A Generalization of Majority Voting Scheme for Medical Image Detectors

In this paper we propose a method for locating the optic disc (OD) in retinal images automatically using a generalization of majority voting scheme. Applying more different optic disc detectors for voting we can achieve better performance for the automatic detection system than for each individual algorithm. The location with maximum number of OD center candidates falling within a radius predefined clinically can be used to localize the OD center. In contrast to the classical voting system we can make good decision if the number of algorithms detecting the optic disc correctly is less than the half of the overall number of algorithms.