Two-dimensional Non-minimum Phase Signal Reconstruction

The existing techniques for reconstruction of one dimensional ( 1-D) signals from samples of its bispectrum are not efficient for reconstruction of (2-D) signals. In this paper we I;resent a new technique for 2-D nonminimum phase signal reconstruction which uses a small segment of the 4-D bis pectrurn. The approach uses the facts that the phase of the Fourier transform of a 2-D discrete signal is periodic and odd while the magnitude is periodic and even. Thus, they mu be expanded using Fourier series. The Fourier series coefficients of these expansions are then recovered from the Fourier series of bispectrum phase and magnitude. The implementation of the proposed technique is greatly simplified by using the FFT for computation of the Fourier series coefficients.