Abstract: Blindness is a growing problem worldwide. The major causes of blindness are glaucoma and diabetic retinopathy. Increased intraocular pressure causes glaucoma. In glaucoma detection, it is very difficult to identify the edge of the optic cup because the image is blurred where blood vessels pass through the optic cup. Current methods do not effectively address the issue of peripheral blurring of the blood vessels surrounding the optic cup. In this paper, it is recommended to automatically detect glaucoma in retinal images using an efficient method. Initially, optic disk and cup segmentation is done by the Convolution Neural Network (CNN) and Modified Region Growing Mechanism (MRG). Then, texture features are extracted from the separated results. Finally, a neural network is used to diagnose glaucoma. The experimental results demonstrate that the proposed approach achieves better glaucoma detection result (accuracy, sensitivity and specificity) compared to few other approaches.