Diabetic retinopathy, a hazardous complication of diabetes, is triggered by elevated blood sugar levels and can lead to blindness. In its initial phases, diabetic retinopathy usually doesn’t show symptoms, underscoring the significance of detection. Yet, ophthalmologists frequently utilize Fundus photography, which is unable to detect small variations essential for identifying tiny nerves and clots. Moreover, a fundus photo could display an irregular retinal surface or curvature. This article presents a fresh image processing framework that merges various filters like entropy, range, Gaussian, and Deriche filters, with image enhancement methods like CLAHE, LBP, SIFT, and HOG, to enhance the detail and quality of retinal images significantly. The proposed design, which employs a variety of unique filtering processes that overlap, could benefit in future diagnosis and treatment. This study makes use of the IDRiD dataset, which contains 516 images of various stages of diabetic retinopathy. These images are further pre-processed and supplied to increase the dataset. After examining numerous filters, this study applies a combination of four different filters: the Gaussian filter, the median filter, the CLAHE method, and finally, the canny edge detection technique.