Why am I getting these incorrect values for the dct2 function?

Hello,

Basically, I'm taking the dct2 function of certain fingerprint images and then comparing them. Naturally, the image of the same person's fingers should have the lowest difference. THis is my code:

For inputting my main comparison image:

A=imread('r1.jpg');
A=imresize(A,[128,128]);
A=dct2(A);

For the image of another to be compared::

B=zeros(128,128,5);
C=imread('m1.jpg');
C=imresize(C,[128,128]);
C=dct2(C);
B(:,:,1)=C;

There are five such images. My comparison loop is:

val=999;
for i=1:5
    mean2((A)-(B(:,:,i))).^2
    if mean2((A)-(B(:,:,i))).^2<val
        val=(mean2((A)-(B(:,:,i))).^2)
        m=i
    end
end
disp(m)

Now, all five pictures are from different people other than A except the fifth one.

1: 0.0186 2: 0.0037 3. 9.3263e-004 4. 0.0103 5. 0.0017

Now, the answer should be the fifth one. But due to the 3rd erratic value, it comes as 3. Why do I get such a value? What is the remedy for it? Is it the image...or the code? However, I get values such as this after comparing:

[Oleg Komarov]

Please format the code: http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup#answer_18099

Answer 1

[Oleg Komarov]

You have to you have to square the differences not the mean:

mean2((A - B(:,:,i)).^2)

or the means of symmetric distributions (of differences) with same first central moment will be equal.

Or use imabsdiff

EDIT

vals = zeros(5,1);
for ii = 1:5
    vals(ii) = mean2((A - B(:,:,ii)).^2);
end
[m,idx] = min(vals);

Calculate the differences for all images and just pinpoint the minimum.

[Aditya]

Chief...You have no idea how thankful I am! Any other ideas on how I can improve this?

[Oleg Komarov]

If you think my answer solved your problem please accept it.