An image fusion approach to the numerical inversion of multifrequency electromagnetic scattering data

We propose a numerical method to invert multifrequency electromagnetic scattering data. The data considered are relative to the observation of a given scene. The reflectivity of the scene is assumed to be an unknown function of the frequency of the illuminating radiation. The numerical method proposed is based on the assumption that the reflectivities of the scene at the different frequencies are piecewise constant functions and that their gradients have the same support. This assumption and an image fusion procedure based on nonlinear optimization allow high-quality reconstructions of the unknown reflectivities. The reconstruction procedure proposed is tested on experimental scattering data provided by the Institut Fresnel, CNRS, Marseille, France.