| Objective: | Develop and implement the next generation of multi-frame bind deconvolution approaches that are tailored to work under daylight and strong turbulence imaging conditions. These approaches will push beyond the standard imaging model of a single channel electro-optical system with statistical independence between frames collected.
| Description: | Multi-frame blind deconvolution (MFBD) algorithms have been in use for years for ground to space image enhancement applications. These algorithms generally use a conventional single channel electro-optical imaging model with the assumption of independence between frames. Past algorithms have typically been developed to support imaging from astronomical observatories in low turbulence conditions during the terminator period of the objects orbit. Atmospheric turbulence distorts the incoming wave front and causes a corresponding degradation in the image plane. Adaptive optics (AO) systems are employed to remove a significant amount of the distortion, but they have limitations in the amount of turbulence that can be effectively mitigated. MFBD algorithms can be used without an AO system or in addition to an AO system for additional mitigation of higher order aberrations. MFBD algorithms use the assumption that the object remains the same over a set of images and the relationship between the frames can be used to extract the distortions from each image frame and thus reconstruct the object. With higher frame-rate cameras becoming available, the interframe correlation is increasing and one can no longer assume independence between frames. Research and development efforts are required to push the envelope of high resolution ground to space imagery operating conditions beyond low turbulent conditions and into full daylight. During daylight the heat from the sun increases the atmospheric turbulence. This leads to a significantly more turbulent environment compared to normal night time operations. As the turbulence increases, new constraints might be found that increase the performance of MFBD algorithms to reconstruct the object. Potential constraints could be additional optical channels using embedded information in the photon data such as polarization or using non-standard imaging models that leverage interframe statistical dependence. Past MFBD algorithms using the conventional model have been shown to be mathematically optimal under a certain set of assumptions via Cramer-Rao lower bound information theory. It is expected that this research topic will use similar techniques to identify and implement MFBD constraints to the strong turbulence problem. The effects of the new constraints on the image model should show the ability of that constraint to improve on the algorithms ability to mitigate the increased atmospheric turbulence. Some basic research has been conducted in this area. Additional review of MFBD algorithms in other fields and the constraints used should be made for potential leverage. This topic looks to expand previous findings and increase their technology readiness level and will enable better space situational awareness (SSA).
| | PHASE I: Develop the mathematical basis for new MFBD approaches and constraints that are tailored to daylight imaging in strong turbulence. Use Cramer-Rao bound analysis on new constraints or imaging system models to identify the most appropriate approaches.
| | PHASE II: Implement algorithms discovered in Phase I in a high performance computing-based MFBD software package and demonstrate its performance with real and simulated data. The Air Force Research Laboratory will provide access to real data test cases and associated benchmarks for comparative purposes at no cost to the contract.
| | PHASE III
| | DUAL USE COMMERCIALIZATION:
Military Application: New approaches will address increased atmospheric turbulence degradation in electro-optical (EO) imaging as capabilities push toward daylight imaging and thus will enable better SSA.
Commercial Application: The ability to use EO imaging in higher turbulence regimes has applicability in the astronomy community as well as other imaging technologies such as medical imaging.
| References: | 1. Matson, C.L., et al, “A fast and Optimal Multi-Frame Blind Deconvolution Algorithm for High-Resolution Ground-Based Imaging of Space Objects,” Applied Optics, Vol. 48, No.1, Pages A75-A92 (2009).
2. Doug Hope, Stuart Jefferies and Cindy Giebink, “ Fourier Constrained Blind Restoration of Imagery Obtained in Poor Imaging Conditions,” Proc. AMOS technologies Conference, Maui, HI, 2007.
| | Keywords: | MFBD, Image Enhancement, Daylight Imaging, Turbulence, High Performance Computing |