改进的基于暗原色先验的图像去雾算法  

Improved algorithm on image haze removal using dark channel prior

作　　者：蒋建国[1,2] 侯天峰[1] 齐美彬[1,2] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]安全关键工业测控技术教育部工程研究中心,安徽合肥230009

出　　处：《电路与系统学报》2011年第2期7-12,共6页Journal of Circuits and Systems

基　　金：特种显示技术教育部重点实验室开放课题基金项目(2008HGXJ0350);安徽省信息产业厅信息产业发展基金项目(2008012)

摘　　要：分析讨论了原暗原色先验去雾算法原理,指出其不足之处并推导出改进方法。通过引入一种容差机制,算法能更有效地处理不满足暗原色先验的明亮区域,纠正了这类区域错误估计的透射率,从而克服原算法在处理这些区域时产生的色彩失真。实验结果表明,这样的修改切实可行,恢复图像消除了色彩失真,视觉效果得以显著提高。Images captured in fog suffer from poor contrast and visibility.It is important to remove weather effects from the degraded image in order to make vision systems more reliable and more robust.An improved method is proposed to overcome the weakness of original algorithm.By including tolerance mechanism and adaptive efforts to fog,the improved algorithm can effectively deal with the sky,white object,and so bright area that do not meet the assumptions of dark channel,to overcome color distortion which is arise by using the original algorithm dealing with these areas.Experimental results show that such changes is feasible,to eliminate color distortion of the image,visibility can be significantly enhanced.

关 键 词：去雾 暗原色先验 色彩失真 容差 

分 类 号：TN911.73[电子电信—通信与信息系统]