An Efficient Quick Algorithm for Computing Stable Skeletons

A new method to obtain high quality skeletons of binary shapes is proposed in this paper. First, a small set of salient contour points is computed by Discrete Curve Evolution (DCE). These salient points are the stable endpoints of the skeleton. Second, the skeleton is grown between pairs of the endpoints. Examining every eight-connected point of the current skeleton points, Select the point, that have equal distance to the contour parts which are partitioned by the two endpoints, as the new skeleton point. The skeleton path continues growing in this way until it reaches the other endpoint or another skeleton branch. The main idea is that the skeleton points are always the center of the maximal disks, and the endpoints of the skeleton are those contour points with high global curvature which is stable to noise and shape variations. The obtained skeletons are in accord with human visual perception and stable, also connected and one pixel thick. They do not require any pruning or any other postprocessing. The experimental results clearly demonstrate that the proposed method significantly outperforms other well-known methods for skeleton computation.