In order to identify blood vessels in retinal fundus images, we developed a contour detection-based image processing system utilizing Mamdani (Type-2) fuzzy rules. The method uses a median filter to remove background noise and Contrast-Limited Adaptive Histogram Equalisation (CLAHE) to boost contrast in green channel data from eye fundus images. Mamdani (Type-2) fuzzy rules are applied to the image gradient value generated from non-local and non-singular fractional order kernels in order to find edges. Results from experiments on the drive dataset, HRF dataset, and real-world photos from a local hospital show how beneficial the suggested approach is as a flexible methodology that can be used for a variety of edge detection/contour-based applications. Our findings indicate that the suggested method outperforms most listed techniques and achieves this with significant computational efficiency. The output of the suggested method may be used as input in deep learning techniques, which will be further applied in the clinical application of diabetic retinopathy and glaucoma to discover abnormalities likely related to the progression and different stages.