No-Reference Quality Metric of Contrast-Distorted Images Based on Information Maximization

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Image quality assessment (IQA) is crucial for digital imaging.
• Evaluating image quality without a reference image (blind IQA) presents significant challenges, especially for contrast distortion.

Purpose of the Study:

• To develop a novel no-reference/blind metric for assessing image quality specifically affected by contrast distortion.
• To improve the accuracy and reliability of blind image quality evaluation.

Main Methods:

• A novel blind image quality assessment (IQA) metric is proposed.
• Local image details are analyzed using visual saliency and entropy after removing predicted regions.
• Global image information is assessed by comparing image histograms with uniform distributions using Kullback-Leibler divergence.

Main Results:

• The proposed training-free blind IQA method effectively estimates the quality of contrast-distorted images by integrating local and global features.
• Experiments on five databases demonstrate the method's superiority over state-of-the-art full-reference and no-reference IQA techniques.
• The developed model significantly enhances the performance of general-purpose blind quality metrics.

Conclusions:

• The novel blind IQA metric provides a robust and accurate solution for evaluating contrast distortion.
• This approach offers a valuable tool for image quality analysis and improvement in various applications.
• The method's ability to improve existing blind IQA metrics highlights its versatility and effectiveness.