In classic dark channel prior image defogging algorithm, the inaccurate calculation of atmospheric light value is apt to cause color deviation, and the deviation of dark channel value in image edge position will lead to halo effect in the restored image. In this paper, by analyzing the gray values’ mean and variance characteristics of sky regions in foggy images, a novel atmospheric light value estimation algorithm based on dichotomy search is adopted. It can obtain more accurate atmospheric light value quickly. An improved method for calculating dark channel values is presented in this paper. First, according to the gray difference between the minimum image and the dark channel image, the adaptive threshold is obtained. Next, the edge position of the image with drastic change in depth of field is determined and the gray value is corrected. The subjective and objective evaluation of the method’ implementation effects are done and the results show that it can effectively remove the halo effect , correct the color deviation, improve the contrast, and retain more context details in the restored image.