A new scene-based nonuniformity correction (SBNUC) algorithm using spatial transform relationship between adjacent image frames is proposed. Based on spatial transform relationship between adjacent image frames, the gray value of current frame can be got by the weighted combination of some corrected pixels in the former frame. Moreover the gain and the offset are adaptively updated with the respective error gradient. The algorithm can avoid ghosting artifacts without considering other conditions. Corrected results of real infrared image sequences show that under the condition of ensuring image sharpness, the algorithm can decrease FPA nonuniformity high frequency information for increasing the image contrast, meanwhile bad pixels can be compensated automatically. This algorithm is proved to be stable and robust theoretically and practically.