Fundus Image Classification Using Two Dimensional Linear Discriminant Analysis and Support Vector Machine

( Vol-2,Issue-10,October 2016 )

Author(s): Wahyudi Setiawan, Fitri Damayanti

Total View : 1104
Downloads : 164
Page No: 1766-1771
DOI:

Keywords:

Diabetic Retinopathy, Two Dimensional Linear Discriminant Analysis, Support Vector Machine, Fundus Image, MESSIDOR.

Abstract:

It is constructed in this study a classification system of diabetic retinopathy fundus image. The system consists of two phases: training and testing. Each stage consists of preprocessing, segmentation, feature extraction and classification. The tested image comes from the MESSIDOR dataset which has a total of 100 images. The number of classes to be classified consists of four classes with each class consists of 25 images. The classes are normal, mild, moderate and severe of Diabetic retinopathy. In this study, the level of preprocessing uses grayscales green channel, Wavelet Haar, Gaussian filter and Contrast Limited Adaptive Histogram Equalization. The level of segmentation uses masking as a process of doing the subtracting operation of between the original image and the masking image. The purpose of the masking is to split between the object and the background. The feature extraction uses Two Dimensional Linear Discriminant Analysis (2DLDA). The classification uses Support Vector Machine (SVM). The test results of some scenarios show that the highest percentage of accuration of the system is up to 90%.

Cite This Article:

Share: