Michael Werth, The Boeing Company; Trent Kyono, The Boeing Company; Jacob Lucas, The Boeing Company; Justin Fletcher, SMC/DirSP-G; Ian McQuaid, Air Force Maui Optics and Supercomputing Site; Michael Brannon, Air Force Research Laboratory
Keywords: QWID, MFBD, LUCID, atmospheric physics, imaging, imaging through turbulence, deep learning
Abstract:
Multi-frame blind deconvolution (MFBD) techniques are often developed without foreknowledge of the incoming data quality. If a multi-frame ensemble contains a mix of high- and low-quality images, the image reconstruction quality may be worse than if only the high-quality images are processed. We report on a new MFBD algorithm that uses image quality scores to reduce processing time and improve reconstruction quality. These scores are automatically provided by a neural network that has been trained to assess the quality of ground-based observations of LEO satellites. This implementation, titled Quality-Weighted Iterative Deconvolution (QWID), is demonstrated on realistic ground-based observations.
Date of Conference: September 15-18, 2020
Track: Adaptive Optics & Imaging