Based on the analysis of fixed pattern noise on the basis of frequency distribution, a correction method using stationary wavelet was proposed. The infrared image sequence was decomposed by a proper wavelet function, thereby the gain and offset coefficients were estimated and eventually the infrared focal plane array nonuniformity correction was realized. The resolution of low-frequency part was improved by the virtue of the wavelet with its multi-resolution properties, so the “artificial ghost” phenomenon which appears in some scene-based NUC algorithm was restrained. Experiments on real infrared image sequence prove the superiority of the algorithm, and have achieved satisfactory results.