基于大气散射模型的图像去雾算法研究

发布时间：2018-05-11 05:32

本文选题：大气散射模型 + 大气耗散函数　；参考：《西安科技大学》2017年硕士论文

【摘要】：随着电子电路技术和光学成像技术的发展,获取的图像分辨率、清晰度越来越高。然而,在雾霾等恶劣天气条件下,捕捉的图像会发生严重退化,给航空航天、工业生产、交通监管及生物医学等领域带来极大的影响。导致图像降质的主要原因是场景目标光线在传播过程中与空气中悬浮的颗粒发生交互作用,从而使得图像细节信息丢失、色调偏移、饱和度降低,无法满足人类的视觉要求。为了提高户外成像系统适用性和稳定性,本文基于大气散射模型的图像去雾算法进行仔细研究,并提出适用性强、稳定性高的去雾算法。本文主要研究内容如下:针对传统偏振去雾算法认为偏振度为全局不变量,导致的图像整体偏暗、层次感低、色调低沉、丢失图像细节信息,提出了改进的基于均值滤波的偏振图像去雾算法。首先利用偏振成像探测系统获得最好与最差状态下的两幅偏振图,然后通过四叉树细分法估算大气光强,实现了大气光强等有关参数的自动寻优,并利用均值滤波的变形形式估算出大气散射光,再根据去雾模型复原无雾图像,最终实现图像去雾。实验表明该算法场景适应力强,具有更高的对比度,更清晰的图像细节和更丰富的色调,有一定的视觉优势。针对基于中值滤波的快速去雾方法存在的不足,提出了一种基于总变差的快速雾天图像复原算法。首先利用总变差对大气耗散函数进行估计,有效的避免了边缘残雾和局部偏暗现象;然后引入直方图修正机制下的自适应保护因子,更正明亮区域的大气散射函数;同时运用何算法通过暗通道快速准确的获取大气光值,降低了时间复杂度,避免了雾图中高亮物体对大气光值估计的影响。最后对复原结果进行亮度调整,得到颜色更加丰富,细节更加清晰的去雾效果图,且算法的时间复杂度是图像像素数的线性函数,在计算速度上取得了较大的提升。
[Abstract]:With the development of electronic circuit technology and optical imaging technology, the resolution and sharpness of the obtained images are becoming higher and higher. However, under severe weather conditions such as haze, the captured images will degenerate seriously, which will have a great impact on the fields of aerospace, industrial production, traffic regulation and biomedicine. The main cause of image degradation is the interaction between the light of the scene target and the suspended particles in the air during the propagation process, which results in the loss of the details of the image, the color offset and the decrease of the saturation, which can not meet the visual requirements of human beings. In order to improve the applicability and stability of outdoor imaging system, the image de-fogging algorithm based on atmospheric scattering model is studied carefully in this paper, and a suitable and stable de-fogging algorithm is proposed. The main contents of this paper are as follows: according to the traditional polarization de-fogging algorithm, the degree of polarization is considered as a global invariant, which leads to the overall dark image, low hierarchy, low tone, loss of image detail information. An improved de-fogging algorithm for polarization images based on mean filter is proposed. First, two polarizations in the best and worst states are obtained by using the polarization imaging detection system, and then the atmospheric light intensity is estimated by the quadtree subdivision method, which realizes the automatic optimization of the relative parameters, such as atmospheric light intensity. The atmospheric scattering light is estimated by the deformed form of mean filter, and then the fog free image is reconstructed according to the defog model, and finally the image is de-fogged. Experimental results show that the algorithm has better adaptability, higher contrast, clearer image details and richer hue, and has some visual advantages. A fast fogging image restoration algorithm based on total variation is proposed to overcome the shortcomings of the fast de-fogging method based on median filter. Firstly, the total variation is used to estimate the atmospheric dissipation function, which effectively avoids the edge fog and the local dark phenomenon, and then the adaptive protection factor based on histogram correction mechanism is introduced to correct the atmospheric scattering function in the bright region. At the same time, we use the algorithm to get the atmospheric light value quickly and accurately through dark channel, which reduces the time complexity and avoids the influence of the highlight object in the fog map on the atmospheric light value estimation. Finally, the brightness of the restoration results is adjusted to obtain a more colorful, more clear detail de-fogging effect, and the time complexity of the algorithm is a linear function of the number of pixels of the image, and the computational speed has been greatly improved.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】