| The problem of fixed pattern noise (FPN), or spatial nonuniformity, arises due to the differing responses of each photodetector within a focal-plane-array (FPA) sensor. Although the response of each FPA detector is nonlinear, they are typically modeled linearly, having both a gain and bias component. The problem of FPN is complicated by the fact that the response of each FPA detector changes due to a variety of factors, causing the nonuniformity pattern to slowly drift in time. Thus, it is required that the FPN be continuously estimated and compensated for in order to guarantee clean, temperature accurate imagery throughout sensor operation.; In this work, a novel registration-based algebraic bias nonuniformity correction (NUC) technique is developed. A special radiometric form of the algorithm is also presented that allows for the non-obstructive calibration of the entire FPA. Two approaches for estimating gain nonuniformity are also presented. Motion estimation techniques, which are used by the various NUC algorithms to register consecutive image frames, are also discussed, along with the derivation of a new projection-based shift estimation algorithm. This novel technique has a high computational efficiency and is able to obtain reliable shift estimates in the presence of FPN.; The high-quality correction abilities of the presented NUC algorithms are demonstrated through application to real infrared data obtained from both cryogenically-cooled and uncooled infrared FPA sensors. A comprehensive theoretical and experimental error analysis is performed to study sources of error that degrade the nonuniformity compensator estimates produced by the NUC algorithms. The error analysis examines four sources of error that affect the algorithm's performance, namely, bilinear interpolation error, residual perimeter nonuniformity, shift estimation error and gain nonuniformity. Some specific applications where the NUC algorithms may be applied are considered, namely, image stabilization, SNR enhancement, resolution enhancement and polarization-based infrared imaging techniques. The capabilities of the Infrared Imaging Laboratory at UNM are also discussed and finally, avenues of future work are considered including possible NUC algorithm extensions, performance studies and other potential applications of the NUC algorithms. |
