Robust Detection of Exudates Using Fundus Images

Human eye is a combination of different components to provide a 3 dimensional color image. Macula is considered as the most sensitive part of the eye which enables the crispest vision, needed for reading, driving, watching screens and other sensitive tasks. Exudates are formed due to leakage of fluids or blood and can be formed in the eye as well. The proposed system is using different mixture of features and by using SVM for segmentation of exudates around the macular region of the eye. If the exudates are present near the macula, they can be prove to be harmful for the vision of the person. We have tested our system on four databases namingly DRIVE[1], DIARETDBl[2], MESSIDOR[3] and a locally generated database from Armed Forces Institue of Ophthalmology (AFIO) Rawalpindi, Pakistan. We have achived 100
^%
accuracy using DRIVE database, 95% accuracy for DIARETDB1 & MESSIDOR respectively. While 94% for AFIO database. The average accuracy acheived using 1791 images in total was 95.48%.