Signal processing based method for solving inverse scattering problems

The problem of reconstructing an image of the permittivity distribution inside a penetrable and strongly scattering object from a finite number of noisy scattered field measurements has always been very challenging because it is ill-posed in nature. Several techniques have been developed which are either computationally very expensive or typically require the object to be weakly scattering. I have developed here a non-linear signal processing method, which will recover images for both strong scatterers and weak scatterers. This nonlinear or cepstral filtering method requires that the scattered field data is first preprocessed to generate a minimum phase function in the object domain. In 2-D or higher dimensional problems, I describe the conditions for minimum phase and demonstrate how an artificial reference wave can be numerically combined with measured complex scattering data in order to enforce this condition, by satisfying Rouche's theorem. In the cepstral domain one can filter the frequencies associated with an object from those of the scattered field. After filtering, the next step is to inverse Fourier transform these data and exponentiate to recover the image of the object under test. In addition I also investigate the scattered field sampling requirements for the inverse scattering problem. The proposed inversion technique is applied to the measured experimental data to recover both shape and relative permittivity of unknown objects. The obtained results confirm the effectiveness of this algorithm and show that one can identify optimal parameters for the reference wave and an optimal procedure that results in good reconstructions of a penetrable, strongly scattering permittivity distribution.