Iterative hard thresholding for compressed data separation (Q2190470)

The authors discuss the problem of compressed data separation i.e., the possibilities of reconstructing the subcomponents of a signal that is sparse in two morphological different dictionaries. Their approach is based on an iterative thresholding algorithm based on the dictionary structure. The main result is that if the measurement data satisfy the RIP condition with respect to the global dictionary and the two dictionaries are subject to a mutual coherence condition, an approximate reconstruction of the distinct components is possible.