The nonlinear detector response of the infrared focal plane arrays (IRFPA) and its drift characteristics will result in the error of nonuniformity correction(NUC). In view of this fact, an S-curve response model based scene adaptive NUC algorithm is proposed. The algorithm takes logarithm to linearize the output signal of IRFPA, and employs adaptive filtering technique to correct the linear signal. The corresponding exponential transform is used to recover the corrected signal. The performance of the proposed algorithm is tested with artificial blackbody images and real infrared video sequence, respectively. The experimental results validate that the new proposed algorithm can promote the correction precision and expand the dynamic range of detector response considerably.