Suppose you store the images as column vectors of length NxN (the
number of pixels in each image) and that you have M images, you'll end
up with a matrix G (N^2 x M) :
1
CovMat = -------- G G^T
M - 1
this is what you are trying to find the eigenvalues/eigenvectors for,
right?
There is a trick widely used when doing PCA on sets of images, it is
based on the fact that the (M x M) matrix:
1
lmCovMat = -------- G^T G
M - 1
has the same non-zero eigenvalues of CovMat. So what you do is to
compute the eigenvalues and eigenvectors of this low memory covariance
matrix and then, for the eigenvectors, compute:
E = G lmE
where E is a matrix containing the wanted eigenvectors and lmE is the
matrix of the eigenvectors compouted from the lmCovMat.
This is implemented already in the OpenCV libraries.
Hope it helps.
number of pixels in each image) and that you have M images, you'll end
up with a matrix G (N^2 x M) :
1
CovMat = -------- G G^T
M - 1
this is what you are trying to find the eigenvalues/eigenvectors for,
right?
There is a trick widely used when doing PCA on sets of images, it is
based on the fact that the (M x M) matrix:
1
lmCovMat = -------- G^T G
M - 1
has the same non-zero eigenvalues of CovMat. So what you do is to
compute the eigenvalues and eigenvectors of this low memory covariance
matrix and then, for the eigenvectors, compute:
E = G lmE
where E is a matrix containing the wanted eigenvectors and lmE is the
matrix of the eigenvectors compouted from the lmCovMat.
This is implemented already in the OpenCV libraries.
Hope it helps.
clear
[x]=load_imgs('training');
x=x(1:100,:);
a1=x*x';
b=x'*x;
[va,p]=eig(a1/27);
[vb,q]=eig(b/27);
%b/27*vb(:,1)==q(1)*vb(:,1);