Code for work: Differential ghost imaging with learned modulation patterns.
All codes run on the Pytorch framework.
Here, DGI-SLNN stands for differential ghost imaging - single layer nerual network.
If you find this project helpful or interesting, please cite our work:
@article{PhysRevApplied.22.014023,
title = {Differential ghost imaging with learned modulation patterns},
author = {Wang, Xiao and Cheng, Pengxiang and Chen, Huaijian and Zhao, Shupeng and Ma, Guangdong and Zhang, Yongchang and Zhang, Pei and Gao, Hong and Liu, Ruifeng and Li, Fuli},
journal = {Phys. Rev. Appl.},
volume = {22},
issue = {1},
pages = {014023},
numpages = {14},
year = {2024},
month = {Jul},
publisher = {American Physical Society},
doi = {10.1103/PhysRevApplied.22.014023},
url = {https://link.aps.org/doi/10.1103/PhysRevApplied.22.014023}
}Unlike conventional imaging with 2D array sensors featuring millions of pixels, ghost imaging enables the use of advanced detector technologies, in turn, was given advantages such as high signal-to-noise ratio, wide spectral range and robustness to light scattering. This causes an extremely time-consuming measurement process, which is difficult to meet the needs of high-quality real-time imaging. This paradox becomes notably salient, especially in the context of utilizing non-orthogonal modulation patterns, such as the speckles generated by rotating ground glass. Efficient modulation patterns and advanced reconstruction algorithms are widely studied as two main ideas to solve the above problem. Here, we perform real-time, high-fidelity differential ghost imaging (DGI) at a low sampling ratio of 6.25% by proposing a compact physically guided single-layer neural network with the DGI algorithm embedded. Simulations and experiments show that, once the learned modulation patterns are obtained, our scheme can achieve fast, high-quality, and noise-robust DGI without the need for complex iterative optimization algorithms or subsequent optimization neural networks. Our scheme opens up new horizons for exploring more efficient modulation patterns for ghost imaging by deeply combining physical priors.
One need to download the corresponding datasets and perform data preprocessing before running the demo: Demo_train.ipynb.
One can create the same environment using the 'environment.yml' file.
python=3.9.1
pytorch=2.0.1
pytorch-cuda=11.8
We are including a color version of the target image used in our experiment. Enjoy!
The author retains all rights to this photograph.
For academic and non-commercial use only.