Related Experiment Video
Updated: Jul 29, 2025
09:19
Fabrication and Characterization of Disordered Polymer Optical Fibers for Transverse Anderson Localization of Light
Published on: July 29, 2013
11.5K
Unsupervised full-color cellular image reconstruction through disordered optical fiber.
Xiaowen Hu1, Jian Zhao2, Jose Enrique Antonio-Lopez1
1CREOL, The College of Optics and Photonics, University of Central Florida, Orlando, FL, 32816, USA.
Light, Science & Applications
|May 23, 2023
Summary
This study introduces unsupervised full-color imaging using disordered optical fibers, overcoming limitations of supervised deep learning for hard-to-reach areas. This method enables high-fidelity cellular imaging without paired data, enhancing flexibility and robustness.
Area of Science:
- Optics and Photonics
- Biomedical Imaging
- Machine Learning
Background:
- Supervised deep learning enhances fiber-optic imaging but requires paired data, limiting flexibility.
- Existing fiber optic methods struggle with high sampling density needed for unsupervised reconstruction.
- Disordered fibers offer a novel approach for high-quality imaging through Anderson localization.
Purpose of the Study:
- To demonstrate unsupervised full-color imaging with cellular resolution through meter-long disordered fibers.
- To develop a flexible and robust fiber-optic imaging system independent of paired data collection.
- To advance unsupervised image reconstruction techniques for challenging imaging scenarios.
Main Methods:
- Utilized meter-long disordered optical fibers for image transmission.
- Implemented a two-stage unsupervised image reconstruction process.
- Employed pixel-wise standardization and a generative adversarial network (GAN).
Main Results:
- Achieved unsupervised full-color imaging with cellular resolution in transmission and reflection modes.
- Demonstrated high-fidelity cell imaging within a 4 mm working distance.
- Showcased imaging robustness with a bent fiber (60° central angle) and enhanced cross-domain generality.
Conclusions:
- Unsupervised reconstruction with disordered fibers offers a flexible alternative to supervised methods.
- The developed technique provides high-fidelity, robust, and versatile cellular imaging capabilities.
- This approach broadens the applicability of fiber-optic imaging in various conditions and for unseen objects.
