Objective To segment coronary arteries from coronary computed tomography angiography (CTA) images of patients and realize the three-dimensional visualization, so that doctors can observe patients’ coronary arteries in three-dimension and are able to appeal diagnosis and treatment to cardiovascular diseases caused by coronary artery lesion. Methods This paper proposed a semi-automatic method of vessel segmentation based on traditional region growth algorithm. Firstly, Preliminary growth of the aorta based on gray value was proposed. Then the coronary arteries were segmented by regional growth based on self-adaption threshold. Finally, we did certain post processing. Results The segmentation results demonstrated that three-dimensional visualization modeling results in this paper could provide clear information about vessels’ size, shape, and whether stenosis and cut. Also, the results could be used for further related computation. Conclusions This semi-automatic vessel segmentation based on traditional region grow algorithm its adaption and effectiveness and makes a better segmentation of coronary arteries.