IRJET - Detection of Diabetic Retinopathy using Local Binary Pattern

This document presents a method for detecting diabetic retinopathy using local binary patterns. It involves three main stages: pre-processing of fundus images which includes disk segmentation and vessel segmentation, feature extraction using LBP and calculating texture features like variance, mean, skew and standard deviation, and classification of images into DR and non-DR categories using a support vector machine. The method aims to analyze texture of retina background and classify images without requiring segmentation of retinal lesions. It achieved 98% accuracy in detecting diabetic retinopathy using this approach on a dataset of fundus images.