Visual stream connectivity predicts assessments of image quality

Area of Science:

• Neuroscience
• Computer Vision
• Psychophysics

Background:

• Early vision mechanisms are extensively studied, yet new discoveries continue.
• Understanding human perception of image similarity is crucial for various applications.

Purpose of the Study:

• To introduce a novel formal treatment of image similarity psychophysics.
• To develop a model derived from early visual pathway connectivity patterns.
• To explain human perceptual similarity judgments.

Main Methods:

• Developed a differential geometry formulation based on visual pathway connectivity.
• Applied the model to predict human perceptual similarity judgments.
• Improved predictions using regression on human behavioral data.
• Constructed hypothesized neural connectivity patterns.

Main Results:

• The differential geometry model accurately accounts for human perceptual similarity.
• The model's predictions were enhanced through regression analysis.
• The approach outperforms a standard measure of perceived image fidelity.
• The study provides clear explanatory principles for image similarity findings.

Conclusions:

• The novel formal treatment offers accurate and explanatory accounts of image similarity perception.
• The proposed method advances the understanding of early vision and image fidelity.
• This approach has implications for both neuroscience and computer vision fields.