Image restoration of forward-looking infrared (FLIR) imagery has the p
otential to significantly improve the quality of images used by an aut
omatic target recognizer (ATR) or human observer. This study investiga
tes the feasibility of real-time image restoration algorithms and the
problem of measuring image quality as it relates to target acquisition
performance. This paper describes a technique for deriving small kern
el filters that efficiently restore and reconstruct. Subject to implem
entation constraints associated with efficient application, the filter
s optimize image fidelity to an 'ideal' close-range image, The paper d
escribes simulation experiments employing an end-re-end imaging system
model, experiments with actual images using a model-based characteriz
ation of an actual imaging system, and simulation experiments that ill
ustrate the utility of the system model and filtering in FLIR imaging
system design. (C) 1997 Published by Elsevier Science B.V.