eigenfaces : how to compute weights and how to classify new image
2 次查看(过去 30 天)
显示 更早的评论
hello i want to classify images use eigen faces i transposed the images of train set to eiganface but now i have issue of weights computation and classifiying new image
%%----------------------- % here the issue i am not sure % 1 computing the weights Image = sum(weights* eigenface) ? % 2 new image classification process ?**
//=================
% i have M=4 images N*N N=50
[ st.data{1} st.data{2} ; st.data{3} st.data{4} ];
% average image
for k=1:M
st.data{k} = im2single(st.data{k});
avImg =avImg + (1/M)*st.data{k};
end
%%normalize (remove mean)
for k=1:M
st.dataAvg{k} = st.data{k} -avImg;
end
%%generate A = [ img1(:) img2(:) ... imgM(:) ];
A = zeros(N*N,M);% (N*N)*M 2500*4
for k=1:M
A(:,k) = st.dataAvg{k}(:);
end
% covariance matrix small dimension (transposed)
C = A'*A;
%%eigen vectros in small dimension
[ Veigvec,Deigval ] = eig(C);% v 4*4 e 4*4 only diagonal 4 eigen values
% eigan face in large dimension A*veigvec is eigen vector of Clarge
Vlarge = A*Veigvec;% 2500*4*4*4 =2500 *4
% reshape to eigen face
eigenfaces=[];
for k=1:M
c = Vlarge(:,k);
eigenfaces{k} = reshape(c,N,N);
end
x=diag(Deigval);
[xc,xci]=sort(x,'descend');% largest eigenval
* **%%-----------------------
% here the issue i am not sure
% 1 computing the weights Image = sum(weights* eigenface) ?
% 2 new image classification process ?** *
%%weights
for mi=1:M % image number
for k=1:2 % eigen face for coeff number
wi(mi,k) = sum(A(:,mi).* eigenfaces{xci(k)}(:)) ;
end
end
%%classify new img mic
testFaceMic = imread('C:\Users\michaels.DSI\Desktop\faces\class\mic3.jpg','jpg');
testFaceMic =rgb2gray(testFaceMic);
testFaceMic = imresize(testFaceMic,[N N]);
testFaceMic = im2single(testFaceMic);
Aface = testFaceMic(:)-avImg(:);
w1face = sum(Aface.* eigenfaces{xci(1)}(:)) ;
w2face = sum(Aface.* eigenfaces{xci(2)}(:)) ;
0 个评论
回答(0 个)
另请参阅
类别
在 Help Center 和 File Exchange 中查找有关 Statistics and Machine Learning Toolbox 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!