Modeling cortical mechanisms of border ownership coding

Purpose: Border ownership is the phenomenon that contrast edges, lines, or illusory contours are perceived as belonging to one of the adjacent image regions as if they were contours of a 3D object. The aim of this study was to develop a model that accounts for the main findings on neural border ownership coding (Zhou et al., 2000) and can be expanded to explain object-based attention. Methods: Numerical solution of systems of linear equations. Results: The model uses an array of edge detector cells E in which each position and orientation is represented by a pair of cells corresponding to the two sides of ownership. Edges excite both members equally, T junctions excite only one member of a pair. Grouping cells G sum the signals of E cells according to roughly circular templates and enhance the gain of the same E cells. Each E cell communicates only with G cells on one side of its receptive field (the ownership side), and the two members of an E pair inhibit each other. Thus, two groups of G cells effectively compete at each edge location. Common activation of the two E cells represents contour strength, differential activation represents border ownership. With two sizes of templates the model assigns border ownership correctly for squares and rectangles over a range of sizes, as well as for C-shaped figures and for different configurations of overlapping figures. Furthermore, the G cells provide handles for object-based selection. By activating a G cell one can enhance the edge signals of one of the figures in the display. Conclusion: A simple model explains how the visual cortex assigns border ownership and how visual cortical mechanisms might provide a structure for object-based attention.