A novel method for vessel skeleton extraction

A novel vessel skeleton extraction method is presented in this work. Our work consists of three steps in a coarse-to-fine style: Firstly, by modeling the distance transform and its gradient vector field, the average outward flux of the gradient vector field is computed to coarsely label all image points. Then we introduce a topology-based shape thinning algorithm for extracting vessel skeleton tree. At last, a skeleton tree refinement algorithm is applied to get a precise extraction of vessel skeleton. The proposed method is parameter-free and computationally efficient. The validity and efficiency of the method is tested on two public database of human eye retina vessel images.