The fixed pattern noise of uncooled infrared focal plane arrays has the similar characteristics with the stripe noise. By researching the moment matching and temporal high-pass filter, a novel multiscale temporal moment matching nonuniformity correction algorithm was proposed. In the proposed algorithm, global motion was first identified by the Gaussian pyramid of the adjacent uncorrected images. Then the temporal moment matching was performed in all levels of Gaussian pyramids and Laplacian pyramids. The experimental results with the real infrared video sequences have shown that the proposed algorithm can significantly increase the convergence speed and reduce the ghosting artifacts.