In this paper, a model based segmentation method is introduced for extracting blood vessel structures in poor quality coronary angiograms. This method extracts blood vessels in the angiograms by exploiting the spatial coherence existing in the image. Here, a circular sampling method is employed to exploit the coherence. This method uses a collection of 2D patterns to represent the 3D structures of vessels. The segmentation method employs the circular sampling method to produce the 2D slice samples at certain depths on each pixel on a varying background on the image, so several 2D sample slices of the 3D pattern of blood vessels are collected. These 2D slices are compared with certain original patterns in order to check whether a slice is part of a blood vessel. Finally, results from several overlapping 2D slices are evaluated collectively and checked whether they represent a 3D blood vessel histogram. To produce the final segmented image, incorrectly segmented noisy parts and discontinuous parts are eliminated by using circular filtering methods. The performance of the method is examined on various qualities of X-ray angiograms and synthetic images. Results indicate that the proposed method yields a good performance in automatic segmentation of such images.