We test this model on three tasks: non-uniform motion blur removal, lossy-compression artifact reduction