Retinal images give unique diagnostic information not only about eye disease but about other organs as well. Retinopathy is a diabetic-related complication which is the main cause of vision loss of the diabetic patients and the prevalence is rising day by day. Early detection of glaucoma can prevent vision loss by regular screening and proper treatment. In order to overcome the difficulties, an automated diagnosis of glaucoma methods are preferred for glaucoma diagnosis. The extraction of robust features are plays an important role in developing a robust system. The former techniques are expensive and require experienced clinical persons to use them. So, there is a need to diagnose glaucoma accurately with low cost. Hence, in this proposed system we implement the new technique in which the Empirical Wavelet Transform is applied on the images to form the Sub band which is also called as the decomposed images. Then from the decomposed image Correntropy features are obtained. These extracted features are normalized based on the feature values by feature selection process. Then, these features are used for the classification of normal and glaucoma images using Least Squares Support Vector Machine (LS-SVM) classifier. After the classification process the performance of the system is evaluated by using the Performance measure values such as Accuracy, Sensitivity, Specificity, Precision, Recall, F-measure values. And these performance results shows that the proposed system which shows the better results than the Existing systems.