Due to the complex nature of vascular structures, the best way to represent and analyse the topology of vasculature is to use its skeleton curves, as the skeleton of an object has the ability to naturally capture important shape characteristics in three-dimensional contexts. Thus, the extraction of vascular tree skeleton plays an important role in the area of computer-aided vascular surgery, such as the reconstruction of vasculatures, virtual angioscopy, and so on. This paper presents a simple and automatic method to extract the vascular tree skeleton from segmented vessel datasets. In the proposed method, the segmented vessel dataset is re-initialized to be a Signed Distance Function (SDF). And then, the moving sphere along the vessel tree can easily and automatically detect bifurcations and predict the location of next skeleton point with the constraint of SDF. Some experiments and application have been carried out to demonstrate the strengths of our proposed method.