Segmentation of blood vessels in retinal images is an important part in retinal image analysis for diagnosis and treatment of eye diseases for large screening systems. In this paper, we addressed the problem of background and noise extraction from retinal images. Blood vessels usually have central light reflex and poor local contrast, hence the results yield by blood vessel segmentation algorithms are not satisfactory. We used different preprocessing steps which includes central light reflex removal, background homogenization and vessel enhancement to make retinal image noise-free for post-processing. We used mean and Gaussian filtering along with Top-Hat transformation for noise extraction. The preprocessing steps were applied on 40 retinal images of DRIVE database available publically. Results show the darker retinal structures like blood vessels, fovea, and possible presence of microaneurysms or hemorrhages, get enhanced as compared to original retinal image and the brighter structures like optic disc and possible presence of exudates were get removed . The presented technique will definitely improve automatic fundus images analysis also be very useful to eye specialists in their visual examination of retina.

Blood Vessel Segmentation Noise Extraction DRIVE Database Top-Hat Transformation.