Microwave tomography for compact-support dielectric objects

A technique for imaging compact-support dielectric objects from microwave backscattering data is proposed. It is based on a linear data model, which permits the evaluation of the object spectrum on a band-pass region, and the application of the Projected Landweber method for spectrum extrapolation. The performance of the resulting algorithm is then investigated as a function of the relaxation parameter and the possible preconditioning strategy. Several experimental results are shown both for noiseless and noisy cases. This algorithm is shown to be an additional possibility to be taken into consideration, alongside the classical linear algorithms and the fully nonlinear ones. The experimental results show that the solutions are far better than the ones obtained by linear methods, yet require far fewer computations than the nonlinear approaches. 9© 1998 John Wiley & Sons, Inc.