Nighttime hazy image usually has the non-uniform illumination, low contrast and serious color deviation. The existing dehazing methods are mainly proposed for daytime images, which don't fit well with the conditions of most nighttime hazy scenes. Nighttime image dehazing is more difficult. We explore the imaging characteristics under nighttime conditions and propose a new nighttime image dehazing method based on low-pass filtering and joint optimization of multi-feature. Firstly, in order to handle the non-uniform illumination of nighttime scenes, the image is filtered by the low-pass filtering. And then the minimum-maximum filtering is applied to the low frequency components to estimate the local atmospheric light. Secondly, for the current daytime dehazing algorithm prior is not suitable for nighttime image, an effective transmission estimation method is presented based on the joint optimization of multi-feature which combines contrast, saturation and information entropy. Finally, for the non-uniform color deviation exists in nighttime images, the non-overlapping blocking local Shade of Gray is proposed. Experimental results demonstrate that the proposed algorithm has a good subjective visual effect, and the objective evaluation indexes are superior to other algorithms in contrast and color deviation degree. The proposed algorithm can significantly remove haze, improve the contrast and recover more details with the natural color and better visual effect.