Abstract: In this study, an automated screening system to diagnose the severity of diabetic retinopathy is recommended. The proposed system consists of 3 stages; the preprocessing being the first one is done to make it reliable for extracting features. In the second stage, features like area of blood vessels, exudates, micro aneurysms and texture features are extracted from the retinal images and classification, the last stage is done using the ELM classifier. The above procedures were implemented and evaluated using images available in DIARETDB1 and DRIVE database. Our proposed method shows a high accuracy of 95% and overcomes the slow training speed when compared with other classifiers.