How SVD works to find similar blocks in image?

I am writing a code to find out which part of an image has been duplicated, so i need to find similar blocks in image. I need to use SVD (singular value decomposition) to find which blocks match together . Do you know how i can apply SVD on image to find similar b开发者_开发问答locks?

Please help me . Thanks

SVD can be used to compare images (blocks) by applying SVD to two images and then comparing only the SVD approximations of them, up to a certain order. This can extended to more than two images by treating the SVD result as a feature vector (-> machine learning) for e.g. k-nearest neighbor or Cluster analysis.

SVD based image compression ("approximation") is e.g. described here: http://online.redwoods.cc.ca.us/instruct/darnold/laproj/fall2001/adamdave/textwriteup.pdf .