Optic disc segmentation using circular hough transform and curve fitting

We present a technique to segment the optic disc (OD) boundary from a color retinal fundus image. The technique used involves the extraction and removal of blood vessels using a top hat transform and an inpainting process. Then, a circular Hough transform is applied to the detected edges to obtain a coarse boundary of the OD and following which probable points of the optic disc are fed to a curve fitting algorithm which uses a higher order polynomial to draw the final boundary of the optic disc. The optic disc segmentation is a crucial part in estimating the CDR (Cup-to-disc ratio) which can be used as an early indicator of glaucoma and for following the progression/remission of the disease.