COOPERATION BETWEEN EDGES AND JUNCTIONS FOR EDGE GROUPING

Edge detection is a fundamental stage in order to facilitate the analysis or the interpretation of an image. However classical edge detectors usually yield gaps in the contour image. To restore incomplete contours, criteria based on perceptual grouping have been proposed. Almost all existing closing algorithms based only on these criteria fail at discontinuity points junction. In this perspective we propose an original algorithm which combines perceptual grouping criteria and junction points. Furthermore, grouping is done by preprocessing iteratively in a bottom-up, local-to-global fashion.