We consider an instance of the phase-retrieval problem, where one wishes to recover a signal (viewed as a vector) from the noisy magnitudes of its inner products with locally supported vectors. Such measurements arise, for example, in ptychography, which is an imaging technique used in lense-less X-ray microscopes and in optical microscopes with increased fields of view.
Starting with the setup where the signal is one-dimensional, we present theoretical and numerical results on an approach that has two important properties. First, it allows deterministic measurement constructions (which we give examples of). Second, it uses a robust, fast recovery algorithm that consists of solving a system of linear equations in a lied space, followed by finding an eigenvector (e.g., via an inverse power iteration). We also present extensions to the two-dimensional setting.
This is joint work with M. Iwen, B. Preskit, and A. Viswanathan.