The fog image is divided into non-sky area and sky area. For the non-sky area, a method of optimization of dark channel prior thought, namely the opening dark channel algorithm, is presented. For the sky area, an improved boundary constraint algorithm is put forward by introducing bilateral filter algorithm which has the role of edge-preserving and denoising. The two different algorithms are utilized to estimate the more accurate transmission map in their own areas respectively. Afterwards, the two defog area can be obtained by exploiting the atmospheric physical scattering model. Finally, the haze image is restored by combining the two defog areas. The experimental results show that the proposed algorithm has greatly improve the image contrast and ameliorate the color distortion problem in haze images, especially for haze images which contain the sky area.