How could I compare two images?

Question

0 votes

FacialRecognizer.zip

Hi, I'm new using matlab and I was sent an assignment that consists on creating a face recognizer. I have almost finished the code but I'm having problems at the end. The steps that I'm doing are

1st. Load images and convert them to gray scale

2nd. Create an array of each image and from those arrays, create a single matrix of all the images

3rd. Do the SVD (I know that I can do with the eigen faces but they told us to use this technique instead)

4th. Load an image that will be compared to the matrix of images

5th. Calculate a similar image

The problem that I have is that it isn't showing the similar image that was found on the matrix. I'm doing the norm between the components of A (the matrix of images) in base of U transposed and the test image. And then for looking for the minimum value between the A and the result of the norm I'm using the function called min. And after that I reshape the result and show it on screen. Here is the code, I think that everything is ok till the part that I calculate a similar image.

    clear all, close all;clc;
    N = 200; %Size of the images
    M = 12; %Number of images
    %1. Load the images
    %George Clooney
    C1 = imread('C1.jpg','jpg');
    C1 = rgb2gray(C1);
    C1 = double(C1);
    C1 = reshape(C1,40000,1);
    C2 = imread('C2.jpg','jpg');
    C2 = rgb2gray(C2);
    C2 = double(C2);
    C2 = reshape(C2,40000,1);
    C3 = imread('C3.jpg','jpg');
    C3 = rgb2gray(C3);
    C3 = double(C3);
    C3 = reshape(C3,40000,1);
    %Sandra Bullock
    B1 = imread('B1.jpg','jpg');
    B1 = rgb2gray(B1);
    B1 = double(B1);
    B1 = reshape(B1,40000,1);
    B2 = imread('B2.jpg','jpg');
    B2 = rgb2gray(B2);
    B2 = double(B2);
    B2 = reshape(B2,40000,1);
    B3 = imread('B3.jpg','jpg');
    B3 = rgb2gray(B3);
    B3 = double(B3);
    B3 = reshape(B3,40000,1);
    %Jared Leto
    L1 = imread('L1.jpg','jpg');
    L1 = rgb2gray(L1);
    L1 = double(L1);
    L1 = reshape(L1,40000,1);
    L2 = imread('L2.jpg','jpg');
    L2 = rgb2gray(L2);
    L2 = double(L2);
    L2 = reshape(L2,40000,1);
    L3 = imread('L3.jpg','jpg');
    L3 = rgb2gray(L3);
    L3 = double(L3);
    L3 = reshape(L3,40000,1);
    %Will Smith
    S1 = imread('S1.jpg','jpg');
    S1 = rgb2gray(S1);
    S1 = double(S1);
    S1 = reshape(S1,40000,1);
    S2 = imread('S2.jpg','jpg');
    S2 = rgb2gray(S2);
    S2 = double(S2);
    S2 = reshape(S2,40000,1);
    S3 = imread('S3.jpg','jpg');
    S3 = rgb2gray(S3);
    S3 = double(S3);
    S3 = reshape(S3,40000,1);
    %2. join all the images
    std.data{1} = C1;
    std.data{2} = C2;
    std.data{3} = C3;
    std.data{4} = B1;
    std.data{5} = B2;
    std.data{6} = B3;
    std.data{7} = L1;
    std.data{8} = L2;
    std.data{9} = L3;
    std.data{10} = S1;
    std.data{11} = S2;
    std.data{12} = S3;
    A = zeros(N*N,M);
    %convert the images to a single matrix
    for k=1:M
       A(:,k) = std.data{k}(:); 
    end
    save('matrixOfImages.mat', 'A');
    %3. do the SVD for compressing the data
    [U,S,V] = svd(A);
    p = 5;
    Ut = U(:,1:p);
    St = S(1:p,1:p);
    Vt = V(:,1:p);
    W = St*Vt';
    save('componentsOfA.mat','W');
    %4. test an image to compare
    test = imread('C1.jpg','jpg');
    test = rgb2gray(test);
    imshow(test);title('Image to compare:')
    test = double(test);
    test = reshape(test,40000,1);
    test = Ut'*test;
    %calculate a similar image
    for m=1:12
       Y(m) = norm(W(:,m)-test);
    end
    u = min(A,Y);
    t = reshape(u,200,200);
    imshow(t);title('The similar image is:');