An integrated approach to automated analysis of coronary lesions from cineangiograms is being developed. A vessel segment between two selected endpoints is tracked with a sector-search method. Tracking points in the vessel are located by examination of the pixel values along a box perimeter over a limited range defined by an angular sector which is oriented toward one of the endpoints. Tracked points are fitted by a polynomial curve that yields a smooth vessel centerline, which is used for generation of a straightened vessel image. The image is then corrected for the nonuniform background in the cine image, which is estimated by a two-dimensional fit of the region surrounding the vessel. For each line in the corrected image, the vessel size is determined by the iterative deconvolution technique, which takes into account the line-spread function of the imaging system. With this integrated approach, stenotic lesions can be assessed accurately, even in the presence of complicated vascular anatomy, such as bifurcations, branches, or overlapping vascular structures.