Optic Disk Segmentation Using Histogram Analysis

Abstract

In the field of disease diagnosis with ophthalmic aids, automatic segmentation of the retinal optic disc is required. The main challenge in OD segmentation is to determine the exact location of the OD and remove noise in the retinal image. This paper proposes a method for automatic optical disc segmentation on color retinal fundus images using histogram analysis. Based on the properties of the optical disk, where the optical disk tends to occupy a high intensity. This method has been applied to the Digital Retinal Database for Vessel Extraction (DRIVE)and MESSIDOR database. The experimental results show that the proposed automatic optical segmentation method has an accuracy of 55% for DRIVE dataset and 89% for MESSIDOR database

Optic Disc Localization in Retinal Fundus Images Based on You Only Look Once Network (YOLO)

“…From Table 3 we conclude that all the previous method has various accuracy for each type of dataset, while the proposed method has 100% accuracy for all datasets. The approaches presented in Table 3 can be categorized into five main classes: approaches based on OD features such as brightness and circularity (article [29]), approaches that relied on just vascular extraction (article [4]) or combined with OD characteristics (article [8]), approaches that employed some transformation (articles [7,27]) or analysis image features for extraction OD (article [5]) and companies with first class, approaches that utilized the template matching technique for OD (article [28]), and approaches that utilized the deep learning for localization OD (article [30]).…”

“…Fig. 6 shows a sample of the OD detection and cropping by [29] for the image in the Messidor1 dataset, compared with the proposed method.…”