Phase imaging with an untrained neural network

TLDR

Conventional computational imaging neural networks rely on supervised training with large datasets, which is impractical in many optical applications due to limited ground‑truth availability and system stability requirements. The study aims to eliminate the need for large labeled datasets by embedding a physical model into a deep neural network for phase imaging. The network integrates a complete physical model of image formation into a conventional deep neural network, enabling automatic optimization during inference. PhysenNet can reconstruct phase from a single diffraction pattern without prior training, automatically optimizing via the interplay between the network and the physical model, demonstrating a new paradigm for physics‑enhanced neural networks.

Abstract

Abstract Most of the neural networks proposed so far for computational imaging (CI) in optics employ a supervised training strategy, and thus need a large training set to optimize their weights and biases. Setting aside the requirements of environmental and system stability during many hours of data acquisition, in many practical applications, it is unlikely to be possible to obtain sufficient numbers of ground-truth images for training. Here, we propose to overcome this limitation by incorporating into a conventional deep neural network a complete physical model that represents the process of image formation. The most significant advantage of the resulting physics-enhanced deep neural network (PhysenNet) is that it can be used without training beforehand, thus eliminating the need for tens of thousands of labeled data. We take single-beam phase imaging as an example for demonstration. We experimentally show that one needs only to feed PhysenNet a single diffraction pattern of a phase object, and it can automatically optimize the network and eventually produce the object phase through the interplay between the neural network and the physical model. This opens up a new paradigm of neural network design, in which the concept of incorporating a physical model into a neural network can be generalized to solve many other CI problems.