Phase imaging with an untrained neural network

By Fei Wang, Y. Bian, Haichao Wang, Meng Lyu, G. Pedrini, W. Osten, George Barbastathis, Guohai Situ. Light: Science & Applications, 2020

369 citations recorded by the source corpus.

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.

View Paper: https://www.semanticscholar.org/paper/408a6d45f32c493e95e09e9f77ac6dba1f06e8cc

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Chao Zuo
2. Shijie Feng
3. Jiaming Qian
4. Fei Wang
5. Wei Yin
6. Qian Chen
7. Chenglong Wang
8. Yixuan Li

Citations in the corpus

• Far-field super-resolution ghost imaging with a deep neural network constraint (2022, methodology): “Note that in the studies of phase imaging using untrained neural networks 34,35 , we did not observe such a phenomenon of resolution enhancement.”
• Deep learning in optical metrology: a review (2022, background): “Temporal phase retrieval: Wang et al.312 introduced a deep-learning scheme to the phase-shifting technique in FPP.”