A scheme for morphological skeleton pruning

摘要

This paper reports a pruning scheme specifically for morphological skeletons. We polished the boundary/axis ratio significance measure by H. Blum and R. N. Nagel to a contour length significance measure, CLSM. We analyzed its advantages compared to two recently proposed pruning approaches, and designed an adequate algorithm for it, based on a firm mathematical MAT theory by H. I. Choi et al. After a thorough analysis on skeletons generated by a typical morphological skeletonization function, Matlab bwmorph, we proposed and implemented suitable preprocessing techniques such as single lining and end extending. We then formulated a perfect pruning scheme and showed its power via experiments on a standard data set, the Part B of the MPEG-7 CE-Shape-1 data set. Algorithmic analysis showed that the scheme has linear computational complexity with respect to n, the pixel number in the initial skeleton.