On the relative detectability of configural properties

Area of Science:

• Cognitive Psychology
• Computer Vision
• Geometry

Background:

• Understanding how humans perceive geometric transformations is crucial for fields like computer vision and artificial intelligence.
• The Klein hierarchy categorizes geometric properties based on their invariance under different transformations.

Purpose of the Study:

• To investigate the perceptual salience of various geometric transformations in human shape discrimination.
• To determine if human visual perception of shape changes aligns with the mathematical hierarchy proposed by Felix Klein.

Main Methods:

• A match-to-sample shape-discrimination task was designed.
• Foils were created by applying topological, projective, affine, and Euclidean transformations to standard objects.
• Observer sensitivity thresholds for each transformation type were measured.

Main Results:

• Perceptual salience decreased in the order: topological, projective, affine, and Euclidean transformations.
• This order directly corresponds to the levels of the Klein hierarchy of geometries.
• Common shape-difference metrics did not correlate with observed perceptual thresholds.

Conclusions:

• Human visual perception of shape transformations is structured according to the Klein hierarchy.
• The visual system prioritizes more fundamental geometric properties (like topology) over less fundamental ones.
• Current computational shape-difference metrics may not fully capture human perceptual judgments of shape transformations.