Data-driven superresolution imaging in disordered media

We propose a methodology that exploits large and diverse datasets to accurately estimate the ambient medium's Green's functions in strongly scattering media. Given these estimates, excellent imaging results are achieved, with a resolution that is better than that of a homogeneous medium. This phenomenon, known as superresolution, arises because the ambient scattering medium effectively enlarges the physical imaging aperture. While superresolution has been demonstrated and analyzed extensively in the context of physical time reversal, time reversal itself is not imaging. Our proposed methodology, based on either conventional optimization methods or neural networks, makes it possible to achieve superresolution imaging in complex media.