The Temporal High-pass Filtering Nonuniformity Correction (THPF-NUC) algorithm has three major drawbacks, namely only correcting the offset coefficient, difficulty in choosing a proper time constant and unclear detail of corrected image. By analyzing the time constant’s effect on the correction quality, an improved THPF-NUC algorithm is proposed. Firstly, to guarantee the uniformity of the gain coefficient, the noise background image is subtracted from the output image of Infrared Focal Plane Arrays (IRFPA) as preprocessing. Then according to the motion degree obtained from the motion detection technique, adaptively adjust the time constant to design the low-pass filter. Lastly the high-boost filtering is adopted to improve the brightness of the whole IR image and enhance the details. The experiment on a real IR video shows that the proposed algorithm eliminates the ghosting artifact and the target fade-out effectively. And the corrected image is high quality with fast convergence rate, giving advantage to real-time engineering application.