A Mild Data-Driven Approach Based on a Lebesgue-Space Inversion Procedure for Microwave Imaging Applications

In this paper, a mild data driven inversion technique for microwave imaging applications is proposed. The developed approach relies upon the use of a variable-exponent inversion procedure, in which a-priori information about the class of inspected targets is included through a data-driven regularization term. The developed technique has been preliminarily tested considering the characterization of inclusions in partially known host structures, showing enhanced reconstruction capabilities with respect to the bare inversion scheme.