The technology of automated detection of diabetic retinopathy (DR) on fundus retinal images can not only make mass screening possible, but also offer a powerful adjunct for early diagnosis, treatment of diabetic retinopathy, and scientific research on human vision. For this purpose, an algorithm based on mathematical morphology for automated detection of diabetic retinopathy was proposed. Firstly, the optic disc was segmented by mathematical morphology and threshold in order to find candidate regions possibly containing lesions. Secondly, some methods such as morphological reconstruction were applied to find the exact contours of lesions. Finally, the true lesions were found out exactly. Experimental results showed that the algorithm was fast and effective in detecting diabetic retinopathy of fundus retinal images.