摘要
针对现有图像去雾方法处理效率低,天空部分处理效果欠佳以及去雾图像视觉效果不理想等问题,提出了一种快速大气光幂去雾算法。提出的算法是对全球环境光值和大气光幂值求取方法的改进。首先,采用高斯低通滤波求取雾气图像低频区,应用循环四分图算法在低频区得到全球环境光值A;其次,采用暗原色优先算法获取初始大气光幂值,结合自适应各向异性高斯滤波处理大气光幂;最后,采用色调调整增强图像细节,使图像逼近于无雾场景。实验结果表明,本文算法能消减图像景深突变处的晕轮效应,亮度、对比度和细节信息处理效果较好,不仅较完整地保留了边缘细节,而且显著提升了处理效率,同时具有较高的鲁棒性和实时性。
A fast atmosphere veil de-hazing method was proposed to overcome the shortcomings of ex- isting de-hazing methods in lower process efficiency, poorer treatment result of sky part and a bad vis- ual effect for images. The method focuses on the improvements for global atmospheric light values and the acquiring method of atmospheric veil values. Firstly, the Gaussian low-pass filter method was used to gain a low frequency area of an image, and the circle quarter figure algorithm was used to ob- tain the global atmospheric light value of the low frequency area. Then, the dark channel prior algo- rithm was employed to achieve the initial atmospheric veil value and the atmospheric veil was pro- cessed by the adaptive anisotropie Gaussian filter. Finally, image details were enhanced by utilizing tone mapping, by which the image could approximate to the no fog scene image. The experimental re- sults show that the proposed method restricts the halo effect of depth mutation area in the image and well processes lightness, contrast and detail information for the image. It keeps edge details of the im- age perfectly ,improves the image processing efficiency and has good robustness and real-time ability.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2016年第8期2018-2026,共9页
Optics and Precision Engineering
基金
四川省教育厅重点项目(No.15ZA0118)
特殊环境机器人技术四川省重点实验室开放基金资助项目(No.13zxtk0505)
西南科技大学博士基金资助项目(No.13zx7112)
关键词
大气光幂雾图像
图像复原
低通滤波
循环四分图
色调调整
晕轮效应
atmosphere veil haze image
image de-hazing
low-pass filter
circle quarter figure algo- rithm
tone mapping
halo effect
