基于S曲线的多尺度Retinex算法的井下图像增强研究

发布时间：2018-06-25 12:28

本文选题：煤矿井下 + 图像增强　；参考：《辽宁工程技术大学》2015年硕士论文

【摘要】：由于煤矿井下的环境特殊,生产中需要对井下情况进行监控,但是井下图像的恶劣环境导致井下图像存在质量差、可读性低、昏暗、分辨率低、模糊等问题。造成这类问题的主要的原因有：井下无自然光,全靠灯光照明,光线昏暗,光强度较低,且光线不均匀；矿井作业导致环境中粉尘多,致使清晰度不高,视野模糊。为了解决上述问题,提高煤矿井下工人生产作业的安全性和生产工作的效率,需要对井下的图像进行一定的增强处理。近年来,随着研究的不断深入,Retinex图像增强算法的优越性日益凸显,特别是针对如井下图像这类对比度较大的图像,这类图像的灰度空间分不出现两极分化的现象,Retinex算法专门针对该种图像有非常优质的增强效果。应用最广泛的是多尺度的Retinex增强算法(MSR),该算法能有效地对井下图像进行增强,使之亮度、清晰度、都得到提高,但同时也存在过增强、光晕等不足。为此,本文提出了一种改进的算法,应用S曲线的多尺度Retinex井下图像增强算法,这里使用的S型函数是对sig函数的变形,函数图像具有S形状,为不对称函数,灵活性较好,可对不同对比度的图像进行不同程度地增强。最后为了验证算法的有效性,以及相对于其他算法的优越性,利用Matlab R2099a对几幅低质图像分别用传统方法和本文算法对其进行了仿真处理,得出几组处理后的对比图像,并且对原图和处理的结果图像进行量化,实验结果表明：应用S曲线的多尺度Retinex井下图像增强算法处理后的图像不论是在亮度、对比度、清晰度,还是细节信息的保留上其处理效果都明显优于传统算法,该结论从量化后的数字上更能够直观地体现,本文算法能够获得了最佳的视觉效果。
[Abstract]:Because of the special environment of underground coal mine, it is necessary to monitor the downhole condition in production, but the bad environment of downhole image leads to the problems of poor quality, low readability, dim, low resolution, fuzzy and so on. The main reasons for this kind of problems are as follows: there is no natural light underground, all rely on lighting, the light is dim, the light intensity is low, and the light is uneven; the mine operation results in more dust in the environment, resulting in low definition and blurred vision. In order to solve the above problems and improve the safety and efficiency of underground workers' production, it is necessary to enhance the image processing. In recent years, the superiority of Retinex image enhancement algorithm is becoming more and more obvious with the development of research, especially for those images with large contrast, such as downhole image. There is no polarization in gray space of this kind of image. Retinex algorithm has excellent enhancement effect for this kind of image. The multiscale Retinex enhancement algorithm (MSR) is the most widely used algorithm, which can effectively enhance the brightness and sharpness of underground image, but it also has some shortcomings such as enhancement and halo. For this reason, this paper presents an improved algorithm, which applies the multi-scale Retinex image enhancement algorithm of S curve. The S-type function is used to deform the sig function. The function image has S-shape and is an asymmetric function, so it has good flexibility. Images with different contrast can be enhanced to different degrees. Finally, in order to verify the validity of the algorithm and its superiority over other algorithms, several low-quality images are simulated by using Matlab R2099a, respectively, and some groups of processed contrast images are obtained by using the traditional method and the algorithm in this paper. And quantization of original image and processed result image, the experimental results show that the image processed by the multi-scale Retinex image enhancement algorithm of S curve is not only in brightness, contrast, clarity, but also in brightness, contrast and sharpness. The processing effect of this algorithm is obviously better than that of the traditional algorithm on the retention of detail information. This conclusion can be more intuitively reflected from the quantized numbers, and the algorithm in this paper can obtain the best visual effect.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD67;TP391.41

【相似文献】