Related Experiment Video
Updated: Jun 28, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.9K
Graph-Represented Distribution Similarity Index for Full-Reference Image Quality Assessment
Summary
We introduce a novel Graph-Represented Image Distribution Similarity (GRIDS) index for image quality assessment. This method accurately predicts perceived image quality by analyzing graph-based distribution patterns.
Area of Science:
- Computer Vision
- Image Processing
- Machine Learning
Background:
- Image quality assessment (IQA) is crucial for evaluating visual data.
- Existing methods often struggle with capturing complex perceptual features.
- Graph-based representations offer a powerful way to model image content.
Purpose of the Study:
- To propose a new full-reference (FR) image quality assessment index called GRIDS.
- To measure perceptual image distance by comparing graph-based distribution patterns.
- To enhance the accuracy and reliability of image quality prediction.
Main Methods:
- Transforming images into graph-based representations (vision graphs).
- Decomposing graphs into cliques to derive joint probability distributions.
- Comparing feature distribution distances across different graph model depths.
Main Results:
- The GRIDS index demonstrates competitive performance against state-of-the-art methods.
- Achieved exceptional predictive accuracy in image quality tasks.
- Exhibited consistent and monotonic behavior in quality prediction.
Conclusions:
- The proposed GRIDS index is an effective tool for full-reference image quality assessment.
- Graph-based image representation captures essential perceptual features for IQA.
- GRIDS offers a promising approach for accurate and reliable image quality prediction.
