an urgent question about PCA on image!
I've trained an eigenvector matrix from 100 pictures through PCA algorithm.Then I tried to recover one of these 100 images using the matrix I gained,and succeed. It,however,was failure when I tried to recover those images other than in that 100 training samples.That is,information loss greatly.BTW,recovering in there means projecting a picture and then back project it.
I am sorry, but your question is far from understandable. Can you please improve its quality?
I've trained an eigenvector matrix from 100 pictures through PCA algorithm.Then I tried to recover one of these 100 images using the matrix I gained,and succeed. It,however,was failure when I tried to recover those images other than in that 100 training samples.That is,information loss greatly.BTW,recovering in there means projecting a picture and then back project it.Please let me know if you are clear this time so I can edit the question again.
Euhm is that not normal when doing dimensionality reduction? You are explicitly throwing away tons of information!
yeah,I'm trying to extract the main information in order to reduce the noise and meaningless information in the pictures.As you told,if I can't do so or there are other ways to extract the main information using PCA?