An Automated Segmentation Approach from Colored Retinal Images for Feature Extraction

Segmentation from colored retina images plays a vital role in stable feature extraction for image registration and detection in many ocular diseases. In this study, the authors will look at the segmentation of the blood vessels from fundus images which will further help in preparation of digital template. Here, images are passed through the preprocessing stages and then some of the morphological operators for thresholding are applied on the images for segmentation. Finally, noise removal and binary conversion complete the segmentation method. Then, a number count on blood vessels around the optic disk is done as a feature for further processing. The authors will ensure whether the segmentation accuracy, based on comparison with a ground truth, can serve as a reliable platform for image registration and ocular disease detection. Experiments are done on the images of DRIVE and VARIA databases with an average accuracy of 97.20 and 96.45%, respectively, for segmentation, and a comparative study has also been shown with the existing works.