A Novel Scheme for the Fovea Localization on Retinal Images

Fovea is at the center of the macula, hard to be observed, but the mainly occurring region of diabetic retinopathy. Parabola-like vascular geometry model is the most common way used to detect fovea. Most studies took the center of an optic disc (OD) as the vertex of the parabola-like vasculature. According to the anatomic observation, the ideal vertex is the entry point of the main blood vessels (i.e. vessel origin (VO)), which is usually beside the OD center. Moreover, VO does not coincide with the OD center. Hence, a VO-parabolic fovea localization scheme, with VO as the vertex of the parabola-like vasculature, is proposed to get better fovea localization. In order to exactly extract vessel origins from retinal images, four features based on the characteristic of vessel origin are selected. Once the vessel origin in a retinal image is identified, the SR-Hough transform employs the vessel origin as the vertex to obtain the fittest parabola. Finally, the VO-parabolic scheme uses the symmetry axis of the fittest parabola to locate the fovea. Two public available databases, DRIVE and STARE, are employed to evaluate the proposed scheme. The experimental results show that the average Euclidean distance between the located position of fovea and the one marked by experts are respectively 9.8 pixels and 30.7 pixels in these two databases.