, Arun Kumar Patel

Abstract — Optical disc segmentation and localization is an essential task for diabetic retinopathy and retinal image analysis. The pervious methods of thresholding using histogram equalization are unable to threshold the disc for diabetic images, since they contain multiple connected regions. This paper presents a simple and efficient approach for segmentation of the optic disc in such retinal images using thresholding the morphologically connected regions and boundary extraction. The input optical image is enhanced using Bi-histogram equalization. This improves the efficiency of the conventional thresholding method. Gray level thresholding is applied on the enhanced retinal images. The desired optical disc is determined by finding the label connected neighbourhood with 8-conect mask. Then morphological boundary extraction and circle fitting method is used for extracting the optical disk. The method is tested on the 40 different optical disc images. It is found to be less complex and efficiently recognizes the boundary of the optic disc and also evaluates the optical canter. The optical canters are tabulated for original and corrupted images. The results are also compared with the existing standard methods