Skeletonization based on the medial-axis and symmetry information

Abstract

The classical and potential thinning issues such as bias effect, boundary noise immunity, and even rotation invariant, are quite worthy of studying in image processing field. In this paper, based on the fundamental medial axis transformation (MAT) concept, we developed an approach including extraction of medial axis with symmetry information, clustering, linking, and post rule-based thinning to investigate the possibility of reducing the bias effect and increasing the boundary noise immunity. A rotation invariant rule-based thinning algorithm was adopted for experimental comparisons. The primary result confirms that the proposed approach is one of the feasible directions to overcome these issues.