Passive superresolution imaging of incoherent objects

Abstract

We investigate Hermite Gaussian Imaging (HGI) -- a novel passive super-resolution technique -- for complex 2D incoherent objects in the sub-Rayleigh regime. The method consists of measuring the field's spatial mode components in the image plane in the overcomplete basis of Hermite-Gaussian modes and their superpositions and subsequently using a deep neural network to reconstruct the object from these measurements. We show a three-fold resolution improvement over direct imaging. Our HGI reconstruction retains its superiority even if the same neural network is applied to improve the resolution of direct imaging. This superiority is also preserved in the presence of shot noise. Our findings are the first step towards passive super-resolution imaging protocols in fluorescent microscopy and astronomy.