In an attempt to improve the convergence rates of band-limited image restoration, we derive and apply two algorithms employing nonlinear projections onto closed convex sets. The performances of these algorithms are compared with the well-known Gerchberg—Papoulis (GP) procedures for several cases and are shown to have superior initial convergence rates although eventually they behave like the GP procedures. Both algorithms are shown to converge weakly (i.e., inner product convergence) to the unknown image.
© 1982 Optical Society of America
H. Stark, S. Cruze, and G. Habetler, "Restoration of optical objects subject to nonnegative spatial or spectral constraints," J. Opt. Soc. Am. 72, 993-1000 (1982)