Computing the curve-skeletons of images

Abstract

In this paper, an automatic algorithm that extracts a curve-skeleton (1-D centreline) from 2-D binary images is described. From the discretized data of the image/object, the potential field is computed resulting in a vector field defined over the entire object. The topology of the vector field is analysed and path-lines are seeded to extract the curve-skeletons. Seeds start from high-curvature boundary values, high-divergence values, and saddles. A hierarchy of curve-skeletons results and different curve-skeletons can be chosen based upon the application. The skeletonization algorithm presented is robust to noise and generates smooth skeletons with topological characteristics. These are the main contributions in this paper and this work is an extension of our previous work on thinning of binary images using divergence of the potential field. The algorithm presented in this paper was tested on numerous 2-D images and the results are shown.

Keywords:

• Skeletonization
• Potential field
• Centreline
• Curvature
• Critical points

Acknowledgements

One of the authors, Balasubramanian Raman, gratefully acknowledges the financial support of the Sponsored Research and Industrial Consultancy (SRIC), Indian Institute of Technology, Roorkee, India under the Grant Number FIG-100421-MAT.