2D thinning algorithms with revised endpixel preservation

Abstract

Skeletons are shape descriptors that summarize the general forms of objects. Thinning is a frequently applied technique for digital binary pictures to extract skeleton-like shape features. Most of the existing thinning algorithms preserve endpixels that provide relevant geometrical information relative to the shape of the objects. The drawback of this approach is that it may produce numerous unwanted side branches. In this paper we propose a novel strategy to overcome this problem. We present a thinning strategy, where some endpixels can be deleted.