Image Quality Assessment via Subjective Feature Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image quality assessment systems are limited in their ability to provide assessments that reflect perceptual quality as perceived by human observers, especially when a reference image or distortion type is unknown, and they often exhibit bias across distortion types.
Innovation Solution
A methodology that uses subjective scoring to predict human observer ratings by mapping image features to subjective scores, enabling the generation of image quality scores that reflect perceived quality, even without a reference image or known distortion type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If PSNR quality measures are used, then quantitative image quality assessment is achieved, but the measures do not correlate well with perceptual quality
Solution Approach 1:
The patent changes the parameter space from traditional PSNR metrics to a transformed feature space that includes natural image statistics, texture features, and distortion indicators. This parameter transformation enables the system to capture perceptual quality attributes that PSNR cannot measure, resolving the contradiction between quantitative assessment capability and perceptual correlation.
Solution Approach 2:
The patent introduces a learned mapping function as an intermediary that transforms objective image features into subjective quality predictions. This intermediary layer bridges the gap between conventional quantitative metrics and human perception, allowing the system to maintain quantitative assessment while improving perceptual correlation.
2Ease of operation
If reference image-based measures are used, then image quality assessment is simplified, but the measures are useful only in limited situations where a reference image is available
Solution Approach 1:
The patent creates a universal quality assessment system that can function in multiple scenarios: with reference images, without reference images, with known distortion types, and with unknown distortion types. The system uses a combination of natural image statistics, texture features, and distortion indicators that work across all these cases, eliminating the limitation of reference-image-based approaches.
Solution Approach 2:
The system performs self-service quality assessment by using the image itself and its features to generate quality predictions without requiring external reference images. The learned mapping function processes the image features directly to produce subjective quality scores, enabling the system to assess quality autonomously in real-world scenarios where reference images do not exist.
3Measurement precision
If distortion specific image quality measures are used, then image quality assessment can be performed, but the measures exhibit bias across distortion types
Solution Approach 1:
The patent applies local quality analysis by extracting and evaluating multiple types of features (natural image statistics, texture features, distortion indicators) from different regions and aspects of the image. This localized feature extraction allows the system to assess quality characteristics specific to each distortion type while maintaining overall fairness across all distortion types through the comprehensive feature set.
Solution Approach 2:
The system uses a composite approach by combining multiple feature types (natural image statistics, texture features, distortion indicators) into a unified quality assessment framework. This composite feature set eliminates bias toward any single distortion type by ensuring that all distortion types are evaluated using the same comprehensive set of features, just as composite materials provide balanced properties across different applications.
Data Source
AI summary
Methods and systems for image quality assessment are disclosed. A method includes accessing an image, identifying features of the image, assessing the features and generating subjective scores for the features based upon a mapping of the features to the subjective scores and based on the subjective scores, generating an image quality score. Access is provided to the image quality score.