A Skeleton Pruning Approach Using Contour Length as the Significance Measure

To understand image by means of pervasive devices, efficient high quality skeleton producing scheme is highly expected. This paper introduces a novel significance measure, the contour length measure for skeleton pixels, abbreviated as CLMSP, which measures the part of a contour supporting a skeleton pixel and is used for skeleton pruning. This measure is shown exhibiting features of significance representation, superfluous hairy branch differentiation capacity and fairness. To ease the search of the nearest contour pixels related to a skeleton pixel, all the contour pixels are organized into a kd-tree. Complexity analysis combined with experimental practice shows that the average complexity of the algorithm is about nlogn, the complexity of building the kd-tree, where n is the number of contour pixels. Applying the approach to skeletons generated by morphological thinning, with certain smoothing on both searching of nearest contour pixels and distance computing, high qualified skeletons have been obtained. Experiments also demonstrate that the approach has high stability against affine transformation and strong noise removal capability.