Brain-Inspired Image Perceptual Quality Assessment Based on EEG: A QoE Perspective.

Area of Science:

• Computer Vision
• Neuroscience
• Signal Processing

Background:

• Effective image perceptual quality metrics are crucial for human-oriented image communication.
• Traditional metrics are limited by subjective human cognitive activities.
• There is a need for objective and reliable image quality assessment.

Purpose of the Study:

• To develop a brain response-based image perceptual quality metric.
• To create a brain-inspired network for assessing image perceptual quality.
• To improve the accuracy of image quality assessment by considering brain responses.

Main Methods:

• Established electroencephalography (EEG) datasets by recording brain signals while subjects viewed distorted images.
• Designed a measurement model to extract EEG features reflecting human perception, creating the EEG perceptual score (EPS).
• Developed a prediction model to generate EPS directly from images, embedding the brain perception process.

Main Results:

• The proposed measurement and prediction models demonstrated superior performance.
• The EEG perceptual score (EPS) effectively captures human brain responses to image quality.
• The brain response-based metric provides a more accurate assessment of image perceptual quality.

Conclusions:

• The developed brain response-based metric offers a more objective and accurate method for image perceptual quality assessment.
• This approach overcomes limitations of traditional metrics by directly measuring brain states.
• The brain-inspired network and EPS hold promise for advanced image communication systems.