【作者】 刘峰；

【导师】 刘晓玉；

【作者基本信息】 武汉科技大学 ， 控制科学与工程， 2016， 硕士

【摘要】 在雾霾天气条件下,户外计算机视觉系统采集的图像会出现对比度下降,色彩失真等退化现象。近年来雾霾天气频繁出现,这对户外计算机视觉系统的正常使用造成了很大影响,因此有必要针对雾霾天采集的图像进行去雾研究。本文主要从图像增强和图像复原两方面对图像去雾进行研究,并以“保证一定去雾效果的同时尽量提升算法速度”为指导思想,针对基于图像增强的Retinex算法和基于图像复原的暗原色先验算法进行研究,并将去雾算法应用于实际场合中,验证算法去雾的有效性。本文的主要工作如下:(1)在图像增强方面,主要分析了直方图均衡化算法和Retinex算法对雾化图像的去雾效果。针对Retinex算法对雾化图像增强时会出现色彩失真的不足,本文采用HSV颜色空间而非传统的RGB颜色空间对雾化图像进行多尺度Retinex增强,利用HSV颜色空间明度分量V与色调分量H之间存在正交性这一特性,在对明度分量V进行增强时,可保持增强前后的图像色调H不变,以保持图像的色彩。实验结果表明,改进方法增强后的图像色彩保真效果较好,同时由于在HSV颜色空间只需对明度分量进行增强处理,算法速度也得到了提高。(2)在图像复原方面,重点研究了暗原色先验去雾算法。原算法为了消除去雾后的图像中产生的光晕现象,采用软抠图的方法修复透射率图,有效解决了这一缺陷,但同时导致算法复杂度高,运算速度慢。本文采用自适应双边滤波取代原始算法中的软抠图对透射率图进行平滑处理,并使用下采样的方法进一步提速。实验表明,改进算法能有效消除光晕现象,算法速度也得到很大程度提高。(3)在理论研究的基础上,将去雾算法应用于道路交通监控场合。根据道路交通监控视频的特点,在不同的雾天等级下使用合适的去雾算法对监控视频序列图像进行去雾,使系统能针对不同的雾天视频有针对性地进行去雾。更多还原

【Abstract】 In foggy conditions,the images collected by the outdoor computer vision system will be greatly degraded,such as contrast decline,color distortion.In recent years,the fog weather frequently appears,this result in a significant negative impact on the use of outdoor computer vision system,so it’s of great value to have a study on image defogging.Based on fog image enhancement and fog image restoration,paper has a study on image defogging technology,establishes the guidelines that “ensure the defogging effect and enhance the speed as far as possible ”,makes some improvements for the related algorithms.Applying the image defogging algorithm in practical application to verifies it’s effectiveness.The main work of the paper is as follows:(1)In terms of image enhancement,paper analyzes the defogging effect of histogram equalization algorithm and Retinex algorithm.Enhanced by Retinex algorithm,the foggy image will appear color distortion,aiming at this defect,paper use multi-scale Retinex which based on HSV color space but not RGB color space to enhance the foggy image.In HSV color space,lightness component V and hue component H are uncorrelated,based on this characteristics,when enhancing the component V,component H remains unchanged.Experimental results show that the proposed algorithm have a better enhance result on colorful image,at the same time,as only to enhance component V,the proposed algorithm is faster.(2)In respect of image restoration,the paper is focusing on the studying of the image defogging algorithm based on dark channel prior.In order to eliminate the halo effect,original algorithm use the soft matting method to repair the transmittance map,this lead to the complexity of the algorithm highly,and computing speed slow.In this paper,smoothing the transmittance map is processed by adaptive bilateral filtering algorithm which replace the original soft matting,and use the down-sampling method to make the speed faster.Simulation results validate that the proposed algorithm can effectively eliminate the halo effect and the speed of the algorithm has been greatly improved.(3)Based on the theoretic research,the defogging algorithm is applied to the situation of road traffic monitoring.Based on the features of traffic video monitoring,using a suitable defog algorithm in different foggy weather grade so that it enables the system to have a well-targeted defogger in different foggy weather conditions.更多还原