A Bayesian observer replicates convexity context effects in figure-ground perception.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Visual Perception

Background:

• Figure-ground perception is crucial for object recognition.
• Convexity context effects, where perception of convexity influences figure-ground assignment, were previously observed but their origins were unclear.

Purpose of the Study:

• To develop a computational model that replicates convexity context effects in figure-ground perception.
• To investigate the role of environmental regularities in visual perception.

Main Methods:

• A two-parameter Bayesian observer model was created.
• The model incorporated prior expectations about object convexity and background color homogeneity.
• The observer estimated the probability of a 3D scene and figure status for convex regions, responding stochastically.

Main Results:

• The Bayesian observer successfully replicated convexity context effects.
• These effects were observed only when concave regions were homogeneously colored, mirroring human performance.
• The model's performance approximated that of average human subjects.

Conclusions:

• Object convexity and background color homogeneity are likely environmental regularities utilized by the human visual system.
• Figure-ground perception may involve interpreting ambiguous visual input based on these expected regularities.
• The proposed Bayesian observer provides a plausible mechanism for these perceptual phenomena.