International Journal of Computer Applications (0975 – 8887) Volume 111 – No 6, February 2015 5 Identifying Abnormalities in the Retinal Images using SVM Classifiers Shantala Giraddi BVB College Of Engg.and Technology Dept of CSE, Hubli, India Jagadeesh Pujari, Ph.D SDM College Of Engg.and Technology Dept of ISE, Dharwad, India Shivanand Seeri BVB College Of Engg.and Technology Dept of CSE, Hubli, India ABSTRACT Automated early detection of exudates in retinal images is a challenging task. With the global diabetic population increasing at an alarming rate, there is need for development of automated systems for detection of exudates. The main obstacle in exudates detection is extreme variability of color and contrast in retinal images that depends on the degree of pigmentation, size of the pupil and illumination. The aim of this paper is to develop and validate systems for detection of hard exudates and classify the input image as normal or diseased one. The authors have proposed and implemented novel method based on color and texture features. Performance analysis of SVM and KNN classifiers is presented. Images classified by these classifiers are validated by expert opthamalagists. . Keywords Diabetic Retinopathy, SVM classifier, KNN classifier Exudates. 1. INTRODUCTION Diabetes mellitus (DM) is a major cause of blindness all over the world. It has been found that patients with diabetic retinopathy (DR) are 25 times more likely to become blind than non-diabetics [1]. There were 31.7 million diabetics in India in year 2000 which is expected to reach 79.4 million by year 2030[2]. Both type 1 and type 2 diabetes can cause Diabetic Retinopathy. Both eyes can be affected by Diabetic retinopathy. Often there are no early signs of diabetic retinopathy. Symptoms may only become noticeable once the disease advances. Typical symptoms of retinopathy are [3].  Dark strings floating in your vision (floaters)  Blurred vision  Fluctuating vision  Dark or empty areas in your vision  Vision loss  Difficulty with color perception Diabetic retinopathy may be classified as early diabetic retinopathy and advanced diabetic retinopathy. Early diabetic retinopathy is called non proliferative diabetic retinopathy (NPDR). In this stage , there are no major symptoms. Retinal swelling may be present to some extent. In this stage tine capillaries become semi permeable membranes, later leaking or oozing fluid and blood into the retina. As the disease progress, the smaller vessels may close and the larger retinal vessels may begin to dilate and become irregular in diameter. Advanced diabetic retinopathy also called Proliferative diabetic retinopathy (PDR) is the most severe type of diabetic retinopathy. At this stage, damaged blood vessels begin to break, leak blood into the clear, jelly-like substance that fills the center of eye. They are not able to supply the nutrients to the retina. The starvation of nutrients in the retina causes growth of new blood vessels. This growth of new capillaries is called neo vascularization. Diabetic patients who have had diabetes for more than five years are likely to develop some form of Diabetic Retinopathy. Only regular screening can result in early detection and effective management of DR. Patients should get their both eyes screened at least once in a year. These screening programs generate large number of images and processing of which is time consuming process. Automated diabetic Retinopathy can save time and reduce the workload of opthmalagists. Developing strategies for screening large population for early detection of DR is engaging attention of several groups in India. Automated grading is less costly and of similar effectiveness, it is likely to be considered a cost‐ effective alternative to manual grading. In several patients, the only visible symptoms of DR are Exudates [4]. Hard exudates occurring in the macula can cause significant visual impairment. The main obstacle in exudates detection is extreme variability of color and contrast in retinal images that depends on the degree of pigmentation, size of the pupil and illumination. These factors affect the appearance of exudates in the retinal images. Many techniques such as clustering, morphological operations, pixel wise classification using various classifiers like Back propagation neural network (BPNN), Support vector machines (SVM) have been employed for the exudates detection. All these techniques have high computational requirement. In their previous work, authors have proposed and validated novel method for exudate detection based on color, texture features using Back propagation neural network [5]. In this paper comparative analysis of SVM and KNN classifiers is presented. The rest of the paper is structured as follows. In section 2, related works are discussed. Section 3 deals with the proposed methodology. In Section 4 experimental results are explained in detail. Paper is concluded in section 5 2. RELATED WORK Hussain F.Jaafar et.al [6] proposed a method based on top- down image segmentation and local thresholding by a combination of edge detection and region growing. Grading of hard exudates is performed. Nidhal K et.al [7] proposed a system in which Exudates are found by thresholding and false exudates are separated from true exudates using first order texture features. Ratio of red to green channel is used to identify optic disc region. Xiwei Zhang et.al [8] Segmented exudates based on mathematical morphology and characterized candidates based on classical features as well as contextual features and validated their system on e-ophtha EX database. In Another approach proposed by Juan Martin Cardenas [9] et.al image preprocessing is done with mean shift filtering and region growing algorithm is performed from local maxima regions taken as seeds to get final results. M. Usman Akram et.al [10] used filter banks to extract the