Improving the Incoherence of a Learned Dictionary via Rank Shrinkage | MIT Press Journals & Magazine | IEEE Xplore