Least-squares phase recovery from the bispectrum phase: an algorithm for a two-dimensional object

We discuss the problem of Fourier phase recovery from measurements of the bispectrum phase and present the algorithm for a two-dimensional object by extending the method introduced by Marron et al. [ J. Opt. Soc. Am. A7, 14 ( 1990)]. The main objective is to unwrap the measured bispectrum phase so that the signal parts of the unwrapped bispectrum phases are linked to the Fourier phases by a linear system of equations: These linear equations are solved in a least-squares sense to recover the Fourier phases. However, in the unwrapping process we have to solve an extremely large linear system of equations, which creates a serious difficulty. To cope with this difficulty, we construct a recursive method that permits us to solvethese equations efficiently without constructing their coefficient matrix. By computer simulations, we verify that the proposed Fourier-phase-recovery algorithm is effective.