Infrared focal plane devices generally present non-uniformity problems, and implementing non-uniformity correction on IRFPA is of great significance for improving image quality. The traditional IRFPA multi-point correction algorithm is of such problems as large amount of computation, low real-time performance and poor practical effect. Therefore, combining with the human visual characteristics, a multi-point correction method is proposed for IRFPA non-uniformity. The human eye vision has a threshold limitation on the gray resolution capability of the image, and with this resolution threshold, the calibration point of the IRFPA can be effectively compressed to generate a pixel number-correction coefficient table. Then, the IRFPA non-uniformity real-time compression correction can be implemented by looking up the coefficient table. The experimental results show that the proposed algorithm has better real-time performance than traditional algorithms, and the non-uniformity is reduced by 0.203%.