Like I said in your duplicate question http://www.mathworks.com/matlabcentral/answers/130082#comment_234403
"Keep working on it. Did you see the link to papers from people have have presumably solved it? It looks like color alone won't solve it if you need to find sclera that is reddish so you may have to combine with other methods like spatial location or edge detection. Maybe you have to locate the iris first and look outwards from there. I can see that it could be a very challenging problem to automatically get every possible pathological image absolutely correct. As a last resort you may have to resort to using imfreehand(). Let me know if you want a demo of that. The good thing is that if you resort to doing it manually, there aren't many images since each patient will have only a few images at most, not hundreds or thousands."
