Isthmus-based 6-directional parallel thinning algorithms

Abstract

Skeletons are widely used in pattern recognition and image analysis. A way to obtain skeletons is the thinning approach, consisting in iteratively removing points from the object without changing the topology. In order to preserve geometric information, it is usual to preserve curve end points (for curve skeletons) or surface end points (for surface skeletons). In this paper we propose a new fast directional parallel thinning scheme, preserving isthmuses (a generalization of curve/surface interior points), and providing skeletons with low amount of noise. We also prove the topology preservation of our approach. © 2011 Springer-Verlag.