Bessel Filter Based Blood Vessel Segmentation for Retinal Images

Abstract

In the era of digital technology, the blood vessels extraction from the retina becomes a vital source of information for the detection of various diseases related to retinal pathology. The aim of this work is to provide a novel technique to assess patients for diabetes based on four parameters: Sensitivity, Accuracy, Specificity and Area under Curve. These parameters are obtained by performing extraction of blood vessels from a retinal fundus image.  Firstly, the outer ring from the input fundus image is removed by using erode morphological operator. Following this process of erosion on input image, this paper applies the concept of Bessel filter for the purpose of enhancement of blood vessels. As a next step, this enhanced blood vessels are extracted by obtaining principal curvature of the image using Hessian matrix and Eigenvalues. Finally, CLAHE operation is applied on the segmented image to improve the contrast of the blood vessels. The outcome of the proposed method is then validated with the pre-segmented retinal image to determine the diabetic retinopathy based on the above mentioned parameters.