Abstract: In automatic retinal image analysis, the location of the optic disk and macula fovea is the prerequisite for a system of computer aided diagnosis or automatic screening of diabetic retinopathy. This paper proposes a novel method to detect macular firstly before optic disk location or vessels detection using directional local contrast filter and local vessel density feature, which improves effectively the accuracy of the detection of the macular and optic disk, and is different from existing algorithms that detected optic disk firstly before macula fovea location. 169 color images with diabetic macular edema from HEI-MED public database are used to evaluate the performance of the developed method. From the experimental results, the accuracy rate both of macular and optic disk detection is found to be 98.2%. The proposed method is simple and unsupervised, and has high practical value.