Retinal Fundus Features Hybrid Detection based on a Genetic Algorithm

In this paper, an analysis of images of the retinal fundus by means of a combination of soft-computing techniques is carried out in order to extract from blood vessels their bifurcation and crossover points. These retinal features are unique for each individual and therefore they are useful for a successive process of personal identification. In particular, the implemented method is based on four steps. In the first step, a pre-processing of the input image by using, in sequence, several filters is carried out: the retinal image is processed by means of pixel to pixel elaboration algorithms to obtain a black and white image representing the template of retinal veins and arteries. In the second step, applying to the so obtained image a genetic algorithm, the edges’ template is extracted. To this point, in the third step, the method continues with a skeleton process of the contours making use of an empirical algorithm. Finally, in the last step, the goal is caught up by characterising the searched points by means of a tracking algorithm.