
189Automated Cup-to-Disc Ratio Estimation for Glaucoma Diagnosis
However, the eye edges, that is, the border of the eye and the black background, could lead
to false centers outside the OD area (see Figure 11.2c). To avoid this error we apply an aniso-
tropic diffusion lter to the G channel of the image [65], as in Figure 11.2d. Next, we threshold
this image with a threshold value xed to the maximum of the blurred image ponderated to
90%. An example of this thresholding is shown in Figure 11.2e. A binary mask is obtained
(Figure 11.2f), wherein the white pixels correspond with candidate OD center pixels.
Finally, the center of the OD is considered ...