Retinal vessel segmentation using a multi-scale medialness function

Recently, automated segmentation of retinal vessels in optic fundus images has been an important focus of much research. In this paper, we propose a multi-scale method to segment retinal vessels based on a weighted two-dimensional (2D) medialness function. The results of the medialness function are first multiplied by the eigenvalues of the Hessian matrix. Next, centerlines of vessels are extracted using noise reduction and reconnection procedures. Finally, vessel radii are estimated and retinal vessels are segmented. The proposed method is evaluated and compared with several recent methods using images from the DRIVE and STARE databases.