CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleResearch on restoration of fog-degraded images
Thesis Advisor彭思龙
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword浓雾 雾霭 对比度提升 多次散射 空间变化的模糊 图像复原 Fog Haze Contrast Enhancement Multiple Scattering Spatial Variant Deblurring Image Restoration
Abstract雾天拍摄的户外图像,由于大气粒子的散射作用,图像对比度低,颜色失真,场景内容模糊,能见度变差。雾天降质图像复原技术已经成为计算机视觉和图像处理领域的研究热点,并应用于交通监控系统,地形测量系统,智能车辆,室外目标检测和识别,水下、航空和卫星成像等领域。 本文深入分析雾天图像成像过程,结合多次散射引起的图像模糊,提出新的图像退化模型,实现雾天降质图像的复原并去除模糊。本文主要做了以下工作: (1) 分析大气成像机理,研究不同雾浓度的天气情况下,不同的大气散射性质。在浓度天气和长距离成像过程中,多次散射造成的图像模糊不能忽略。在分析单次散射和多次散射机理的基础上,本文提出新的有雾图像退化模型,有效的刻画各种雾浓度的成像过程。 (2) 基于新的退化模型,有雾图像的复原过程变成分离大气光和去模糊两个过程。在分离大气光过程中主要考虑单次散射的影响,建立MAP(最大后验概率)框架,引入自然图像统计的稀疏先验和软抠像约束,采用迭代重加权最小二乘过程估计透射率图。在去模糊过程,由于大气粒子的多次散射依赖于场景深度,模糊核建模为依赖于场景深度的大气点扩散函数然后利用MAP方法复原得到清晰的无雾图像。 结果表明,本文提出的模型和算法能够有效的提高图像能见度,消除图像模糊,取得满意的复原效果。
Other AbstractImages from outdoor foggy scenes are degraded by atmosphere mediums. Due to scattering of atmospheric particles, degraded images lost contrasts and color fidelity, and appear poor visibility and blurred scene contents. Image dehazing has been concerned in computer and image processing applications, such as traffic surveillance and control system, topographic survey system, intelligent vehicles, outdoor objects detection and recognition. Based on the deeply analysis of the images formation process in the foggy weather, this paper proposes a new image degradation model incorporating burring in images caused by multiple scattering and implements images restoration with blurring removed. The main work includes: (1) Analyzed the image formation process, atmospheric scattering has different properties in various weather conditions. In dense fog and long distant imaging, blurring in images caused by multiple scattering could not be ignored. Based the mechanisms of single scattering and multiple scattering, a new foggy image degradation model has been proposed to interpret the image formation process in foggy weather. (2) Base on the model, the image restoration problem has been separated in two parts: removing airlight and deblurring. Single scattering has main influence on images in the process of removing airlight, then applying the MAP (maximum a posteriori) with the natural statistics prior and a soft matting regulation, the estimation of the transmission map could be generated through the iterative re-weighted least squares process. In the process of deblurring, since multiple scattering of atmospheric particles depends on the depth map, the blur kernel is modeled as the atmospheric point spread function, and then haze-free images are recovered using the MAP method. Results demonstrate that the new degradation model and algorithm could enhance visibility and remove blurring in images effectively, and achieve satisfying restoration effects For the designed embedded vision positioning system, many experiments are conducted, such as color block identification and segmentation, feature extraction, objects positioning, object approaching and obstacle avoidance for a mobile robot. The experimental results verify the effectiveness of the proposed embedded vision positioning system.
Other Identifier200828014628045
Document Type学位论文
Recommended Citation
GB/T 7714
李霞. 雾天降质图像的复原技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2011.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20082801462804(9163KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李霞]'s Articles
Baidu academic
Similar articles in Baidu academic
[李霞]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李霞]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.