Retinal Blood Vessels Segmentation using ISODATA and High Boost Filter

Automatic segmentation of retinal vessels tree is a prerequisite step in the field of ophthalmology for computer-aided diagnosis. This paper describes a robust vessel segmentation algorithm that is efficient enough to segment retinal vessels tree automatically and accurately. The proposed method includes several steps. Initially, we splitted the color fundus image into three channels Red, Green and Blue channels. On Green channel, CLAHE is used for contrast enhancement and top-hat morphological operation are applied to remove noise. Meanwhile High Boost and Frangi filters are introduced to get the prominent vessels pixels. Finally, the ISODATA thresholding technique is applied to get final binarized output segmented image. The proposed method outperforms recently published methods in terms of statistical parameters, such as Accuracy, Sensitivity, and Specificity on DRIVE and STARE datasets. In addition, the proposed method showed a remarkable scientific value for future computer-aided diagnosis applications.