基于超像素分割的图像去雾算法  

Single Image Dehazing Based on Superpixel Segmentation

作　　者：郑良缘 王平[1] 高颖慧[1] 

机构地区：[1]国防科技大学电子科学与工程学院ATR重点实验室,长沙410073

出　　处：《重庆理工大学学报（自然科学）》2015年第6期100-106,共7页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(61103082)

摘　　要：针对基于暗原色先验去雾算法中出现的光晕效应以及透射率计算的复杂性,提出一种基于超像素分割的图像去雾算法。首先根据超像素分割的原理将图像分割成若干区域;然后利用区域内的暗原色值来估算区域的透射率,从而得到图像的透射率图;最后采用引导滤波代替软抠图算法来提高透射率计算速度,并根据大气散射模型恢复雾天图像。实验结果表明:该算法在提高图像去雾运算速度的同时能有效抑制光晕效应,使图像更加清晰,且能较好地保持边缘细节。According to the halo artifacts and the long time computation when using the dark channel prior image dehazing algorithm,we proposed a new dehazing method from a single image using a segmentation algorithm based on superpixel. Firstly,we divided an image into many small parts. Secondly,we got the more accurate transmission map in every small region based on the dark channel prior algorithm. Thirdly,we used the guiding filter in place of the soft matting algorithm in dark channel prior. Lastly,we recovered the haze image based on the atmosphere attenuation model. Experiments show that the presented method processes the haze image fast and removes the halo artifacts effectively,and it not only enhances the contrast,but also keeps the edges sharp.

关 键 词：大气散射模型 暗原色先验 超像素 图像分割 去雾 

分 类 号：TN911.73[电子电信—通信与信息系统]