My understanding is that you need to apply masks at the sampling stage, and there are legitimate researchers trying to play with that concept using single pixel cameras. Never heard of anyone else try compressed sensing with regular CCD or CMOS sensor array cameras.