A NEW NONLINEAR OPTIMIZATION TECHNIQUE FOR RECONSTRUCTING DIELECTRIC SCATTERERS WITH POSSIBLE HIGH CONTRASTS

In this article we present a new time-domain nonlinear optimization te chnique for the profile reconstruction of two-dimensional dielectric s catterers. Conventional reconstruction techniques, for example, the Bo rn iterative and conjugate gradient methods, employ the free-space Gre en's function, and are useful primarily for low-contrast scatterers. A lthough the distorted Born iterative method uses the Green's function from the last iteration step, and therefore has a faster convergent ra te, it also is only suitable for reconstructing weakly scattering diel ectric objects. In contrast, the method reported herein is based on a nonlinear optimization technique that employs an updated Green's funct ion at each iteration stage, and is capable of solving stronger nonlin ear problems involving objects with high contrasts. To illustrate the application of the technique, we present the numerical reconstruction results for a scatterer with relative permittivities up to 13.5 and sh ow that the algorithm has excellent convergence properties. (C) 1997 J ohn Wiley & Sons, Inc.