Vascular segmentation based on CT images plays an important role in many practical applications, such as disease diagnosis, surgical planning and so on. Due to limitations of individual differences and image forming apparatus, angiographic images still remain low contrast and strong noise. The paper provides a method of vessel segmentation, which first do the image preprocessing using histogram, then improve vesselness function of the vessel enhancement using Hessian matrix to make it more sensitive to small and fuzzy tubular structures, finally region growing algorithm is employed to extract the richer in vessel branching. The experiments proved the segmentation method can be achieved vessel segmentation accurately. It can avoid the error segmentation effectively and has a better robustness.