With the development of IRFPA techniques, IR staring imaging system has been applied in more and more fields. In these applications, the residual non-uniformity and the non-uniformity correction(NUC) algorithms of the output IR images are of major concern. In order to resolve this problem, the non- uniformity sources and their characteristics in spatial frequency domain were studied firstly in this paper, and then the advantages and the disadvantages of both Calibration-based and Scene-based NUC algorithms were taken into account, and finally a combined non-uniformity correction algorithm including Blind-pixel Detection and Compensation was put forward to implement the non-uniformity correction, and the experiments results were given as well. The results shows that the combined non-uniformity correction algorithm can prevent the RNU to increase with the changes of time and working environments well, and has excellent performances in NUC technical applications.