基于颜色衰减先验的去雾方法研究

发布时间：2018-06-06 21:01

本文选题：颜色衰减先验 + 非线性模型　；参考：《南昌航空大学》2017年硕士论文

【摘要】：雾天环境下获得的图像和视频,存在清晰度差、对比度低、颜色失真等严重的退化现象,直接影响和限制了户外视觉系统的作用。因此改善雾天退化图像和视频的质量,降低雾霾等天气环境对户外视觉系统的影响,对有雾图像和视频进行有效的去雾处理,有着重要的现实意义。本文主要研究了图像去雾算法和视频去雾算法,具体研究内容如下:(1)针对基于颜色衰减先验去雾算法的不足,主要从场景深度估计和大气散射系数估计两个方面进行改进。首先详细阐述了关于场景深度估计的改进,通过建立关于场景深度的非线性模型,并利用训练样本学习得到模型的最佳系数,从而得到更加准确的场景深度。然后详细阐述了关于大气散射系数估计的改进,通过定义关于大气散射系数的函数,并利用大量图像实验得到函数的最佳参数,从而得到更加准确的大气散射系数。(2)针对现有的视频去雾方法存在去雾视频各帧在色彩和结构方面不协调,去雾处理时间长的问题,本文提出基于通用大气光值和通用透射率图的视频去雾方法。该方法利用监控视频的特性,将通用大气光值和通用透射率图应用于有雾视频的各帧图像中进行去雾。该方法首先对有雾视频提取背景图像,然后利用单幅图像去雾算法获取背景图像的大气光值和透射率图,最后将该大气光值和透射率图应用到有雾视频的每一帧的去雾处理中,从而实现视频去雾。(3)将本文图像去雾算法和暗原色先验理论去雾算法、颜色衰减先验理论去雾算法进行实验对比,从主观和客观两个角度分析本文算法的优越性。并从实验的角度说明本文提出的通用大气光值和通用透射率图的视频去雾方法的有效性。
[Abstract]:The images and videos obtained in fog environment have some serious degradation phenomena such as poor definition low contrast color distortion and so on which directly affect and limit the function of outdoor visual system. Therefore, it is of great practical significance to improve the quality of degraded image and video, reduce the influence of haze and other weather environment on outdoor visual system, and effectively remove fog from fog image and video. This paper mainly studies the image de-fogging algorithm and the video de-fogging algorithm. The detailed research contents are as follows: 1) aiming at the shortcomings of the prior de-fogging algorithm based on the color attenuation, we mainly improve the depth estimation of the scene and the estimation of the atmospheric scattering coefficient. Firstly, the improvement of depth estimation of scene is discussed in detail. The nonlinear model of depth of scene is established, and the best coefficient of the model is obtained by using training samples to get more accurate depth of scene. Then the improvement of the estimation of atmospheric scattering coefficient is described in detail. The function of atmospheric scattering coefficient is defined, and the optimum parameters of the function are obtained by a large number of image experiments. Thus, a more accurate atmospheric scattering coefficient. (2) aiming at the problem that the existing video de-fogging methods do not coordinate the color and structure of each frame of the de-fogging video, and the processing time of de-fogging is long. This paper presents a video fog removal method based on the general atmospheric light value and the general transmittance map. Based on the characteristics of surveillance video, the general atmospheric light value and the general transmittance map are applied to each frame image of fogged video to remove fog. Firstly, the background image is extracted from the fogged video, then the atmospheric light value and the transmittance map of the background image are obtained by using a single image de-fog algorithm. Finally, the atmospheric light value and transmittance map are applied to the de-fogging process of each frame of the fogged video. Thus, the image de-fogging algorithm and dark priori theory de-fogging algorithm and the color attenuation priori theory de-fogging algorithm are compared experimentally. The advantages of this algorithm are analyzed from the subjective and objective aspects. The effectiveness of the proposed video de-fogging method for the general atmospheric light value and the general transmittance map is illustrated from an experimental point of view.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】