In the field of imaging through a single multimode fiber (MMF), various methods have been explored to reconstruct the object. In all these methods, deep learning is more robust and the imaging system is less complex. In this paper, we verified the reconstruction of images from a few-mode fiber with deep learning methods based on the intensity of the light filed with simulation and experiments. Compared with another broadly used method that depends on the measurement of optical fiber transmission matrix, the reconstructed images with deep learning method are not affected by the circular field of view of the fiber aperture.

Date Published: 5 November 2018
PDF: 10 pages
Proc. SPIE 10816, Advanced Optical Imaging Technologies, 108160H (5 November 2018); doi: 10.1117/12.2502770

Published in SPIE Proceedings Vol. 10816:
Advanced Optical Imaging Technologies
Xiao-Cong Yuan; Kebin Shi; Michael G. Somekh, Editor(s)