Skeletonization algorithm using discrete contour map

Abstract

The skeleton of a binary object can be considered as an alternative to the object itself; it describes the object in a simple and compact manner that preserves the object topology. In this paper, we introduce a new definition for discrete contour curves, and we propose a new approach for extracting a well-shaped and connected skeleton of twodimensional binary objects using a transformation of the distance map into contour map, which allows us to disregard the nature of the distance metric used. Indeed, our algorithm can support various distances such as the city-block distance, the chessboard distance, the chamfer distance or the Euclidean distance. To evaluate the proposed technique, experiments are conducted on shape benchmark dataset.