Perceptual Salience of Contours Due to Horizontally-Mediated Synchronization of Modelled Cortical Cells
Shih-Cheng Yen and Leif H. Finkel
Department of Bioengineering,
University of Pennsylvania,
Philadelphia, PA.
Abstract
Purpose. We present simulations of a cortical network which accounts for recent experimental results on contour salience (Polat and Sagi, 1993; 1994; Kapadia et al., 1995; Field et al., 1993; Kovács et al., 1996; Pettet et al., 1996).
Methods. Based on anatomical data, we employ two zones of excitatory connections, one flaring out along the orientation axis of the cell (co-axial) and the other extending orthogonally to this axis (trans-axial). Excitatory horizontal connections facilitate cells which receive strong input to their classical receptive field. Inhibitory horizontal connections limit these facilitatory effects to a spatial domain, the size of which scales inversely with the density of stimuli in the image. Facilitated cells enter a "bursting mode" (Gray and McCormick, 1996) in which they can synchronize with other, strongly connected, bursting cells. The perceptual salience of a contour is determined by the summed activity of the synchronized set of cells responding to the contour.
Results. The synchronization mechanism can account for results of Kovács and Julesz (1993) showing that smooth closed contours are significantly more salient than open contours. In simulations, synchronization occurs faster and at greater inter-element separations in response to closed versus open chains of stimuli. In addition, as reported psychophysically, we observe a precipitous change in salience when the last element of closed circular chain is added. These results appear to hold for a variety of synchronization mechanisms: phase-coupled oscillators (Kopell and Ermentrout, 1986), relaxation oscillators (Sommers and Kopell, 1993) and burst-based mechanisms (Traub et al., 1996).
Conclusions. Synchronization mechanisms may be used to detect global object properties such as closure, as well as measuring salience in complex images.
Supported by ONR N00014-93-1-0242 and The Whitaker Foundation