Abstract: The scattering of particles(e.g. haze particles and fog droplets) in the atmosphere degrades images of outdoor scenes and further harms image quality. Image dehazing can significantly gain contrast, increase scene visibility, correct color shift introduced by the airlight and improve visual effect. Nonetheless, image dehazing is prone to noise amplification. Noise of the dehazed image is obvious in regions with low transmission, especially for distant sky region. A noise inhibition method during image dehazing was proposed to solve this problem. The hazy image was smoothed by the bilateral filter with varying radius which was adapted to the transmission map. The new transmission map was calculated again corresponding to the smoothed hazy image, and then used to estimate the denoised recovery image. Combining the noise estimation method for single color image, experimental results demonstrate our method's performance with noise inhibition.