How to perform template matching between matrix and image?

Question

I've been working on a simple machine learning on Matlab
My dataset looks like:

There are 10 pictures in a folder, 5 for training, and 5 for testing.
Totally 40 folders, or 400 pictures in my dataset.

First, I read 200 images(32*32pixel) into a matrix called FFace which size is 200x1024.
Then, perform PCA on FFace and get pcaTotalFace(200x50).
Next, perform LDA and get prototypeFace(200x50).
I've successfully reduce FFace(200x1024) to prototypeFace(200x50).

My question is : How to do template matching between remaining 200 images and prototypeFace?

Below are my PCA and LDA code for reference. PCA:

function [FFace, TotalMeanFace, pcaTotalFace, projectPCA, eigvector, 
prototypeFace]=PCALDA_Train

people=40; 
withinsamp=5; 
principlenum=50; %reduction to dimension of 50
FFace=[]; %store every traning picture(200*1024)

for k=1:1:people
    for m=1:2:10
        matchstring=['dataset' '\' num2str(k) '\' num2str(m) '.bmp'];
        matchX=imread(matchstring);
        matchX=double(matchX);
        if(k==1 && m==1)
            [row, col]=size(matchX);
        end
        matchtmpF=[];
        % arrange the image into a vector
        for n=1:row
            matchtmpF=[matchtmpF, matchX(n,:)]; %1*32 row concat 32 times
        end
        FFace=[FFace;matchtmpF]; % col concat 
    end
end
    TotalMeanFace=mean(FFace);
    FFaceNor=FFace-TotalMeanFace;
    covPCA=FFaceNor'*FFaceNor; 
    [Vec, Val]=eig(covPCA);
    eigval=diag(Val); 
    [junk, index]=sort(eigval, 'descend');
    PCA=Vec(:,index); 
    eigval=eigval(index);
    projectPCA=PCA(:,1:principlenum); %extract the principle component
    pcaTotalFace=[];
    for i=1:1:withinsamp*people
        tmpFace=FFaceNor(i,:);
        tmpFace=tmpFace*projectPCA;  
        pcaTotalFace=[pcaTotalFace; tmpFace];  
    end

LDA:

classMean=[];
SW=[];
for i=1:withinsamp:withinsamp*people
    withinFace=pcaTotalFace(i:i+withinsamp-1,:);
    if(i==1)
        meanwithinFace=mean(withinFace);
        withinFace=withinFace-meanwithinFace;
        SW=withinFace'*withinFace %cov(withinFace)
        classMean=mean(withinFace);
    end
    if(i>1)
        meanwithinFace=mean(withinFace);
        withinFace=withinFace-meanwithinFace;
        SW=SW+withinFace'*withinFace;
        classMean=[classMean;mean(withinFace)];
    end

end
pcaTotalMean=mean(pcaTotalFace);
classMean=classMean-pcaTotalMean;
SB=classMean'*classMean;
[eigvector, eigvalue]=eig(inv(SW)*SB);
eigvalue=diag(eigvalue);
[junk, index]=sort(eigvalue, 'descend');
eigvalue=eigvalue(index);
eigvector=eigvector(:,index);
prototypeFace=pcaTotalFace*eigvector;

end

Answer 1

I figured it out on my own.

function [correct]=PCALDA_Test(projectPCA, LDA, prototypeFace, 

TotalMeanFace)

principleNum=50;
ldaNum=30;
people=40;
withinsamp=5;
totalcount=0;

correct=0;
projectLDA=LDA(:, 1:ldaNum); %LDA 
FFaceTest=[];
for i=1:1:people
    for j=2:2:10
        s=['ORL3232' '\' num2str(i) '\' num2str(j) '.bmp'];
        totalcount=totalcount+1;
        tempF_Test=[];
        matchX=imread(s);
        matchX=double(matchX);
        if(i==1 && j==2)
            [row, col]=size(matchX);
            
        end
        for n=1:row
            tempF_Test=[tempF_Test, matchX(n,:)];
            
        end
        FFaceTest=[FFaceTest; tempF_Test];    
    end
end

meanTotalTest=mean(FFaceTest);
TestNor=FFaceTest-meanTotalTest;
pcaTotalFace_Test=[];
for i=1:1:withinsamp*people
    tmpFace=TestNor(i,:);
    tmpFace=tmpFace*projectPCA;
    pcaTotalFace_Test=[pcaTotalFace_Test;tmpFace];
end
prototypeFace_Test=pcaTotalFace_Test*projectLDA; %map PCA testSet to LDA

[n,m]=size(prototypeFace);
C=floor(prototypeFace);
for id=1:1:withinsamp*people
    P=prototypeFace_Test(id,:);
    Distance=zeros([1,n]);
    Distance=sqrt(P.^2*ones(size(C'))+ones(size(P))*(C').^2-2*P*C');
    [minValue,minrow]=min(Distance);
    if abs(id-minrow) <=4   
        correct=correct+1;
    end
end
        
Recognition_rate=double(correct*100/totalcount);
fprintf('Number of correction : %d\nNumber of incorrection : %d\nRecognition Rate : %.4f', correct,  200-correct, Recognition_rate);

disp(' %');