Neural evidence for image quality perception based on algebraic topology

Area of Science:

• Neuroscience
• Image Processing
• Data Analysis

Background:

• Assessing image quality traditionally relies on objective metrics, often failing to capture subjective human perception.
• Understanding the neural basis of image quality perception is crucial for developing advanced assessment methods.

Purpose of the Study:

• To investigate the algebraic topological characteristics of brain networks derived from electroencephalogram (EEG) signals in response to different image quality levels.
• To propose a novel neurophysiological approach for image quality assessment by integrating EEG analysis with topological data analysis.

Main Methods:

• Collected EEG data while participants viewed images with varying distortions (e.g., JPEG compression, Gaussian blur).
• Applied topological data analysis to extract algebraic topological features from EEG signals.
• Analyzed frequency band differences, particularly in the beta band, and phase transition variations.

Main Results:

• Statistically significant differences were found in the algebraic topological characteristics of EEG signals between clear and unclear images, especially in the beta frequency band.
• Human sensitivity to JPEG compression was found to be more pronounced than to Gaussian blur, evidenced by greater phase transition differences in brain networks.
• The study successfully identified neurophysiological markers related to perceived image quality.

Conclusions:

• Algebraic topological characteristics of EEG signals can effectively differentiate between various image quality levels.
• The proposed neurophysiological approach offers a promising avenue for objective and perception-based image quality assessment.
• Brain network responses show differential sensitivity to specific image distortions, highlighting perceptual differences.