September 2005
Volume 5, Issue 8
Vision Sciences Society Annual Meeting Abstract  |   September 2005
A neural network model of Gestalt-like visual processing
Author Affiliations
  • Maria Pietronilla Penna
    University of Cagliari, Dept. of Psychology, Italy
  • Anna Montesanto
    University Politecnic of Marche,DEIT, Italy
  • Vera Stara
    University of Cagliari, Dept. of Psychology, Italy
  • Massimiliano Dasara
    University of Sassari, Dept. of Sciences of Languages, Italy
Journal of Vision September 2005, Vol.5, 975. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Maria Pietronilla Penna, Anna Montesanto, Vera Stara, Massimiliano Dasara; A neural network model of Gestalt-like visual processing. Journal of Vision 2005;5(8):975.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

The perceptual phenomena evidenced by Gestalt theory can be explained by assuming the existence, within perceptual space, of a suitable vector field. The latter would act upon the single stimulation elements, by organising them into wholes, endowed with properties of global nature. The contours of perceived patterns would coincide with suitable lines of force associated to the equilibrium configurations of this field which, in turn, should depend on the form of the boundaries of perceptual space itself. Stadler proposed to detect these lines of force by resorting to a task in which a subject must graphically reproduce on an empty sheet of paper a point previously observed on another sheet. The displacement between the reproduced and the observed point lets us individuate the direction and the magnitude of the tangent vector to the line of force at a given position.


In this work we introduce a neural network model able to reproduce the performance of human subjects in the task quoted above and to represent Gestalt-like properties of spatial pattern processing. This model is based on a general architecture consisting of: a retina receiving input pattern, a spatial memory designed to process retinal output values, and a system of filtering networks designed to detect the domains of the spatial scene whose properties are the most important for the task to be done by the whole system. Our model includes also a motor network describing motor action issued by subjects as a consequence of their spatial memory content. Model's performance has been tested through a comparison of its motor output with the one of human subjects on visual pattern reproduction.


The findings obtained from the latter evidenced that the majority of lines of force thus individuated crossed in a small number of points, to be identified with the attractors of the perceptual field. We found clear evidence for the presence of two attractors located near the two corners on the upper part of the sheet.

Penna, M. Montesanto, A. Stara, V. Dasara, M. (2005). A neural network model of Gestalt-like visual processing [Abstract]. Journal of Vision, 5(8):975, 975a,, doi:10.1167/5.8.975. [CrossRef]

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.