Under a haze and mist weather condition, the scattering and absorption due to atmospheric particles tremendous influence on contrast and visibility of the surveillance images. In this paper, the image fog removal principle using atmospheric veil is discussed, for the bilateral filter leads to gradient reversal artifacts while atmospheric veil estimate is refined using a bilateral filtering, an improved single image dehazing method using atmospheric veil is proposed. Firstly, the atmospheric scattering model was simplified. Then, the estimated atmospheric veil is refined using an anisotropic diffusion filtering to overcome the weakness of the bilateral filter. Finally, the tolerance mechanism is used in algorithm, the error estimates of atmospheric veil in sky region is corrected, so as to overcome the color distortion in these region. Experimental results show that improved algorithm can efficiently improve the visual effects of dehazed image.