Matrix formulation of the reconstruction of phase values from phase differences

TLDR: The phase reconstruction problem is formulated in terms of a vector-matrix multiplication and it is shown that previous solution methods are equivalent to this general description, and the errors in reconstruction are analyzed.

Abstract: Methods in speckle imaging and adapative optics, as well as a new technique in digital image restoration, require the calculation of the Fourier phase spectrum from measurements of the differences on a two-dimensional grid of the phase spectrum. The calculation of phases from phase differences has been analyzed in the literature and relaxation mechanisms for computing the phase have been derived by least-squares analysis. In the following paper we formulate the phase reconstruction problem in terms of a vector-matrix multiplication, and we then show that previous solution methods are equivalent to this general description. We also analyze the errors in reconstruction and reconcile previously published error results based on simulations with an analytical error expression derived from Parseval’s theorem. Finally, we comment upon the rate of convergence of phase reconstructions, and discuss numerical analysis literature which indicates that the methods previously published for phase reconstruction can be made to converge much faster.