Presentation
3 March 2022 Stabilized imaging through dynamic scattering media based on generative adversarial networks
Xiaowen Hu, Jian Zhao, Jose Enrique Antonio-Lopez, Stefan Gausmann, Rodrigo Amezcua-Correa, Axel Schülzgen
Author Affiliations +
Abstract
The dynamic nature of scattering media, such as living tissues, greatly degrades the images of hidden objects reconstructed by deep neural networks. We show that if the dynamic scattering medium is followed closely, the relationship between distorted reconstructed images and objects can nevertheless be found by generative adversarial networks. We numerically verify this method in a general case where the dynamic scattering medium is formulated as an evolving transmission matrix. Then we experimentally apply the method in imaging cell samples through a disordered optical fiber system with imaging depth variations. Increased robustness in imaging is observed in both cases.
Conference Presentation
© (2022) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Xiaowen Hu, Jian Zhao, Jose Enrique Antonio-Lopez, Stefan Gausmann, Rodrigo Amezcua-Correa, and Axel Schülzgen "Stabilized imaging through dynamic scattering media based on generative adversarial networks", Proc. SPIE PC11974, Biomedical Applications of Light Scattering XII, PC119740B (3 March 2022); https://doi.org/10.1117/12.2607202
Advertisement
Advertisement
KEYWORDS
Scattering media

Scattering

Speckle

Imaging systems

Machine learning

Neural networks

Numerical simulations

Back to Top