A Method for Quantitative Comparison of 2D Skeletons

Skeletons are widely used shape descriptors which summarize the general form of binary objects. There exist numerous skeletonization techniques that produce various skeleton-like features for the same object. Despite of the fact, that some researchers have made efforts to compare skeletons and evaluate skeletonization algorithms, we propose a new similarity measure that is based on the concept of normalized distance maps. In addition, a novel method for the quantitative comparison of skeletons is also presented. The reported method uses a high resolution dataset containing pairs of elongated objects and their expected skeletons. Our method is validated with the help of generalized morphological skeletons driven by neighborhood sequences. Based on the proposed method, we compared and ranked nineteen existing 2D thinning algorithms.