Diabetic Retinopathy is a severe and widely spreadeye disease which can lead to blindness where one of the main symptoms for vision loss is Exudates and it could be prevented by applying an early screening process. In the Existing systems, a Fuzzy c-Means Clustering techniques is used for detecting the exudates. The main objective of this paper is efficiently detecting the Exudates. In this a Three Stage – [TS] approach is introduced for detecting and extracting the exudates automatically for screening the Diabetic retinopathy. TS have functions on the image in three levels like Preprocessing, Image Enhancement and Detecting the Exudates. The detected exudates are classified using GLCM method for finding the accuracy. The TS approach is experimented in MATLAB software and the performance evaluation can be proved by comparing the results with the existing approach’s result and hand-drawn ground truths from expert ophthalmologist.