To improve the effectiveness of infrared image dehazing, an improved dark channel was proposed. First, because the penetration ability of near-infrared light in the sky region is different from that in the non-sky region, and the infrared energy in the sky region is smaller than that in the non-sky region, the region was divided into sky and non-sky regions using the energy difference. Second, the atmospheric light value of the sky region was calculated using the average pixel brightness of the sliding window, the near-infrared wave attenuation was considered for the transmittance, and the atmospheric light value and transmittance of the non-sky region were calculated using the improved dark channel algorithm. Finally, the dehazed image was recovered from the atmospheric light value and transmittance of each region. The experimental results show that the dehazing output of the infrared image was clearer, the image detail information was better, and the evaluation index was better than those of other algorithms.