Morphological and Neural Network Based Approach for Detection of Exudates in Fundus Images

Abstract

This paper presents a technique for detection of hard exudates in fundus images. Hard exudates is one of the anomaly formed in retina due to diabetic retinopathy. Early detection of hard exudates may prevent the vision loss of diabetic retinopathy patient. In this work, optic disc (OD) is extracted with the help of morphological operators. OD is masked in green component image to avoid the misclassification between OD region and hard exudates region. Then features of green component image are computed and applied to neural network for detection of hard exudates. Experimental results show the better competency of algorithm on DIARETDB0 and DIARETDB1 databases.