Segmentation of Blood Vessels by Local Analysis of 2D Image Patches

Retinal vasculature can be considered a major indicator of various opthalmologic disorders and diseases. Therefore, it is important to accurately detect and segment the retinal blood vessels. We present a novel technique for the segmentation of arteries and veins from a retinal fundus image. Local 2D image patch-based analysis using intensity gradients is used as the basis for detection of vessels. Two stages of preprocessing are implemented before detection for ease of approach. The proposed model carries several advantages over the state-of-the-art. The computational cost of the proposed model is low. Also, the average execution time per image is 1.3 seconds which is competitive with the state-of-the-art. The performance of the proposed model is compared with the standard approaches on two publicly available fundus image databases - DRIVE and STARE, using three popular metrics such as sensitivity, specificity, and accuracy.