图像去雾技术研究与实现

发布时间：2018-01-18 07:20

本文关键词：图像去雾技术研究与实现　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机软硬件的快速发展,计算机视觉系统已广泛应用于智能交通、视频监控、目标跟踪等领域。然而,在雾天环境下,由于受到大气中悬浮颗粒的散射,摄像设备捕获到的图像严重退化,能见度低,直接限制和影响了计算机视觉系统的正常运行。因此,为了保证视觉系统的正常工作,改善雾天图像的视觉效果,对雾天图像进行去雾处理,有着非常重要的现实意义。本文在研究雾天图像降质物理成因的基础上,提出了新型的图像去雾方法,主要工作如下:(1)针对当前图像去雾方法没有考虑不同雾天环境对去雾效果的影响问题,提出了基于物理成因的自适应单图像去雾方法。对于雾天物理模型中的重要变量透射率,使用散射系数和场景深度直接求取。在求解散射系数时,将雾粒子直径作为一个重要参数进行处理,利用雾天图像的暗通道图来自动识别雾浓度,并确定雾粒子直径的值。而场景深度使用最大化后验概率的方法进行估计,从而使得不同雾天环境下的降质图像都可以获得良好的去雾效果。(2)针对复杂深度的雾天图像使用最大化后验概率方法估计的深度图像不准确的问题,提出了基于双目视觉立体匹配的去雾方法。利用立体匹配技术估计出更为精确的深度图,同时使用提出的快速导向滤波方法提高运行速度,获得了更为良好的去雾效果。(3)基于提出的去雾方法,设计并实现了一个图像去雾系统。用户可以直观地看到去雾效果和一些评价指标数据,并进行参数的调整。
[Abstract]:With the rapid development of computer software and hardware, computer vision system has been widely used in intelligent transportation, video surveillance, target tracking and other fields. The image captured by the camera is seriously degraded, and the visibility is low, which directly limits and affects the normal operation of the computer vision system. Therefore, in order to ensure the normal operation of the vision system, improve the visual effect of fog image. It has very important practical significance to defog the fog image. Based on the study of the physical cause of the fog image degradation, a new image de-fog method is proposed in this paper. The main work is as follows: 1) in view of the current image de-fogging methods, the influence of different fog environments on the de-fogging effect is not considered. An adaptive single image de-fogging method based on physical origin is proposed. For the transmittance of important variables in the fog physical model, the scattering coefficient and scene depth are directly obtained. The diameter of fog particles is treated as an important parameter, and the fog concentration is automatically identified by the dark channel map of fog image. The value of the particle diameter is determined, and the depth of the scene is estimated using the method of maximizing the posterior probability. So that the degraded images in different fog environments can get good de-fogging effect. (2) aiming at the problem of inaccuracy of the depth images estimated by the maximum posterior probability method for the fog images with complex depth. A new de-fog method based on binocular stereo matching is proposed. The stereo matching technique is used to estimate more accurate depth maps and the proposed fast guided filtering method is used to improve the running speed. An image de-fogging system is designed and implemented based on the proposed de-fogging method. Users can directly see the de-fogging effect and some evaluation index data. The parameters are adjusted.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】