Preprocessing digital retinal images for vessel segmentation

Abstract

The information contained in the retinal vasculature is used to diagnose the onset of retinal diseases such as diabetic retinopathy. However, due to non-uniform illumination and variations in imaging modalities, the contrast between the retinal blood vessels network and the background is very low, encumbering the analysis and the diagnosis processes. This prompts the need for preprocessing digital fundus images to remove noise and improve contrast thus increasing the segmentation accuracy of the retinal vasculature. In this study, we address issues of nonuniform illumination and low contrast by developing a framework that implements shade correction, image enhancement and prepares the digital fundus images for the next stage.