电子学报 ›› 2015, Vol. 43 ›› Issue (3): 500-504.DOI: 10.3969/j.issn.0372-2112.2015.03.013

• 学术论文 • 上一篇    下一篇

基于景深约束的单幅雾天图像去雾算法

南栋, 毕笃彦, 马时平, 何林远, 娄小龙   

  1. 空军工程大学航空航天工程学院, 陕西西安 710038
  • 收稿日期:2013-07-18 修回日期:2014-03-13 出版日期:2015-03-25
    • 作者简介:
    • 南 栋 男,1987年10月出生,陕西西安人,博士生,研究方向为图像处理与计算机视觉. E-mail:nd.tian_53@163.com;毕笃彦 男,1962年5月出生,陕西扶风人,教授,博士生导师,研究方向为图像处理与模式识别. E-mail:biduyan@163.com
    • 基金资助:
    • 国家自然科学基金 (No.61175029,No.61203268,No.61202339,No.61372167,No.61379104); 陕西省自然科学基金 (No.2012JQ8034); 博士后特别资助 (No.2012T50879); 博士后基金面上资助 (No.2012M512144)

Single Image Dehazing Method Based on Scene Depth Constraint

NAN Dong, BI Du-yan, MA Shi-ping, HE Lin-yuan, LOU Xiao-long   

  1. Institute of Aeronautics and Astronautics, Air Force Engineering University, Xi'an, Shaanxi 710038, China
  • Received:2013-07-18 Revised:2014-03-13 Online:2015-03-25 Published:2015-03-25
    • Supported by:
    • National Natural Science Foundation of China (No.61175029, No.61203268, No.61202339, No.61372167, No.61379104); Natural Science Foundation of Shaanxi Province,  China (No.2012JQ8034); Postdoctoral Science Foundation funded project (No.2012T50879); Major Project of Doctoral Foundation (No.2012M512144)

摘要:

本文提出一种基于景深约束的单幅雾天图像去雾算法,该算法首先对退化模型进行变换以满足Kimmel变分框架的要求;其次,考虑到人眼视网膜锥细胞对绿光的敏感性,将绿光分量作为大气传输图变分求解模型的输入;最后,利用景深图像特性,在8邻域快速求解中对能量函数进行约束,从而有效提升去雾图像的视觉效果.实验结果表明本文算法在获得最佳去雾效果的同时,具有较好的实用性和较少的计算资源消耗.

关键词: 图像去雾, 退化模型, 景深, 变分模型

Abstract:

We propose a single image dehazing method based on scene depth constraint.Firstly, the initial degradation model is transformed to meet the Kimmel's variational model.Then, considering the sensitivity of the cone in human's retina to the green light, we use it as an input of the variational model of the atmospheric transmission map.Finally, in the 8-neighborhood fast solving we constrain the energy function by scene depth, so as to improve the visual effect of the result efficiently.Experiments demonstrate that the proposed method can effectively remove fog, provide good practicability and take less memory consumption.

Key words: image dehazing, degradation model, scene depth, variational model

中图分类号: