Pre-processing of retinal images acquired from digital fundus cameras for improved performance of diagnostic tools in biomedical engineering

Abstract Early diagnosis of abnormality related to retina such as glaucoma is of utmost significance as it is the common cause of permanent blindness across the world and is predicted to rise further in near future. It is usually diagnosed using fundus images acquired from digital fundus cameras. But the images acquired are prone to some noises that create alternations in outputs of diagnosis by tempering the accuracy. These noises include vessels in the retina, lower contrast of images and irregular illumination that depreciates the performance of cup and disc segmentation which are crucial to diagnose glaucoma. Thus, removal of outliers to pre-process retinal images plays an important role in diagnostics. This manuscript presents a hybrid technique for removal of outliers from retinal images. The suggested approach can be embedded in digital cameras to filter out the outliers and improve the performance of diagnostic tools.