Charles Matson (Air Force Research Laboratory), Charles C. Beckner, Jr. (Air Force Research Laboratory), Kathy Borelli (KJS Consulting), Tom Soo Hoo (Boeing LTS), Shiho You (Boeing LTS), Brandoch Calef (Boeing LTS) Maria Murphy (Maui High-Performance Computing Center), Ron Viloria (Maui High-Performance Computing Center)
Keywords: Imaging
Abstract:
One of the missions of the Air Force Maui Optical and Supercomputing (AMOS) site is to generate high-resolution images of space objects using the Air Force telescopes located on Haleakala. Because atmospheric turbulence greatly reduces the resolution of space object images collected with ground-based telescopes, methods for overcoming atmospheric blurring are necessary. One such method is the use of adaptive optics systems to measure and compensate for atmospheric blurring in real time. A second method is to use image restoration algorithms on one or more short-exposure images of the space object under consideration. At AMOS, both methods are used routinely. In the case of adaptive optics, rarely can all atmospheric turbulence effects be removed from the imagery, so image restoration algorithms are useful even for adaptive-optics-corrected images. Historically, the bispectrum algorithm has been the primary image restoration algorithm used at AMOS. It has the advantages of being extremely fast (processing times of less than one second) and insensitive to atmospheric phase distortions. In addition, multi-frame blind deconvolution (MFBD) algorithms have also been used for image restoration. It has been observed empirically and with the use of computer simulation studies that MFBD algorithms produce higher-resolution image restorations than does the bispectrum algorithm. MFBD algorithms also do not need separate measurements of a star in order to work. However, in the past, MFBD algorithms have been factors of one hundred or more slower than the bispectrum algorithm, limiting their use to non-time-critical image restorations. Recently, with the financial support of AMOS and the High-Performance Computing Modernization Office, an MFBD algorithm called Physically-Constrained Iterative Deconvolution (PCID) has been efficiently parallelized and is able to produce image restorations in only a few seconds. In addition, with the financial support of AFOSR, it has been shown that PCID achieves or closely approaches the theoretical limits to image restoration quality for a variety of scenarios. For these reasons, PCID is now being transitioned to being the site-wide image restoration algorithm. Because the algorithm can be complicated to use, a GUI is being developed to be the front end to the PCID algorithm. This interface, called the Advanced SPeckle Image Reconstruction Environment (ASPIRE) version 2.0, is the next generation of the current ASPIRE GUI used as a front end to the bispectrum algorithm. ASPIRE 2.0 will be the front-end GUI to PCID, the bispectrum algorithm, and the AMOSphere database. In this presentation we describe ASPIRE 2.0 and PCID and how to use them to obtain high-resolution images.
Date of Conference: September 12-15, 2007
Track: Imaging