See Figure 2 in http://statweb.stanford.edu/~candes/papers/RobustPCA.pdf http://statweb.stanford.edu/~candes/papers/RobustPCA.pdf There's lots of work on backg

See Figure 2 in http://statweb.stanford.edu/~candes/papers/RobustPCA.pdf http://statweb.stanford.edu/~candes/papers/RobustPCA.pdf There's lots of work on background estimation. This is just one approach using nuclear norm minimisation.