On the use of regularization techniques in numerical inverse-scattering solutions for microwave imaging applications

This paper deals with integral-equation-based numerical methods for microwave imaging using regularization procedures to overcome ill-conditioning problems. The strong dependence of reconstruction quality on “a priori” information is discussed. Such information is required to select a suitable number of independent columns when using truncated pseudoinversions (or other regularization parameters, in different cases) for accurate dielectric reconstructions. Moreover, a criterion for the choice of the optimal number of independent columns is proposed, and the possibility of making this choice less critical by using a multiview version of the method is explored. Finally, a modified procedure is presented that further increases the range from which to choose the number of independent columns that allows one to achieve acceptable reconstructions