Nonuniformity of infrared focal plane arrays (IRFPA) is a key problem that must be solved in its applications. Based on the theory of wavelet transform, a scene-based nonuniformity-correction (NUC) technique for IRFPA is presented. In this algorithm, the infrared image sequence is decomposed by the discrete sequence wavelet transform with an appropriate wavelet function. Its corresponding statistics are calculated, and then coefficients of the offset and the gain for IRFPA NUC are achieved ultimately. The algorithm achieves good nonuniformity-correction capability, which is validated by experiments of real infrared imaging procedure. And the algorithm can auto-adapt to the slow-change parameters that affect the NUC and can restrain the “artificial ghost” in some degree that appears in other scene-base NUC methods.