摘要
针对雾霾图像中含有高亮、大面积浓雾,天空区域无法清晰识别,求取出的图像偏暗、色彩失真的问题,提出了一种基于暗通道先验的自适应阈值分割和透射率补偿的去雾改进算法,通过OSTU阈值分割将图像分为前景区域和背景区域,求取亮暗通道,并通过统计函数计算出前景区域和背景区域所占的像素比例,加权求取大气光值。通过透射率补偿参数K的引进,使得求取的透射率更接近真实值,最后通过CLEAR法进行色度调整。实验结果表明,该法去雾后的图像细节信息保留完整,失真度减小,视觉上更加真实自然,信息熵平均提高7.03%,SSIM平均提高5.56%,MSE平均减小9.19%。
In view of the problem that the haze image contains high brightness,large area thick fog,and the sky region can not be clearly identified,the resulting image is dark and the color is distorted,in this paper,an improved algorithm based on dark-channel prior adaptive threshold segmentation and transmission compensation is proposed.The image is divided into foreground region and background region by OSTU threshold segmentation,the ratio of foreground and background pixels is calculated by statistical function,and the atmospheric luminous value is calculated by weighting.By introducing the transmission compensation parameter K,the transmissivity obtained is closer to the real value.CLEAR method is used to adjust the chromaticity.The experimental results show that the detail information of the defogged image is retained intact,the distortion is reduced,and the vision is more real and natural.The information entropy,SSIM and MSE are increased by 7.03%,5.56%and 9.19%respectively.
作者
孙乐乐
席一帆
吕悦
SUN Lele;XI Yifan;LV Yue(School of Information Engineering,Chang’an University,Xi’an 710064,China)
出处
《电子设计工程》
2021年第4期48-52,共5页
Electronic Design Engineering
基金
国家自然科学基金面上项目(51878066)。
关键词
去雾
大气散射模型
透射率补偿
阈值分割
CLEAR法
defogging
atmospheric scattering model
transmittance compensation
threshold segmentation
CLEAR method
