Executive Summary : | Non-invasive imaging plays a pivotal role in various science and engineering applications, particularly in earth and medical sciences. It allows us to examine an object without having any adverse impact either on the object itself or on the environment. In many cases, such as deep earth studies, it is not even possible to have direct measurements with current technologies. However, transforming the indirect observations into physical properties requires solving an inverse problem that is ill-posed in nature. Joint analysis of multiphysics data offers valuable advantages in reducing the nonuniqueness of model parameter estimation. The proposed project aims to develop an inverse modeling algorithm that imposes structural constraints on the estimated model parameter in an electromagnetic imaging problem. It is achieved by posing an objective function that strives to obtain a structurally-consistent model. It will be devised by augmenting the objective function with a function that optimizes the cross-gradient of the model parameter to be evaluated and a model utilized for structural information. The model used for achieving structural similarity can either be available before inversion or computed during the current inversion. An inexact Gauss-Newton method will be employed for the optimization as it is memory efficient. The primary goal of the development is to work on earth science-related problems; however, it can be effortlessly extended to medical imaging. |