A non-uniformity correction method for infrared images based on calibration combined scene is proposed to address the problems in the use of calibration and scene based methods. The blackbodies at different temperatures in the laboratory are used to construct second-order calibration parameter of infrared detection systems. Infrared images are performed non-uniformity correction by neural network algorithms to adjust the correction coefficients dynamically. Algorithm verification is performed through constant temperature blackbody and real external scenes, it is proved that the non-uniformity of the image can be reduced effectively, which is an effective method for infrared image non-uniformity correction.