Subjective contours are useful for extracting contours with very weak contrasts

Abstract

Certain figures cause one to perceive clearly visible contours (subjective contours) in the region where there are no physical contrasts in brightness or color. An algorithm is proposed for generating subjective contours automatically from given figures. The algorithm is based on the following idea: 'L' or 'I' type vertices are produced on physical contours because they are occluded by a subjectively perceived plane. And these vertices are candidate points where subjective contours will pass. Then pairs of points to be filled in with subjective contours are determined according to factors of proximity and similarity in Gestalt's rules. The algorithm was implemented as a procedure reflecting neurophysiological and perceptual findings and applied to typical figures successfully. Furthermore, the algorithm was applied to real images where leaves overlapped one another with extremely feeble inner contours. The algorithm could extract the inner contours and separate each leaf.