《通信工程本科毕业设计》由会员分享，可在线阅读，更多相关《通信工程本科毕业设计（113页珍藏版）》请在金锄头文库上搜索。
1、I摘要摘要图像拼接就是将数 幅（本文主要研究的是两幅图像拼接） 有重叠部分的图像(可能是在不同时间、在不同的角度拍摄的)拼接成一幅大型的无缝高分辨率图像的技术,该技术经过计算机视觉、计算机图形学、图像处理等领域的研究，已经广泛应用于 遥感图像分析；虚拟现实技术；医学图像处理；军事夜视成像等领域中。在图像拼接过程中，最 为关键的两个步骤是 图像配准和图像融合。其中图像配准是图像融合的基础 ,而且图像配准算法的计算量一般非常大,因此图像拼接技术的发展很大程度上取决于图像配准技术的创新。图像拼接的方法很多 ,不同的图像拼接算法步骤会有一定差异 ,但大致的过程是相同的。一般来说 ,图像拼接主要包括以下
2、五步 :图像预处理；图像配准；建立变换模型；统一坐标变换以及图像融合。以上的五个步骤本文将在后面进行介绍。而本文重点研究的是图像的配准技术。本文着重介绍了基于变换域 和基于特征点的图像配准方法。 本文在对这两种配准技术的研究中发现基于特征点的图像配准方法不易受光照、旋转等因素的影响，而且特征相对像素数量较少，有利于提高速度；而基于变换域的方法虽然计算量比较大，但是可以为图像拼接提供一个良好的初始配准参数。综合以上，本文 针对用基于特征点的方法配准后的图像间因具有旋转及光线强度差异等现象而导致的拼接效果不佳以及拼接速度慢，并且基于变换域的图像配准算法存在局限性，计算量过大 等问题，提出了一种新的
3、基于特征点配准的图像拼接方法，此方法不但具有实际应用的价值，而且降低了计算量。本文对传统的基于特征点的Harris角点检测算法中的角点预处理和角点响应函数两个方面进行了改进，并且在图像预处理时结合了相位相关技术。具体来采用的是以下四个步骤：（1）用相位相关技术计算出相邻两帧图像之间的平移量，从而得到图像之间的大致重叠区域；（2）用改进的是Harris角点检测算法检测角点；（3）根据相似测度NCC(Normalized Cross Correlation)方法提取出匹配特征点对；II（4）用加权平均值法中的渐进渐出的方法实现拼接图像的融合。关关键键词词：图像拼接；基于特征；基于灰度；基于变换域；
4、图像配准；相位相关；Harris 角点检测;NCC 方法。IIIAbstractImage mosaic technology is a few images that have overlapping portions of the image (possibly different time, different perspectives or different sensors) built a large seamless high resolution image technology，which through computer vision, computer graphics,
5、 image processing and other areas of research, has been used widely in remote sensing image analysis; Virtual reality technology; Medical image processing; and Military night vision in the field of imaging. Image registration and image fusion image mosaic are two key technologies. Image registration
6、 is the basis of image fusion, and image registration algorithm is very large, so the image mosaic technology development depends largely on the image registration technology innovation.Image mosaic method a lot of, different image stitching algorithm steps will have some differences, but roughly th
7、e process is the same. Generally speaking, image stitching mainly comprises the following five steps: image processing; image registration; establishment of transformation model; coordinate transformation; image fusion. In this paper, the above five steps will introduce in the back. And this paper f
8、ocuses on the image registration technology.Image registration methods, this paper introduced emphatically based on the transform domain and image registration based on feature point method. In these two registration techniques found in the study of image registration based on feature points method
9、is not affected by light, rotation and other factors, and the characteristics of relative pixel number, is beneficial to improving the speed; and based on the transform domain method although the computational amount is larger, but can provide a good image stitching initial registration parameters.
10、Comprehensive above, because of the method of feature point with the image registration due to its rotation and between light intensity differences to the phenomenon such as the joining together of the effect not beautiful and joining together slowly, and based on the transform domain image registra
11、tion algorithm limitations, computing dimension and IVso on, this paper put forward a new feature point registration based on the image matching method, this method not only has the value of practical application, and reduce the computation time.This paper improves the Harris corner method of tradit
12、ional feature point corner in detection algorithm pretreatment and angular point response function and the image preprocessing when combined with fragrance related technology two aspects,. Specific to use is the following four steps:(1) Use phase related technical calculated two frames between adjac
13、ent of translation quantity, thus obtains the image of the overlapping area between roughly;(2) Detect algorithm corner detection with the improvement of Harris is corner;(3) Extract the matching feature points according to the similarity measure NCC (Normalized Cross Correlation) method ;(4) Join t
14、ogether image fusion through the weighted average method with the gradual gradually out of the method.Key words: image mosaic based on feature based on gray level; based on transform domain; image registration; Phase correlation; Harris corner detection; NCC method.V目录目录摘要摘要.IIIAbstract.V第第一一章章 绪绪论论.1第第一一节节 引引言言.1第第二二节节 论论文文研研究究的的意意义义 .1第三节第三节 图像拼接技术概述图像拼接技术概述 .
