摘要：超分辨率重建算法利用多帧图像的互补信息，从一系列质量较差的低分辨率图像来重建一幅质量较好的高分辨率图像,从而提高图像的清晰度。
首先，本文介绍了几种经典的超分辨率重建算法，在此基础上分析了超分辨率重建的几个关键要素和算法的质量评价标准。其次，本文使用MATLAB编程语言实现了两种空间域的超分辨率重建算法：一种是凸集投影算法，它在一个矢量空间中定义闭合的凸型集合，任意的初始估计向这些约束集合进行投影就能获得高分辨率估计图像；另一种是传统正则化算法，本文的正则化参数选择为部分重建图像的线性函数，通过引入合理的约束来得到高分辨率图像。
仿真结果表明传统正则化算法性能虽然优于一些算法，但是由于只是对图像边缘情况加以控制，得到的高分辨率图像效果不理想。相比较而言，凸集投影算法运算量较大，重建出的图像内噪声减小了许多，边缘也得到了很好的保持。5332
关键词：超分辨率重建；传统正则化法；凸集投影法
Designing of Railway Surveillance video Super-resolution Study
Abstract: Super-resolution restoration refers to restore a high-resolution image from multiple low-resolution images by the comprehensive utilization of wealthy complementary imformation between the multiple frames to improve image acuity.
We first introduce several classical algorithm of super-resolution reconstruction,then we analysis several key elements of super-resolution restoration. On the basis of it,we discuss the quality evaluation criteria of the algorithm. Secondly,we use Matlab programming language to accomplish two spatial super-resolution reconstruction algorithms：one is Projection onto Convex Sets(POCS),it definits closed convex set in the vector space. Any initial estimate project to the constraint set can get high-resolution estimation image, another is conventional regularization algorithm.In this paper,to get high resolution images by introducing reasonable constraints we select some linear function of the reconstructed image as regularization parameter.
The simulation results show that the traditional regularization algorithm performance, although better than some algorithm, but because the situation is just the edge of the image to be controlled, the resulting high-resolution image is not ideal.In comparison, POCS algorithm is more computation,the noise of the reconstructed image is reduced,and the edge of the image has also been well maintained.
Key Words: Super-resolution restoration; adaptive regularization; POCS
目录
1 绪论 1
1.1 研究背景与研究意义 1
1.2 图像超分辨率重建的应用 2
1.3 国内外研究现状 3
1.4 论文内容及主要工作 3
2 经典的超分辨率重建算法介绍 5
2.1 频率域方法 7
2.2 空间域方法 8
2.1.1 非均匀样本内插值法 8
2.1.2 凸集投影法 9
2.1.3 迭代反投影法 9
2.1.4 最大后验概率估计法 9
2.2 超分辨率重建技术的步骤 10
2.3 超分辨率图像重建的质量评价 10
3 两种空间域超分辨率重建算法 12
3.1 凸集投影超分辨率重建算法 12 MATLAB铁路监控视频的超分辨率技术研究:http://www.751com.cn/tongxin/lunwen_2326.html