A discrete Fourier transform (DFT) based algorithm for solving a quadr
atic cost functional is proposed; this regularized functional allows o
ne to obtain a consistent gradient field from an inconsistent one. The
calculated consistent gradient may then be integrated by use of simpl
e methods. The technique is presented in the context of the phase-unwr
apping problem; however, it may be applied to other problems, such as
shapes from shading (a robot-vision technique) when inconsistent gradi
ent fields with irregular domains are obtained. The regularized functi
onal introduced here has advantages over existing techniques; in parti
cular, it is able to manage complex irregular domains and to interpola
te over regions with invalid data without any smoothness assumptions o
ver the rest of the lattice, so that the estimation error is reduced.
Furthermore, there are no free parameters to adjust. The DFT is used t
o compute a preconditioner because there is highly efficient hardware
to perform the calculations and also because it may be computed by opt
ical means. (C) 1997 Optical Society of America.