The long-range targets in the sky appear as dim and small targets in the infrared detector， which have unconspicuous gray scale and area characteristics in the image.Due to the influence of the clouds in the sky， the image background may include large areas with high brightness or complex texture structure， which may affect the detection probability and false alarm rate of the system.Therefore， the background of image must be suppressed.In this paper， Top-Hat transform is used together with the image segmentation algorithm based on Markov random field for image processing.The simulation results show that this method can effectively restrain the background while retaining the real targets.