovation Solution

A graph-based prediction method using an optimization function, defined by a cost function based on the distribution of eigenvalues of the graph Laplacian and linear combinations of eigenvectors, to improve prediction accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional statistical prediction methods are used for video coding, then prediction can be performed based on experimental statistics, but prediction accuracy deteriorates for complex signals because model parameters change significantly with different sources

Engineering Contradiction:
Improveprediction reliabilityVSAvoidadaptability to different signal sources
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent changes from fixed experimental statistical parameters to adaptive parameters derived from graph signal processing. The graph Laplacian eigenvalues and eigenvectors are computed dynamically based on the specific signal characteristics, allowing the prediction model to adapt to different signal sources while maintaining reliability through the mathematical properties of graph transforms.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If graph-based prediction methods are used to improve prediction accuracy, then prediction performance is enhanced, but computational complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidalgorithm complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the prediction process into distinct stages: graph construction from signal data, computation of graph Laplacian and its eigen decomposition, and application of the transform for prediction. This segmentation allows for optimized implementation where only necessary components are computed, reducing overall complexity while maintaining accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by using graph-based transforms that are adapted to local signal characteristics through the graph structure. The graph Laplacian eigen decomposition provides localized frequency information that can be selectively applied to different signal regions, improving accuracy without requiring full-signal complex processing.

Inventive Principle:
Principle #3Local quality

3Productivity

If conventional prediction methods are used, then implementation is simpler, but compression efficiency is reduced due to higher bit rate overhead

Engineering Contradiction:
Improvecompression efficiencyVSAvoidbit rate overhead
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent replaces conventional mechanical-like prediction operations with graph-based transform operations. The graph Laplacian transform provides a more efficient representation of signal characteristics, allowing for better prediction with reduced residual energy, thereby reducing the bit rate needed to encode prediction errors and improving overall compression efficiency.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3211894B1Method and apparatus for performing graph-based prediction by using optimization function
Publication Date: 2020.03.04 LG ELECTRONICS INC
  • EP3211894B1 patent drawingFigure 1
  • EP3211894B1 patent drawingFigure 2
  • EP3211894B1 patent drawingFigure 3(a)~3(b)

AI summary

The present invention provides a method for performing a graph-based prediction using a graph signal including obtaining a context signal; generating a graph signal based on a graph parameter set; obtaining a graph-based transform matrix based on the graph signal, wherein the graph-based transform matrix includes an eigenvector; obtaining a prediction vector using an optimal transform vector calculated through an optimization function; and generating a prediction signal using the prediction vector, where the optimization function has the context signal, an eigenvalue of a graph Laplacian matrix and the eigenvector as a variable.