Journal of Computer Applications ›› 2017, Vol. 37 ›› Issue (1): 268-272.DOI: 10.11772/j.issn.1001-9081.2017.01.0268

Previous Articles     Next Articles

Single image in-depth dehazing algorithm based on optimization of guided image

DONG Yufei, YANG Yan, CAO Biting   

  1. School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou Gansu 730070, China
  • Received:2016-05-10 Revised:2016-07-19 Online:2017-01-10 Published:2017-01-09
  • Supported by:
    This work is partially supported by National Natural Science Foundation of China (61561030), the Fundamental Research Funds for Gansu Provincial Finance Department (214138), the Teaching Reform Project of Lanzhou Jiaotong University (160012)

基于导向图优化的单幅图像深度去雾算法

董宇飞, 杨燕, 曹碧婷   

  1. 兰州交通大学 电子与信息工程学院, 兰州 730070
  • 通讯作者: 杨燕
  • 作者简介:董宇飞(1989-),男,山东滨州人,硕士研究生,主要研究方向:数字图像处理;杨燕(1972-),女,河南临颍人,副教授,博士,主要研究方向:数字图像处理、智能信息处理、语音信号处理;曹碧婷(1990-),女,甘肃白银人,硕士研究生,主要研究方向:数字图像处理。
  • 基金资助:
    国家自然科学基金资助项目(61561030);甘肃省财政厅基本科研业务费资助项目(214138);兰州交通大学教改项目(160012)。

Abstract: Aiming at the quality loss problems such as degradation in contrast and color distortion of image captured in haze and fog weather conditions, a single image in-depth dehazing algorithm based on optimization of the guided image was proposed. The local mean and standard deviation of the image were adopted to optimize the guided image on the basis of analysing the character of atmospheric veil. Then, the guided image was further filtered by using the dual zone filtering to get smooth and sharp-edged guided image. The atmospheric veil was estimated through the fast guided filtering. At last, a clear image would be recovered based on the atmospheric scattering physical model. The experimental results show that the recovered image is clear and natural, and rich in details. Its close view is dehazed completely, while the dehazing of its distant view is improved greatly. The proposed algorithm achieves good results where the depth of the field has a sudden saltation and improved the visibility and robustness of outdoor vision system.

Key words: in-depth dahazing, dark channel prior, dual zone filtering, guided filtering, guided image

摘要: 针对雾霾等天气条件下获取的图像出现对比度下降、颜色失真等降质现象,提出一种基于导向图优化的单幅图像深度去雾算法。该算法在对大气散耗函数特性进行分析的基础上,引入图像局部均值和标准差优化导向图;再进一步对导向图进行分区域滤波,得到平滑且边缘清晰的导向图;然后采用快速引导滤波估计大气散耗图;最后根据大气散射物理模型恢复清晰图像。实验结果表明,恢复的图像清晰自然,细节丰富,近景去雾彻底,远景去雾有很大提升,在景深突变处的边缘取得较好的效果,提高了户外视觉系统的视见度和鲁棒性。

关键词: 深度去雾, 暗通道先验, 双区域滤波, 引导滤波, 导向图

CLC Number: