Extracting Noise-Resistant Skeleton on Digital Shapes for Graph Matching

Abstract

In order to match shapes using their skeletons, these ones should be thin, robust to noise, homotopic to the shape, consequently, connected. However, these properties are difficult to obtain simultaneously when the shape is defined on a discrete grid. In this paper, we propose a new skeletonization algorithm, which has all these properties. Based on the Euclidean distance map, the algorithm extracts the centers of maximal balls included in the shape and uses the ridges of distance map to connect them. A post-processing is then applied to thin and prune the resulting skeleton. The proposed method is compared to three fairly recent methods to highlight the good properties of the obtained skeleton.KeywordsGraph MatchSkeleton PointMaximal BallElliptic ParaboloidEuclidean Distance TransformationThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.