The National Geospatial Intelligence Agency (NGA) and other Department of Defense (DoD) agencies use geospatial intelligence (GEOINT) derived from motion imagery to support U.S. national security. Full motion video (FMV) captured from airborne platforms is subject to multiple distortions and noise due to atmospheric disturbance, obscuration (dust, haze, smoke, and mist), and transmission defects such as compression artifacts and dropout. This project will develop the software algorithms to maximize the information content of a single image derived solely from information contained within FMV. Lickenbrock will leverage its experience in image registration, image processing, and deconvolution along with the use of a newly developed background-foreground separation algorithm. In Phase I, Lickenbrock will design the proof of concept system with an accompanying demonstration and provide the mathematical basis for the proposed Phase II system. The methods will combine information from multiple temporally related FMV frames to produce a single image superior to the original frames without introducing new artifacts. This will enable human observers to use still frames, rather than search through video for intelligence production. Another benefit is that automated object detector algorithm typically trained on high quality still frames will be effective with FMV.
