A method of segmenting vessels by morphological filters and dynamic thresholding for digital subtraction angiography (DSA) images is presented. The first step is to reduce the noise and enhance the details of image by using morphological operators. The second is to segment vessels by dynamic thresholding combined with global thresholding based on the properties of DSA images. Artificial images and actual images have been tested. Experiment results show that the proposed method is efficient and is of great potential for the segmentation of vessels in medical images.