ABSTRACT
Based on a keynote lecture at CompImage 2006, Coimbra, Oct. 20–21, 2006, an overview is given of our activities in modelling and using surround inhibition for contour detection. The effect of suppression of a line or edge stimulus by similar surrounding stimuli is known from visual perception studies. It can be related to non-classical receptive field (non-CRF) inhibition that is found in 80% of the orientation selective neurons in the primary visual cortex. A computational model of surround suppression is presented. It acts as a feature contrast computation for oriented stimuli: the response to an edge at a given position is suppressed by other edge responses in the surround. Consequently, the responses to texture edges are strongly reduced while the responses to contours are scarcely affected. The model gives results that are in line with perception. A surround suppression step is added to a Gabor energy filter and to the Canny edge detector. In either case it improves considerably the detection of contours. The biological utility of the neural mechanism of surround inhibition might be that of quick pre-attentive detection of object contours in natural environments rich in texture. In computer vision, a surround suppression step can be added to virtually any edge detector with limited local support in order to improve its contour detection performance.
