An approach for reconstructing the three-dimensional geometry of the coronary arteries from a limited number of two-dimensional angiographic views and producing a computer model which can be used to quantify and visualize the coronary vasculature is presented. The approach incorporates a hierarchical database that organizes geometric information from arbitrary arterial structures in a natural fashion. Detecting and locating vascular structure from angiographic images are partially automated through entropy edge detection techniques. The back projection algorithm accurately models the nonparallel geometry inherent in X-ray biplane angiography and produces a quantitative description of vascular structure, including estimates of vessel diameters aid lengths. The approach has been applied in phantom studies, in which quantitative comparisons were made, and patient studies, in which qualitative evaluations were made.<<ETX>>