Mathematical models of visual perception based on cortical architectures

Abstract

We present a mathematical model of figure-ground articulation, which takes into account gestalt laws and is compatible with the functional architecture of the primary visual cortex (V1) to obtain low-level object segmentation. Connectivity kernels, derived from Lie group theory, are used to describe the gestalt law of good continuation. The spectral analysis of connectivity matrix derived from these kernels allows to individuates perceptual units with the highest saliency. Some applications of this model to the problem of individuation of perceptual units in illusory figures are presented, as clear examples of problems of visual perception.