【作者】 祝培；

【导师】 朱虹；

【作者基本信息】 西安理工大学 ， 模式识别与智能系统， 2004， 硕士

【摘要】 随着视频监测技术在各个领域越来越广泛的应用，要求系统的适用性也随着提高。为了使监测系统在恶劣的天气环境下仍然能够正常工作，提出了恶劣天气环境下的清晰化技术，以通过软件手段来达到强化微弱信息的目的。 本文主要针对雾天环境下图像的清晰化技术进行了研究。论文中首先对雾天的分类给出了介绍，并且分析了雾天下的成像模型的特点，以及由此带来的图像清晰化处理的困难。为了解决这些问题，论文分为两大类方法对图像进行处理，一类是基于模型的图像复原技术，一类是非建模的图像增强技术。由于轻雾天气和大雾天气下图像信息损失程度的不同，对它们分别采用不同的手段进行处理。采用图像自适分割方法以及图像局部增强方法对轻雾天气下的图像进行处理，在大幅度提高图像清晰度的同时避免了天空噪声的放大。对单深度图像的处理，则采用人机交互方式选取辐射黑点，后根据大气模型进行处理的方法，求取图像的光学深度，进而利用模型处理的方法恢复图像的信息。对变雾况的两幅同一场景的图像，采用差图像放大的方法，提高图像的清晰度并给出了理论证明。大雾天气下的图像由于其本身灰度范围有限的特点，采用动态范围拉伸的方法进行全局增强，并对其增强后出现的噪声干扰问题利用稀疏编码技术给予解决。 本论文的研究主要利用了计算机技术和图像处理技术来提高雾天下图像的西安理工大学硕士学位论文清晰度，对于提高户外全天候监测系统运行的鲁棒性的提高有着重大的意义，并且对其他恶劣犬气下图像的清晰化处理也有一定的促进意义。更多还原

【Abstract】 With the development of the video surveillance system techniques, the demands of system’s reliability and stability have raised. In order to make video surveillance systems operate normally in bad weather, image clearness technique is proposed in the paper which aims to enhance weak information by software methods.The paper studies on the image clearness technique in fog. Firstly, the classification of fog is introduced. Then the characteristic of atmospheric scattering models and the difficulty caused by these models are analyzed. To solve these problems, the paper presents two classes of ways: One based on image restoration technique and the other based on image enhancement technique. Due to the difference degree of information lost in haze and dense fog, different methods are adopted. Adaptive segmentation and local enhancement methods are used to process images taken in haze. As the result, higher contrast of image is obtained without noise magnification in sky region. Single depth image can easily be processed through models. Fist, black pixel whose radiance is zero is chosen mutually, then image optical depth is computed, at last image information is recovered, two images which* This Work is supported by the Major Project of Science and Technology Research of Ministry of Education of China under Grant No: 01113are taken in different degree of fog are processed by difference image information amplification. Results are satisfactory, besides theoretical proof also has been given. According to the characteristics that limited dynamic range of images taken in dense fog, dynamic range stretch as a global enhancement method is used. In order to reduce the effect of noise it brings, sparse coding is adopted.In this paper, computer techniques and images process techniques are used to enhance the clearness of images taken in fog automatically, which has a great significance in improving the robustness of outdoor surveillance systems and also be helpful to the process of images taken in other bad weather.更多还原