Resolution-improved microwave tomography by means of hyperspectral analysis tools

In microwave tomography, the underlying scattering phenomenon is often linearized to retrieve the associated induced currents maps. These maps are then summed together to provide a good estimation of the target position. This qualitative reconstruction scheme is robust with noise but has a spatial limitation linked to the bandwidth of the radiation operator. We will show that the result of this inverse source problem can be considered in a different way by storing, incidence after incidence, the reconstructed induced current maps. As in hyperspectral imaging, this hypercube can be formalized as several views of the same scene. In hyperspectral imaging, the wavelength is the varying parameter. In harmonic microwave tomography, the incidence angle is the varying parameter. By applying image processing techniques such as principal component analysis, we will show that we can improve the spatial resolution of the qualitative maps. Synthetic results in an aspect-limited configuration will be presented.