Detection and Classification of Exudates and Non-exudates in Retinal Images

Abstract

The retina of human eye plays a key function in the vision, and it is a light-sensitive layer. The optics of eye produces an image figure in the retina. The various eye diseases like diabetic retinopathy, myopia, macular pucker, and macular hole have an effect on the retina. The retina is affected by these diseases which are vascular disease and cause vision mutilation and blindness. These diseases happen due to diabetics, aging, and nearsightedness. Exudates are the pathological condition of the retina. So the early detection of these is very important. In the paper, an efficient methodology like Otsu thresholding method and the K-means clustering method is proposed for the recognition of exudates. After detecting the exudates, various texture feature extraction processes are involved. Finally, the classification method is performed using Backpropagation Neural Networks (BPN). The main spotlight of the projected work is to develop algorithms for exudates recognition and categorization of retinal images in pathological or non-pathological, convalescing investigation of the fundus images. The experimental results acquired from the projected method of extracting the features and classification method exposed that non-diseased cases were recognized with 90% exactness while temperate and severe cases were 99% accurate.