"Pop-Out " of Salient Contours in a Network based on Striate Cortical Connectivity

Shih-Cheng Yen and Leif H. Finkel

Department of Bioengineering,
University of Pennsylvania, Philadelphia, PA.

Abstract

Purpose. It has been suggested that long-range intrinsic connections in striate cortex modulate local contrast sensitivity and contour extraction (Ito, et al., 1995; Knierem and van Essen, 1992). We have developed a network architecture to investigate the role of recurrent excitation and normalization in perceptual pop-out.

Results. Network response to an oriented segment can be enhanced or suppressed depending upon the texture in the stimulus surround. In agreement with the psychophysical results of Field, et al., (1993) and Polat and Sagi, (1994), the network significantly enhances the responses to segmented contours embedded in a field of distracter elements if these contours are perceptually salient. The maximum spacing between separate segments on a contour, Dmax, is greater for contours with constant curvature (circles and lines) than for "snake"-like contours, as supported by the results of Kovacs and Julesz (1993).

Conclusions. Intrinsic striate connectivity may be responsible for several of the Gestalt laws, including good continuation and similarity/proximity. When applied to real images the network emphasizes those contours which are judged most salient by human observers. Such an algorithm can be used to attract visual attention to particular image features in image processing applications.

Supported by ONR N00014-93-1-0242 and The Whitaker Foundation