摘要
传统基于暗通道图像去雾中仅考虑初始透射图对于透射图优化的影响,而未充分考虑深度信息对于透射图优化贡献不同的特点,造成去雾结果在深度不连续处出现"晕"且离视点较远区域的去雾效果有明显下降.针对上述问题,本文提出一种基于边缘特征的加权暗通道先验去雾算法.该方法根据边缘特征的位置估计深度信息的连续性,将边缘点及非边缘点赋予不同权值,对加权透射图优化求解.仿真实验表明,新的去雾算法在恢复图像细节的基础上能够有效抑制"晕"的产生,证实了本文方法的可行性和有效性.
The classical method based on dark channel prior only considers the estimated transmission map and does not take into account the depth information to optimize the soft matting function,which makes the result contain few halo artifacts in depth discontinuity and haze-removal effect in the region far from the viewpoint is degraded.For this issue,a simple but efficient image dehazing method is proposed with weighted dark channel prior based on edge feature.Using this method,depth continuity can be estimated by position of edge points,and refined transmission map is obtained by different weight of edge points and non-edge points.Experimental results show that detail information is well preserved and no halo artifacts in haze-free image,and demonstrates the feasibility and effectiveness of the proposed algorithm.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2012年第3期320-325,共6页
Acta Photonica Sinica
基金
国家自然科学基金(No.61071172
No.60602056
No.60634030)
航空科学基金(No.20105153022)
西北工业大学基础研究基金(No.JC200941)资助
关键词
图像去雾
暗通道先验
边缘检测
深度估计
Haze removal
Dark channel prior
Edge detection
Depth estimation
