Pixel Feature Classification Based Blood Vessel Segmentation in Retinal Image

Automated detection of blood vessel is always been the area of interest for various eye related disease diagnosis. Diabeic Retinopathy is the most common cause of blindness. This paper presents a new optimized method for blood vessel detection in digital retinal images. This method uses a neural network (NN) scheme for pixel classification and computes a 2-D vector composed of gray-level features for pixel representation. Based on the feature vector pixels are classified as vessel and nonvessel. This algorithm is tested on publically available Drive database where vasculature structures are marked by expert. The average accuracy of 0.9361 is achieved which is far superior as compared to the accuracy obtained using rule based method