Centerline Extraction of 2D Vessels Based on Wavelet Magnitude

Centerline extraction is very important in interactive navigation and virtual endoscopy. However traditional centerline extraction techniques, such as minimal path, are sometimes incapable of providing a precise approximation of the ideal centerline of vessels because the minimal cost path is the shortest Euclidean path linking two user-supplied points. In this paper, by using wavelet magnitude, we propose a new algorithm to obtain the central region of the vessels and central points of the vessel junctions. Then the minimal path technique is used to link obtained central points and the user-supplied points within the central region. Experiment results on synthetical and clinical images show that our algorithm can extract high-quality centerline of the vessels.