Any diseased retinal image has three different color groups. Blood vessel tree and haemorrhages are dark red;
Optic Disk, exudates in Diabetic retinopathy and drusens in AMD are bright yellow; mucus and membrane of the
retina are of yellowish red color. So we select the number of clusters as three. After three different clusters aregenerated, then, depending on the abnormality type, an appropriate cluster is selected. If micro aneurysms or
haemorrhages are to be detected from Diabetic Retinopathy images, then the cluster containing the dark red portion
should be extracted from the clustered image and the region of interest should be selected from that. If the exudate is
a region of interest, then the cluster containing bright yellow color is selected. The flow of the algorithm is depicted
in Fig. 1. Images from the public databases were divided into two groups, one consisting the training set comprising
of 10 images of each abnormality ( 60 images) and the other the test set (79 images) out of a total of 139 images.