A new thinning algorithm for binary images

Thinning plays a crucial role in image analysis and pattern recognition applications. It is one of the most frequently used pre-processing methods to analyze different types of images. Thinning consists basically of reducing a thick digital object into a thin skeleton. There are several thinning algorithms for getting a skeleton of a binary image in the literature. The most popular, and well proved one is the ZS algorithm proposed by Zheng and Suen. In the present paper, we propose a new thinning algorithm which combines the directional approach used by ZS and the subfield approach in order to produce a new hybrid thinning algorithm which is more efficient, produces thinner results (skeleton thickness is equal to one) than the ZS algorithm and solves the ZS´s loss of connectivity problem in 2×2 squares. Results of applying the proposed algorithm on a variety of binary images and comparison with ZS algorithm show better results in terms of thinning rate, thinning speed, visual quality and connectivity preservation.