Abstract: Glaucoma refers to the optic nerve damages, causing functional abnormalities in the visual field leading to irreversible loss of vision. The diagnosis of glaucoma at an early stage becomes vital to prevent the damage to optic nerve. In a fundus image, glaucoma is well characterised by recognizable patterns of optic disc and retinal nerve fibre structure. In this study, we present a methodology for diagnosis of glaucoma by using the Local Binary Patterns (LBP) and fractal features from fundus image. The input fundus image is pre-processed using Contrast Limited Adaptive Histogram Equalisation method (CLAHE). The LBP of the enhanced image is obtained and fractal features are extracted from the LBP. Significant fractal feature are selected based on the t-test statistics. Finally, Kernel based Support Vector Machine classifier (K-SVM) is employed on the selected features for classifying the healthy and glaucomatous image. The presented methodology is implemented with the images obtained from FAU and RIMONE database. Simulation results, demonstrate that the fractal features obtained from LBP with K-SVM Classifier identifies glaucomatous eye with higher accuracy when compared with clinical results.
K. Nirmala, N. Venkateswaran and C. Vinoth Kumar, 2016. Kernel SVM Classifier for Detection of Glaucoma Using LBP Based Fractal Features. Asian Journal of Information Technology, 15: 2702-2708.