Enhanced computational imaging with encoding masks optimized by deep neural networks

Bryan I. Vogel; Michael F. Finch; Kevin J. Miller

doi:10.1117/12.2587173

12 April 2021 Enhanced computational imaging with encoding masks optimized by deep neural networks

Bryan I. Vogel, Michael F. Finch, Kevin J. Miller

Proceedings Volume 11740, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XXXII; 117400I (2021) https://doi.org/10.1117/12.2587173
Event: SPIE Defense + Commercial Sensing, 2021, Online Only

Abstract

This research investigates the usage of deep neural networks (DNNs) to optimize encoding masks that, when combined with digital imaging systems, can recover high spatial frequency content that would otherwise be filtered out by the OTF. It has been shown that Fourier ptychographic photography (FPP) can recover high resolution imagery by imaging scenes that have been illuminated by a controlled light source. Instead of a controlled light source, this research utilizes light encoding masks that are optimized by a DNN to allow for decreased blur and increased resolution of recovered imagery. A masks generator is optimized in a generative adversarial network (GAN) where generated masks are used to recover imagery through a FPP phase recovery algorithm. The masks and recovery algorithm are optimized according to a loss function comparing the recovered imagery to the pristine, undegraded imagery as dictated by a Wasserstein critic, a DNN model. This method creates masks optimal for recovering high frequency spatial information of specific imaged object types based on the training dataset used.

Conference Presentation

Citation Download Citation

Bryan I. Vogel, Michael F. Finch, and Kevin J. Miller "Enhanced computational imaging with encoding masks optimized by deep neural networks", Proc. SPIE 11740, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XXXII, 117400I (12 April 2021); https://doi.org/10.1117/12.2587173

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available