A Neuronal Model Detecting Figure—Ground Direction and Contrast Polarity at Illusory Contours

A previous model of visual cortical contour processing simulated the perception of illusory contours and the direction of figure and ground at such contours (Heitger and von der Heydt, 1993 Proceedings of International Conference on Computer Vision). This model was based on ‘complex type’ operators. Hence, it lacked representations of contrast polarity and could not account for brightness effects of illusory figures.

Recently, we found that some neurons of area V2 in monkey visual cortex that signalled illusory contours were also sensitive to the figure — ground direction and contrast polarity at such contours. This allowed us to extend the previous scheme by incorporating the signals of even and odd symmetrical ‘simple type’ operators, particularly at the level of the ‘end-stopped’ operators which served the detection of the inducing elements of illusory figures (line ends, corners). To represent illusory contours, the signals of these operators were grouped with the use of fields of wide spatial support. Line ends and corners were handled differently in this process: at corners, signals parallel to the illusory contour were emphasised (para grouping), whereas at line ends signals orthogonal to the contour (ortho grouping) were emphasised. Further, ortho grouping included only end-stopped operators with even-symmetrical fields, para grouping only operators with odd-symmetrical fields. Thus, both processes contributed to the contrast polarity represented at illusory contours.

Results of simulations showed that this model reproduces the shapes of illusory contours, their figure — ground direction, and contrast polarity. Hence, it can explain the selectivity of cortical neurons at such contours and the perceived brightness of illusory figures.