In order to implement non-uniformity correction of infrared image, the neural network algorithm is commonly used to complete adaptive non-uniformity correction of infrared image at present, this method can update the correction factor automatically, but the algorithm is complex, and the correction output image contrast is low, the ghosting and image blurring appears. An improved neural network non- uniformity correction method is presented, the initial value of calibration coefficients are obtained by the use of high and low temperature blackbody, and the calculation of the output expect value in the iterative process is improved. The test results show that the method is simple and practical, and has a good non-uniform correction effect.