Imaging through unknown scattering media based on physics-informed learning

TLDR

Imaging through scattering media is a hot topic in optics, with impressive results achieved via deep learning, but most DL approaches are purely data‑driven and lack physics priors, limiting generalization. This paper aims to develop a physics‑informed learning method that combines speckle‑correlation theory with deep learning to enable scalable imaging through unknown thin scattering media. The method trains a deep network using only one diffuser, integrating speckle‑correlation theory to guide reconstruction and achieving high fidelity for sparse objects. The approach solves the inverse problem with broad applicability, accurately reconstructing objects of varying complexity and sparsity even with diffusers of different statistical properties, extends the field of view of traditional speckle‑correlation methods, and paves the way for practical scattering imaging.

Abstract

Imaging through scattering media is one of the hotspots in the optical field, and impressive results have been demonstrated via deep learning (DL). However, most of the DL approaches are solely data-driven methods and lack the related physics prior, which results in a limited generalization capability. In this paper, through the effective combination of the speckle-correlation theory and the DL method, we demonstrate a physics-informed learning method in scalable imaging through an unknown thin scattering media, which can achieve high reconstruction fidelity for the sparse objects by training with only one diffuser. The method can solve the inverse problem with more general applicability, which promotes that the objects with different complexity and sparsity can be reconstructed accurately through unknown scattering media, even if the diffusers have different statistical properties. This approach can also extend the field of view (FOV) of traditional speckle-correlation methods. This method gives impetus to the development of scattering imaging in practical scenes and provides an enlightening reference for using DL methods to solve optical problems.