Calculate height of object

Please find attached my image. I need to find the meniscus height. algorithm so far: 1) k-means clustering 2) 8-neighborhood this gives me a binary image. How do i find the height (NOT length) of the meniscus. Also i have many such images, is there some automated way to find height of all of them with same code. Right now i manually find the height by comparing it to a line of 5-7 pixels.

http://www.4shared.com/download/Bghnq8n9ba/me_online.jpg?lgfp=3000

http://www.4shared.com/download/c4HY8whOce/me1.jpg?lgfp=3000

[Image Analyst]

I see an eye. What are you calling the meniscus? And what are the axes for your clusters in (1)? And I don't understand step (2) at all. "8-neighborhood" is not an algorithm for binarizing an image. It's a concept for how different regions might be connected or disconnected from each other.

[Sonalika]

the thin bright white line is the tear meniscus(lower lid) I'm using two clusters using the inbuilt function im giving a ginput manually selecting a point of the meniscus so that i get all the 8-connected points with it which finally gives the binary image with all the similarly illuminated points i.e the meniscus. if you have any other method of segmenting the meniscus and calculating its height, i would really appreciate that

[Sonalika]

 for finding the neighbors:

 if true
    % code
  end
function Phi = segCroissRegion(tolerance,Igray,x,y)
if(x == 0 || y == 0)
    imshow(Igray,[0 255]);
    [x,y] = ginput(1);
end
Phi = false(size(Igray,1),size(Igray,2));
ref = true(size(Igray,1),size(Igray,2));
PhiOld = Phi;
Phi(uint8(x),uint8(y)) = 1;
while(sum(Phi(:)) ~= sum(PhiOld(:)))
    PhiOld = Phi;
    segm_val = Igray(Phi);
    meanSeg = mean(segm_val);
    posVoisinsPhi = imdilate(Phi,strel('disk',1,0)) - Phi;
    voisins = find(posVoisinsPhi);
    valeursVoisins = Igray(voisins);
    Phi(voisins(valeursVoisins > meanSeg - tolerance & valeursVoisins < meanSeg + tolerance)) = 1;
end

if true
  % code
end

Answer 1

[Image Analyst]

I suggest you look into anisotropic diffusion. It's made for smoothing along lines and sharpening edges and is used in things like fingerprint analysis. See attached demo.

[Sonalika]

Will definitely look into it. Thanks so much. But the main question that I'm still struggling with is getting the height in pixels. As its really minute I can't select the two points between which I need the height manually. Is there a way around this?

[Image Analyst]

Take the binary image and process it to get rid of the stuff other than the meniscus sliver. Then use regionprops to get it's area. Then skeletonize and get the length. Divide the area by the length to get the average width.

[Sonalika]

Thank you so much sir. I also tried anisotrophic diffusion and it works beautifully. So thank you once again!