Abstract: The standard gradient-based edge detectors react to all luminance changes, irrespective of whether they are due to the contours of the objects represented in a scene or due natural textures like grass, foliage, water, and so forth. To improve their performance in detection of object contours and region boundaries in natural scenes, an algorithm called switching surround suppression was proposed which can be easily incorporated into them. Different from previous methods that deploy surround suppression, our method only exerts suppression on texture edges while leaving contours unaffected, which further improves the contour detection performance of standard gradient-based edge detectors. The proposed method was evaluated by the natural images and the standard reference contour edge images. The results showed that the proposed method is better than the standard gradient-based edge detectors as well as other methods that deploy surround suppression.

Key words: information processing, edge detection, contour detection, SUSAN criterion, switching surround suppression