Weird Calculation difference using uint16 and double for an image

Hello, I have a function that calculates the "contrast" of a greyscale image

function FM=CalculateBrenner(Image)

        [M N] = size(Image);
        DH = Image;
        DV = Image;
        DH(1:M-2,:) = diff(Image,2,1);
        DV(:,1:N-2) = diff(Image,2,2);  % second order difference between
                                        % columns, i.e., along x-direction. 

        FM = max(DH, DV);     
        FM = FM.^2;
        FM = mean2(FM);

end

when my image is a uint16, I get

FM=  100.39
max=438
min=22

But when my image is cast as a double

I get:

FM = 204.14
max=438.00
min= 22.00

So why is the calculation FM different for both cases, yet the values its using are the same (indicated by the max & min)

Answer 1

[Steven Lord]

d = [1 2 1];
u = uint16(d);
dd = diff(d)
du = diff(u)

The smallest value you can store in an unsigned 16-bit integer is 0. In double 1 - 2 is -1 but in uint16 it is 0.

[Jason]

OK, so I should use a double as it doesn't ignore negative values.

[Steven Lord]

You could use imabsdiff if you have Image Processing Toolbox, or just take the max of A-B and B-A (one will be zero, one may not be) for appopriate pieces A and B of your Image variable.

d = [1 2 1];
u = uint16(d);
max(u([3 2])-u([2 1]), u([2 3])-u([1 2]))

I'll leave that last line for you to generalize.