An adaptive output window technique based on gray redundancy for infrared image is proposed in order to overcome the disadvantage of infrared image gray distribution. The adaptive output window threshold is calculated from the histogram,if the pixels sum of the same gray level is zero,this gray level is compressed. Those gray levels which pixel sum is less than the threshold are replaced by gray level of the threshold. Then effective gray-levels are distributed equidistantly. This algorithm can eliminate gray redundance and no gray-level lost. The results show that the image quality has been improved. Theis algorithm achieved real-time processing.