计算机应用 ›› 2017, Vol. 37 ›› Issue (10): 2916-2920.DOI: 10.11772/j.issn.1001-9081.2017.10.2916

• 计算机视觉与虚拟现实 • 上一篇    下一篇

基于天空分割的单幅图像去雾算法

毛祥宇, 李为相, 丁雪梅   

  1. 南京工业大学 电气工程与控制科学学院, 南京 211800
  • 收稿日期:2017-04-17 修回日期:2017-06-16 出版日期:2017-10-10 发布日期:2017-10-16
  • 通讯作者: 李为相(1973-),男,河南光山人,副教授,博士,主要研究方向:智能决策、社交网络、图像处理,E-mail:lwxlf@njtech.edu.cn
  • 作者简介:毛祥宇(1992-),男,安徽阜阳人,硕士研究生,主要研究方向:图像处理、模式识别;李为相(1973-),男,河南光山人,副教授,博士,主要研究方向:智能决策、社交网络、图像处理;丁雪梅(1993-),女,江苏扬州人,硕士研究生,主要研究方向:模式识别、图像处理.
  • 基金资助:
    江苏省"六大人才高峰"项目(XXR-012)。

Single image dehazing algorithm based on sky segmentation

MAO Xiangyu, LI Weixiang, DING Xuemei   

  1. College of Electrical Engineering and Control Science, Nanjing Tech University, Nanjing Jiangsu 211800, China
  • Received:2017-04-17 Revised:2017-06-16 Online:2017-10-10 Published:2017-10-16
  • Supported by:
    This work is partially supported by the Six Talent Peaks Project in Jiangsu Province (XXR-012).

摘要: 针对暗通道先验算法在天空区域失效和复原图像色彩变暗的问题,提出一种基于天空分割的图像去雾算法。首先,采用基于边缘检测的分割算法将原始图像区分为天空区域和非天空区域;其次,在暗通道先验算法的基础上,改进对大气光和透射率的估计方法,进而对非天空区域采用改进的暗通道先验算法去雾;最后,利用基于成本函数的对比度增强去雾算法处理天空区域。实验结果表明,去雾后图像在方差、平均梯度、信息熵等指标上相对于暗通道先验算法均有较大提升,所提算法在保持较高运行效率的同时,能有效避免天空区域的Halo效应,还原真实的景物色彩。

关键词: 去雾, 大气散射模型, 天空分割, 暗通道先验, 对比度增强

Abstract: To address the problem that dark channel prior algorithm is invalid for sky region and the problem that the color of the restored image became darker, a single image dehazing algorithm based on sky segmentation was presented. Firstly, the segmentation algorithm based on edge detection was used to divide the original image into sky region and non-sky region. Then, based on the dark channel prior method, the estimation method for atmospheric light and transmittance was improved for the dehaze of non-sky region. Finally, the sky region was processed by an optimized contrast enhancement algorithm based on cost function. The experimental results demonstrate that, compared with dark channel prior algorithm, many technical specifications of restored images such as variance, average gradient and entropy are greatly improved. The proposed algorithm can effectively avoid the Halo effect in sky region and restore the true scene color while maintaining high operating efficiency.

Key words: haze removal, atmospheric scattering model, sky segmentation, dark channel prior, contrast enhancement

中图分类号: