基于无参考图像获取能见度的车载视频自适应去雾算法

为解决雾天环境下车辆环境感知困难的问题，针对车行视程去雾算法无法自动获取能见度、复原图像色彩过饱和、细节丢失严重等情况，克服传统器测法和目测法对能见度获取的非实时性及主观性，提出了一种改进的非线性二分求根算法，利用无参考图像空域质量评价指标（BRISQUE）对能见度进行实时修正，最终实现了能见度的自动估值。作者改进了大气能见度与车行可视距离的关系函数，由改进后的车行可视距离求出的透射率值与实际透射率相比误差减小，降低Halo效应的产生、增加了图像细节信息。实验表明，利用能见度求出的透射率估值在大气光散射模型下能够自适应的处理雾霾视频，复原出的视频图像画质清晰，色彩鲜艳亮丽不失真且能保留大量的图像信息，处理过程视频流畅无卡顿，对于在雾天环境中交通场景不断变化的车载视频也有良好的去雾效果。

Self-adaptive Defogging Algorithm for Vehicle Video with Automatic Estimation of Visibility

In order to solve the problem of vehicle environment perception in foggy environment, aiming at the situation that the algorithm of vehicle visual distance defogging can not automatically acquire visibility, restored image color oversaturation, and loss of details seriously, etc, an improved non-linear dichotomy root-finding algorithm, which utilizes the spatial quality of no-BRISQUE revises the visibility in real time, and finally realizes the automatic evaluation of visibility, is proposed to overcome the non-real-time and subjectivity of visibility acquisition by traditional instrument measurement and visual measurement methods. An improved non-linear dichotomy root-finding algorithm is proposed,. The relationship function between atmospheric visibility and vehicle visual distance is improved. The error between the transmittance value calculated from the improved vehicle visual distance and the actual transmittance is reduced, the Halo effect is reduced, and the image details are increased. Experiments show that the transmittance estimation based on visibility can adaptively process haze video under atmospheric light scattering model. The reconstructed video image has clear picture quality, bright color and no distortion, and can retain a large amount of image information. The processing video is smooth and no jamming. It also has good fog removal effect for vehicle video with changing traffic scenes in foggy environment.