Local and Global Gestalt Laws: A Neurally Based Spectral Approach

Area of Science:

• Computational Neuroscience
• Visual Perception
• Mathematical Modeling

Background:

• Figure-ground articulation is crucial for visual perception.
• Existing models often lack integration of local and global Gestalt principles.
• Understanding the primary visual cortex (V1) functional architecture is key.

Purpose of the Study:

• To present a novel mathematical model of figure-ground articulation.
• To incorporate both local and global Gestalt laws.
• To ensure compatibility with the functional architecture of the primary visual cortex (V1).

Main Methods:

• Modeling the local Gestalt law of good continuation using Lie group theory-derived connectivity kernels.
• Quantitatively comparing these kernels with V1 long-range connectivity.
• Introducing global Gestalt constraints via spectral analysis of a connectivity matrix.

Main Results:

• The model successfully integrates local and global Gestalt principles.
• Spectral analysis effectively groups local features and identifies salient perceptual units.
• Numerical simulations demonstrate the model's applicability to various stimuli.

Conclusions:

• The proposed mathematical model offers a unified framework for figure-ground articulation.
• It provides insights into the neural mechanisms underlying visual perception in V1.
• The model is a valuable tool for studying visual processing and perceptual organization.