A feature-based model of symmetry detection.

Area of Science:

• Computer vision
• Computational neuroscience
• Image processing

Background:

• Symmetry detection is a fundamental aspect of biological visual systems across various species, including mammals, insects, and birds.
• Understanding the computational mechanisms underlying symmetry perception can provide insights into visual processing in both biological and artificial systems.

Purpose of the Study:

• To develop and evaluate a novel, two-stage algorithm for detecting image symmetry.
• To quantify the degree of asymmetry in images, particularly human faces.
• To assess the algorithm's performance against human psychophysical ratings of symmetry.

Main Methods:

• The algorithm employs a two-stage approach: first, identifying visually salient features within an image.
• Second, evaluating the long-range symmetry of these features using a Gaussian filter.

Main Results:

• The algorithm successfully detects the axis of maximum symmetry and quantifies asymmetry in arbitrary images, including human faces.
• Evaluation on a human face dataset demonstrated the algorithm's ability to discern subtle symmetry variations.
• The algorithm's quantitative symmetry measurements showed a strong correlation with human psychophysical symmetry ratings.

Conclusions:

• The developed symmetry-detection algorithm is effective in identifying and quantifying image symmetry.
• The algorithm's performance suggests its potential utility in understanding biological visual systems and in computer vision applications.
• The strong correlation with human ratings validates the algorithm's perceptual relevance.