A combinatorial pruning algorithm for Voronoi skeletonization

Abstract

Voronoi skeletons have been used extensively in image processing and analysis due to its fast computation and good properties. However, they are very sensitive to boundary noise which may cause a large number of insignificant branches that need to be pruned. Commonly used measurements of significance can be divided into two types: local and global. Local measurements of significance are context-aware but sensitive to noise. Global measurements of significance are robust to noise but unaware of context information. In this paper, we propose a combinatorial branch pruning algorithm that integrates both local and global measurements. Experimental results show that the proposed method is stable with different shapes and robust to boundary noise. © 2013 Springer-Verlag.