Modeling visual attention's modulatory aftereffects on visual sensitivity and quality evaluation.

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Image Processing

Background:

• Visual noise-shaping applications (compression, error resilience, watermarking, encryption, display) require enhanced performance.
• Incorporating perceptual characteristics is crucial for improving these visual applications.
• Visual attention is a key mechanism of the human brain influencing visual perception.

Purpose of the Study:

• To introduce visual attention for visual sensitivity and quality evaluation in visual applications.
• To propose a novel numerical measure, the perceptual quality significance map (PQSM), reflecting human brain's processing of visual content.
• To enhance the performance of visual sensitivity models and visual quality metrics using PQSM.

Main Methods:

• Analysis of visual attention mechanisms and their modulatory aftereffects.
• Development of the perceptual quality significance map (PQSM) integrating local perceptual stimuli (color contrast, texture contrast, motion) and cognitive features (skin color, face detection).
• Experimental validation using subjective viewing to assess PQSM's impact.

Main Results:

• The proposed PQSM statistically reflects the human brain's processing ability on local visual content.
• PQSM integration led to performance improvements in two visual sensitivity models.
• PQSM-based visual quality metrics demonstrated enhanced performance compared to existing methods.

Conclusions:

• The perceptual quality significance map (PQSM) is an effective tool for incorporating visual attention into image processing.
• PQSM enhances visual sensitivity and quality evaluation in various visual applications.
• This approach offers a statistically grounded method for perceptual quality assessment.