Vessel centerlines extraction from Fundus Fluorescein Angiogram based on Hessian analysis of directional curvelet subbands

This paper presents a novel algorithm for automatic extraction of the blood vessels centerline in Fundus Fluorescein Angiography (FFA) images in different diabetic retinopathy (DR) stages. First, the background normalized images are enhanced by applying a morphological edge detector. Then each of the directional images resulting from curvelet sub-bands is individually processed using Hessian matrix and first order derivative of the directional images information in a multi-scale framework for extracting initial centerline segments. Every resulted candidate segment in previous step is confirmed or rejected based on the length and intensity features and eigenvalues analysis. The final vessels centerline segmentation is obtained by connecting the images subsets in a binary image. The proposed algorithm is tested on 70 FFA images from different DR stages and the performance of method in terms of true positive ratio (TPR) and false positive ratio (FPR) that are obtained .9017 and .0983 respectively.