Abstract: The components of the space noise in an infrared image with the sky background were discussed. The reasons for forming all types of noises were analysed. The noise types of the IR image were classified and their effects on the image were investigated. The analysis indicates that the space nosie in the IR image mainly represents the Gauss noise and spicedsalt noise. An improved median filtering based on a simple threshold is introduced by considering the difference in the noise disposal of the averaging method and the median filtering. For the different thresholds, the operators are different too. The weight selection of every operator in the M×N area depends on the gray median in the area.The more the gray value is close to the median, the more the weight of operator is. The background of the image becomes smoother after the preprocessing.The SNR and improved PSNR increase by 1.2dB. It indicates that the improved median filtering works better than the traditional ones.