Fast Fundus Optic Disc Localization Based on Main Blood Vessel Extraction

Optic disc (OD) is one of the most important structures in retina. The automatic detection of OD is essential to make Diabetic retinopathy examination more effective. In this paper, a fast optic disc localization algorithm is described. Our approach begins with pre-processing, and then the main vessels in fundus images are extracted based on morphology methods. In the next step, parabolic fitting algorithm by the least square method is applied to express the main vessels. Finally, we judge the fitting result and defined the apex as the OD center. The proposed method was evaluated in 3 public datasets. An average accuracy of 99.05% at the average time of 1.08 second was achieved.