Aiming at traditional neural networks non-uniformity correction (NUC) algorithm's disadvantages such as slow convergence, low correction precision and difficulty to meet the real-time engineering application requirements of infrared imaging system, an improved NUC algorithm for infrared focal plane arrays (IRFPA) based on neural network is proposed. The algorithm is based on linear response of detector, and in order to realize fast and synchronization convergence of correction parameters the each original image data is normalized to a value close to one. Experimental results show the method has the faster convergence speed and better vision effect than the traditional algorithms, and it is better applied in practical projects.

A nonlinear piecewise scheme for non-uniformity correction in infrared focal plane arrays (IRFPA) is presented. In this method, utilizing the nonlinear piecewise scheme of detector response has extended the larger dynamic range of IRFPA response and the higher correcting accuracy than the non-uniformity correction algorithms based on linear response model of IRFPA detector. Based on the principle of this method, the mathematical model is established. At last experimental results are given out. The results show that it has higher correction precision, fewer calculations, and is easier to implement real-time non-uniformity correction of IRFPA by hardware circuit.