Extracting retinal vessels from the retinal image is important to the diagnosis of hypertension and arteriosclerosis in clinic. This paper presents a method which can segment retinal vessels effectively with a group of nonlinear image processing algorithms. The procedures of the whole operation are: 1 )Adaptive image enhancement with image’s statistic features.2)Adaptive image segment based on blood model. and 3)Extracting vessels’ centerlines via distance transform. The results of .experiment indicate that the method is really suited for non-fluorescence retinal image.