摘要
雾、霾天气是影响图像和视频质量下降的重要因素,给室外视频任务带来很大不便。考虑到远景视频中地平线的存在,提出一种改进的基于地平线检测的去雾方法。该方法基于暗原色先验理论,用改进的地平线检测算法,从整幅图像分割出天空区域,得到雾天图像退化物理模型中的大气光部分,再引入容差机制,提高图像去雾质量。而在对传输图的修正过程中采用图像引导滤波代替既占内存又耗时的软抠图方法。对远景雾天图像的去雾实验表明,该方法改进了原有基于暗原色先验单幅图像去雾方法中白色场景和物体的存在易导致算法无效的限制,有效减小了日周光光晕现象对图像可视化质量的影响,同时提高了算法速度。
Fog and haze are the important factors that degrade the quality of image and video , which bring great inconvenience to outdoor video tasks .Considering the existence of horizon in vision video , an improved defogging method based on horizon detection is proposed .It detects the horizon with the theory of dark channel and then segments the sky region from the whole image to get the light part of the haze imaging model .Parameters are adjusted by introducing fault-tolerance to improve image quality .At the same time, we use image-guided filtering instead of soft matting which consumes both memory and time in transmission correction.As shown in the experiment , this method improves the vulnerable behavior of origin algorithm in the presence of white scene or objects ,the image quality debased by neglecting the obvious halo of diurnal light in distant view image , and the processing time.
出处
《计算机与现代化》
2014年第9期67-71,共5页
Computer and Modernization
基金
国家863计划项目(2013AA013802)
关键词
远景视频
地平线检测
容差
引导滤波
去雾
vision video
horizon detection
fault-tolerance
guided filtering
haze removal