Python(pytorch) code for the paper: Ziyang Chen, Siming Zheng, Zhishen Tong, and Xin Yuan, "Physics-driven deep learning enables temporal compressive coherent diffraction imaging," Optica, 9(6): 677–680[pdf] [doi]

Coherent diffraction imaging (CDI), as a lensless imaging technique, can achieve a high-resolution image with intensity and phase information from a diffraction pattern. To capture high speed and high spatial-resolution scenes, we propose a temporal compressive CDI system. A two-step algorithm using physics-driven deep-learning networks is developed for multi-frame spectra reconstruction and phase retrieval. Experimental results demonstrate that our system can reconstruct up to 8(20) frames from a snapshot measurement. Our results offer huge potential for visualizing the dynamic process of molecules with large field-of-view, high spatial and temporal resolutions.

Figure 1.Reconstruction results for the complicated object. (a) Coded measurement; (b) Reference images of the moving object; (c) reconstructed spatial spectra; (d) 8 corresponding reconstructed spatial images by HIO algorithm; (e) 8 corresponding reconstructed spatial images by the proposed DNN-HIO algorithm. Boxes of different colors circle the parts where the contrast between the two results is more obvious.

1.Download the all the files via Baidu Drive (access code zsms) or One Drive and directly put the data in TC_CDI_Stage1.

@article{Chen:22,
author = {Ziyang Chen and Siming Zheng and Zhishen Tong and Xin Yuan},
journal = {Optica},
keywords = {Digital micromirror devices; Phase retrieval; Power spectral density; Ptychography; Spatial resolution; X ray imaging},
number = {6},
pages = {677--680},
publisher = {Optica Publishing Group},
title = {Physics-driven deep learning enables temporal compressive coherent diffraction imaging},
volume = {9},
month = {Jun},
year = {2022},
url = {http://opg.optica.org/optica/abstract.cfm?URI=optica-9-6-677},
doi = {10.1364/OPTICA.454582},
abstract = {Coherent diffraction imaging (CDI), as a lensless imaging technique, can achieve a high-resolution image with intensity and phase information from a diffraction pattern. To capture high-speed and high-spatial-resolution scenes, we propose a temporal compressive CDI system. A two-step algorithm using physics-driven deep-learning networks is developed for multi-frame spectra reconstruction and phase retrieval. Experimental results demonstrate that our system can reconstruct up to eight frames from a snapshot measurement. Our results offer the potential to visualize the dynamic process of molecules with large fields of view and high spatial and temporal resolutions.},
}