Proceedings Article | 22 October 2010
KEYWORDS: Nonuniformity corrections, Infrared imaging, Infrared radiation, Imaging systems, Calibration, Independent component analysis, Detection and tracking algorithms, Interference (communication), Complex systems, Temperature metrology
Going with Infrared Focal Plane Array (IRFPA) development, the application of infrared imaging system is more and
more extensive, it's well known that the Non-Uniformity Correction (NUC) is the only necessary data soft processing in
the whole infrared imaging data link, it will be seen from this that the NUC quality stand or fall influences the final
imaging product quality directly, for target detection and identification system, it increases the complexity and timeliness
of the target detection and identification algorithm undoubtedly. Currently, the Non-Uniformity Correction (NUC)
algorithm can be divided two classifications: the one is that Non-Uniformity Correction based on calibration source, this
algorithm assumes the infrared system response characteristic is linear, takes the dark current and gain as the two
correction parameters, but for nonlinear, especially for the response drift characteristic and the ambient temperature
change, the higher the system sensibility is, the greater the influence is and the higher the design requirements for system
stray radiation are. The correction effectiveness is limited seriously; the another is adaptive correction algorithm based
on scene (SBNUC), it can be subdivided time domain, space domain and motion estimation processing algorithms,
although it do not need physical calibration source and also reduces the influence of system response drift to a certain
degree, but the requirement is rigorous for statistics specimen and size, and the rapidity of convergence and stability are
different.
In this paper, according to blind information source decomposition technique, the infrared image is divided to signal and
noise as two information sources, a new Non-Uniformity Correction method based on Fast Independent Component
(FastICA) blind separation is put forward. By means of the experimental contrast analysis for the linear correction
algorithm and constant statistics algorithm of real infrared image, by this new algorithm, the influence of the system
response drift and the ambient temperature change for the linear correction algorithm based on physical calibration
source is not only suppressed, but also the shortages of the scene-based Non-Uniformity correction (SBNUC) in statistics
specimen and size are overcome partly. The experimental result proved the effectiveness of this algorithm in the paper
which effectively separated the signal and noise of the infrared image. At the same time, the algorithm in the paper
supplied a new solution of Non-Uniformity Correction (NUC) by the experiment.
