Figure 16

Segmentation processes of a non-pathological retinal fundus image obtained from our dataset. Mask image (a), image edge obtained using the proposed phase map (b), image structure obtained using the phase map (c), binary image obtained from (b) using Otsu thresholding[43](d), skeletonized version of (d) after eliminating outliers (e), dilated version of (e) (f), segmented result using only green channel of the given image based on the proposed structure-based level set segmentation method without preprocessing (g), segmented image using the proposed method (h), and the level set function after 61 iterations (i).