Skeletonization of two-dimensional shapes via fast numerical calculation of vector fields

摘要

We propose an approach for efficient two-dimensional skeletonization of binary shapes through numerical calculation of vector fields and curvature estimation by using the Weingarten formulae. It can be shown that potential valleys generated by vector fields have a close relationship with the definition of intensity axis of symmetry. Given a binary image, the algorithm consists in generating a grayscale image corresponding to the magnitude of a vector field followed by a search of the points that belong to the bottom of the potential valleys or regions with minimum magnitude. It can be shown that these points provide a good approximation to the medial axis of the object in study. Also, the proposed method demonstrated good performance due to the fact that the vector field can be easily and rapidly calculated using the fast Fourier transform algorithm.