图像去雾还原技术研究

发布时间：2019-04-11 14:36

【摘要】：随着我国网络时代的到临,越来越多的场合会用到图像处理技术,其中,在车牌识别,人脸识别,视频监控等领域作用越来越明显。这就要求在具体实践的过程当中充分还原真实的图像以利于识别,但是在实际情况当中,当遇到雾霾天气,由于空气中的大气粒子的散射作用,将导致所拍摄出来的图像变得模糊,对比度与色彩等视觉效果也将大打折扣。因此,在处于恶劣天气下,如何去除雾霾所带来的影响,提高图象质量便显得尤为重要,这也是本文所讨论的重点。本文讨论了两类图像去雾技术,一类以图像增强为主,图像的增强技术主要表现在图像的空域增强、频域增强以及基于色彩恒常性的Retinex增强算法,其中重点讨论并在不同情境下实验了基于Retinex的多种图像去雾算法;另外一类以图像复原技术为主,通过对图像降质模型的分析,以及对大量无雾图像的统计与实验,形成了基于暗原色先验的图像去雾技术。本文细致分析了雾霾的形成原因以及具体的图像降质模型,并通过实验验证基于暗原色先验的图像去雾技术的效果,对实验结果中表现出的时效性不足的问题进行了修正,采取了一种基于引导滤波的透射率优化算法提高了算法效率。最后本文通过对图像降质模型的的分析与研究,采取了基于类高斯滤波及自适应中值滤波的快速去雾算法,不仅提高了去雾之后图片的质量,与暗原色先验算法相比提高了图像成像的速度。本文各算法的模拟实验均基于MATLAB 2012下运行,对于最终的实验结果通过对比度、平均梯度、信息熵以及EPI这四个量化参数进行具体对比,并分析各算法的优劣。
[Abstract]:With the advent of the Internet era in China, image processing technology will be used in more and more occasions, in which, the role of license plate recognition, face recognition, video surveillance and other fields is becoming more and more obvious. This requires that the real images be fully restored in the process of concrete practice to facilitate identification, but in practice, when the haze weather is encountered, due to the scattering of atmospheric particles in the air, The image will be blurred and visual effects, such as contrast and color, will be greatly compromised. Therefore, in severe weather, how to remove the impact of haze and improve the image quality is particularly important, which is also the focus of this paper. In this paper, two kinds of image de-fogging techniques are discussed, one is image enhancement, the other is image enhancement in spatial domain, frequency domain and Retinex enhancement algorithm based on color constancy. Retinex-based image de-fogging algorithms are discussed and tested in different situations. The other is based on the image restoration technology. Through the analysis of the image degradation model and the statistics and experiments of a large number of fog-free images, an image de-fog technique based on a priori dark primary color is formed. In this paper, the causes of haze formation and specific image degradation model are analyzed in detail, and the effectiveness of image de-fogging technology based on dark primary color priori is verified by experiments, and the lack of timeliness in the experimental results is corrected. A transmission optimization algorithm based on guided filter is adopted to improve the efficiency of the algorithm. Finally, through the analysis and research of image degradation model, a fast de-fogging algorithm based on Gao Si-like filter and adaptive median filter is adopted in this paper, which not only improves the quality of image after de-fogging, but also improves the image quality. Compared with the dark primary color priori algorithm, the image imaging speed is improved. The simulation experiments of each algorithm in this paper are based on MATLAB 2012. The final experimental results are compared by four quantization parameters: contrast, average gradient, information entropy and EPI, and the advantages and disadvantages of each algorithm are analyzed.
【学位授予单位】：宁夏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】