The cause of the target fade-out in the traditional neural network nonuniformity correction algorithms for infrared focal plane array(IRFPA)was studied. A new combinational algorithm for nonuniformity correction was proposed on the basis of analyzing the strengths and limitations of edge-directed NN scheme(ED-NN-NUC) and nonuniformity correction combining one-point calibration and NN-NUC. The new algorithm includes three modules which are pre-correction, rough correction and accurate correction. The real infrared image including dim targets was used to validate the proposed algorithm. The results show that the proposed algorithm can effectively eliminate the target fade-out, and it is less time consuming than ED-NN-NUC.