Region Growing Based Segmentation Using Forstner Corner Detection Theory for Accurate Microaneurysms Detection in Retinal Fundus Images

Abstract

Several diseases have adverb effect on visual system of human visual system (eye) and diabetes is leading one among them. Prolong and uncontrolled diabetic patient is prone to Diabetic Retinopathy (DR). DR is asymptomatic disease hence requires detection in early stages to avoid big loss in vision. It shows immediate necessity of screening system to access eye images and analyze for DR detection. Microaneurysms (MAs) are primary lesion of DR, so their detection can give time for patient and ophthalmologist to prevent further vision loss. Region growing segmentation method is proposed for accurate detection of MAs. The preprocessing of retinal images uses non local means (NLM) filter and contrast limited adaptive histogram equalization (CLAHE) for noise removal and enhancement image quality. In segmentation, region growing algorithm in which the seeds for the grower are selected and positioned by means of Forstner Corner Detection theory is utilized. After segmentation, the redundant areas are removed using morphological operations (Niblack Adaptive Thresholding) and finally the Predator prey optimizer is used for optimizing the features for MA detection.