交通卡口图像的超分辨率重建算法研究
时间:2019-01-27 21:16 来源:毕业论文 作者:毕业论文 点击:次
| 摘要图像的超分辨率重建技术就是利用一组低质量的、低分辨率图像(或运动序列)来产生单幅高质量的、高分辨率图像。图像超分辨率重建技术的应用领域极其宽广。在医学，军事，公共安全，计算机视觉等领域都存在着重要的应用前景。图像超分辨率重建技术可以提高图像的识别精度和识别能力。图像超分辨率重建技术可以实现目标物的专注分析，从而来获取感兴趣区域的更高空间分辨率的图像，而不必直接采用数据量巨大的高空间分辨率图像的配置。在数字图像被广泛应用的领域，高分辨率就意着拥有更多的细节信息，可以更好的进行观察、识别。目前的超分辨率重建技术有三类：基于插值的方法，基于重建的方法，基于学习的方法。本文的研究工作如下： 了解并掌握简单的插值算法：最近邻插值，双线性插值，双立方插值。了解基于重建的方法，这类方法所有可用信息都来自于输入图像序列，不包含任何附加的先验知识。学习一种基于学习的方法，SR算法正是将传统多帧基于重建方法和基于学习方法结合的算法框架，在只有一张图像的情况下在放大较高倍数仍然有较好的重建效果。33166 关键词： 超分辨率重建；结构相似性；插值算法；尺度不变性； 毕业论文设计说明书外文摘要 Title ： Research on super resolution reconstruction algorithm of traffic bayonet image Abstract Super resolution image reconstruction technology uses a set of low quality, low resolution image (or sequence of motion) to produce a single high resolution and high quality image. Super resolution image reconstruction is used in many fields including military, medicine, public security, computer vision etc.. Super resolution image reconstruction technique can improve the recognition ability and the accuracy of image. Super resolution image reconstruction technology can realize the analysis object of attention, so it can get the region of interest (ROI) is a high spatial resolution image, and it is unnecessary to apply large amounts of data with high spatial resolution image configuration. In the field of digital image application, high resolution means with more detailed information, which is good for observation and identification. Super resolution reconstruction currently has three categories: interpolation based method, reconstruction based method, and learning based method. The research work of this paper is as follows: We understand and master the simple interpolation algorithm at first in this dissertation：neighbor interpolation bilinear interpolation, bicubic interpolation. Understanding the reconstruction method based on a learning approach, this kind of method employs all the information available in the input image sequence, without any additional prior knowledge. As learning based approach, SR algorithm is a combination of the traditional multi frame reconstruction method and the learning based algorithm frame. In the case of only one image in a higher magnification it still has a better reconstruction effect. Keywords：Super resolution reconstruction; structural similarity; interpolation algorithm; scale invariance; 目录 1 绪论 3 1.1 研究背景和意义 3 1.2 国内外研究现状 3 1.3 未来发展方向 4 1.4本章小结 5 2 关于超分辨率重建技术的基础 5 2.1图像成像原理 6 2.2 简单的插值算法 7 2.2.1 最近邻插值 7 2.2.2 双线性插值 8 2.2.3 双立方插值 8 2.3 本章小结 9 3 基于结构相似性学习的重建算法 10 (责任编辑：qin) |