Improved infrared image neural network non-uniformity correction algorithm

The responsive of infrared focal plane arrays(IRFPA) is different; it will affect the quality of imaging system seriously. Non-uniformity correction technology will need in practical application. The calibrated images have the problems of blurring and existing ghost artifacts when use the traditional neural network correction algorithm. And it is bad for the observation of the target. After analysis the reasons for the problems in the traditional neural network correction algorithm, proposed the improved algorithm. Replace the mean filter, which used in the traditional algorithm, by the nonlinear filter. The corrected image by the improved algorithm not only a marked improvement in clarity, but also effectively eliminate the problem of artifacts in traditional algorithms.