基于车行视程与大气光快速估值的车载视频去雾算法
Video Dehazing Algorithm Based on the Visual Range of the Vehicle Driving and the Airglow Rapid Estimation
作者：李炎炎(四川大学 制造科学与工程学院, 四川 成都 610065)；龙伟(四川大学 制造科学与工程学院, 四川 成都 610065)；覃宏超(四川大学 制造科学与工程学院, 四川 成都 610065)；付继贤(四川大学 制造科学与工程学院, 四川 成都 610065)
Author：LI Yanyan(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)；LONG Wei(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)；QIN Hongchao(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)；FU Jixian(School of Manufacturing Sci. and Eng., Sichuan Univ., Chengdu 610065, China)
收稿日期：2016-12-26 年卷（期）页码：2017,49(3):217-222
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：车行视程;大气光估值;去雾算法;引导滤波;盒式滤波
Key words：visual distance on vehicle;airglow rapid estimation;defogging algorithm;guided filter;box filter
基金项目：四川省科技支撑计划资助项目（2014Z0007；2010GZ0171）
中文摘要
交通场景中的视频图像去雾处理，是一个实时性极强的不确定反问题，针对雾霾天气下车载视频图像退化严重的现象，分析了交通环境中雾气浓度对车前物景可视度的关系，提出了大气能见度与车行视觉距离之间的关系模型，讨论了降质图像增强的理论与方法，建立了雾霾条件下车行可视距离的线性回归公式和基于大气能见度的透射率快速估值模型。同时，研究了雾霾物像暗通道的基本特征，提出了利用引导滤波对图像实现边缘平滑、细节增强以及利用盒式滤波保持边缘信息，降低时间复杂度；直接利用灰度图像获取大气光像素矩阵的估值方法，建立了基于大气光的快速估计模型，解决了暗原色先验理论方法的“非天空区”假设及其时间复杂度难以适应车载视频图像处理的问题。根据上述提出的透射率估计方法和天空光的估值模型，本文提出了一种鲁棒性好，实时性强的雾霾视频图像去雾的新算法，并完成了基于该方法的视频图像恢复处理流程设计，构造了车载雾霾视频图像恢复处理的综合验证平台，通过信息熵和图像边缘检测的方法，对本文提出的方法与目前已有的几种流行的去雾方法进行了图像恢复质量的比较，结果表明，本文提出的方法在反映图像细节和清晰化等方面都取得了良好的处理效果。
英文摘要
Aiming at the problem of serious degradation of image in the weather of fog and haze, this paper analyzed the relationship between the atmospheric visibility and the visual distance of the vehicle driving, and established the linear regression formula of the visual distance of the vehicle driving under the hazy weather and the transmittivity rapid estimation model based on the atmospheric visibility, according to the severe degradation issue of the vehicle-mounted video image in the hazy weather; at the same time, according to the dark channel features of the objective image, a fast estimation model of atmospheric light is constructed. It built the rapid estimation model of the airglow by using the guided filter and the box filter and solved the problem of method of the dark channel priori theory hypothesis of the "the sky area" and the time complexity is difficult to adapt to the onboard video image processing problems. It proposed a new hazy weather video image dehazing algorithm with good robustness and strong instantaneity. Then designed the video image restoration processing comprehensive verification platform, which conducted the evaluation comparison on the image recovery quality of the theoretical method presented by this article and the current typical dehazing method through the comentropy and image edge detection, gained good processing effects.
【关闭】
