Color image sharpening and application to eye fundus image analysis

This work aims to build an algorithm to sharpen a color digital image based on S-CIELAB extension. We use this method to segment the optic cup inside the optic disc, and the optic disc from the rest of the eye fundus image of glaucomatous eyes. S-CIELAB involves a series of smoothing spatial filters in the opponent color space to approximate the contrast sensitivity functions of the human vision system. The filters are linear combinations of Gaussian masks. We combine these spatial filters with the Laplacian operator in each opponent channel to obtain the sharpened image. The resulting image is then subtracted from the original image in each opponent channel and back transformed to the device independent representation space (XYZ) to obtain the final sharpened image. The application developed to segment the optic cup and the optic disc is intended to give assistance in the cup to disc ratio estimation of glaucomatous eyes. Often the contours of both the optic cup and disc are faint and intersected by entangled veins that make it difficult to draw their silhouettes. The method is based on the information of color, the color differences between neighbor pixels and the geometry of the areas involved. It includes the spatial filtering proposed in the S-CIELAB extension and uses the color image sharpening algorithm.