matlab principal component analysis ( PCA)

image processing matlab

principal component analysis

Characteristic extraction aims to obtain vital information from facial features . The information is taken in the form of data representative of a raw data is represented using principal component analysis method . PCA is used to mengektraksi characteristics of the normalized image . PCA process will produce eigenvectors or characteristic vectors are then used to form the eigenspace ( eigenfaces ) . The steps in the formation of the eigenspace vectors forming faces , and quest weight of the vector eigen faces.

• Formation vector face

Face image which is a two- dimensional image can be viewed as a one dimensional vector . If the length and width of the image is M and N pixels , the number of vector components is MXN . Each pixel is encoded by a vector component . Formation vector of an image is done by placing each line of the image in addition to the other line called lexicographical ordering .

 

• Face Vector Image Formation of Faces

Results of the vector is then used as a column vector . Ie by multiplying the number of face vectors are then obtained vector column face . If the face amount is the vector W is the face ( MXN ) x W.

  listing program.

tic

load hasil\dbdata data

% Principal Component Analysis

[vec col]=size(data);

%mencari nilai tengah

mn=mean(data);

for i=1:vec

    xmean(i,:)=data(i,:)-mn;

end

%menghitung matrix covarians

kovar=xmean*xmean';

%mendapatkan eigen vektor dan eigen value

[v d]=eig(kovar);

%membentuk diagonal matrix sebagai vektor

d1=diag(d);

[d2 index]=sort(d1);

cols=size(v,2);for i=1:cols

    vsort(:,i)=v(:,index(cols-i+1));

    dsort(i)=d1(index(cols-i+1));

end

dsum=sum(dsort);

dsum_extract=0;

p=50;

i=1;

%mengambil 50 vektor eigenfaces

while(i<=p && dsort(i)>0)

    base(:,i)=dsort(i)^(-1/2)*xmean'*vsort(:,i);

    i=i+1;

end

PCA=data*base;

accu=0;

save hasil\dbeig PCA base

tag :

• pcaprincipal componenet analysi
• pca
• image procesing
• matlab code
• face recognition
• image recognition
• feature ektraction