Automated Denoising of Diabetic Retinopathy Images for Enhanced Medical Diagnosis

Globally, diabetic retinopathy (DR) continues to be a major cause of vision impairment. Early diagnosis and detection of DR are essential to preventing blindness and vision loss. But the traditional manual method of screening retinal pictures takes a lot of time and is prone to human mistake. Using image processing to apply automated denoising techniques can increase image quality, make DR lesions more visible, and help with correct diagnosis. In order to improve the quality of images of diabetic retinopathy, this study investigates the ability of four image denoising filters: guided image filtering, Gaussian-bilateral filtering, Gabor filtering, and Haar Wavelet filtering. AGAR300 (IEEE DataPort), Diabetic Retinopathy Detection, Stare, Indian Diabetic Retinopathy Image Dataset, and APTOS 2019 Blindness Detection are the five datasets used to evaluate the efficacy of these filters. The findings demonstrate that noise in DR images may be efficiently reduced by all four filters. The one that shows the biggest overall improvement in lesion visibility and image quality among them is guided image filtering.