摘要图像拼接技术就是将数张有重叠部分的图像（可能是不同时间、不同视角或者不同传感获得的）拼成一幅大型的无缝高分辨率图像的技术。要求最大程度与原始图像接近，失真尽可能小，且没有明显的缝合线。图像拼接技术主要包括两个关键环节即图像配准和图像融合。对于图像融合部分，由于其耗时不太大，且现有的几种主要方法效果差别也不多，所以总体来说算法上比较成熟。而图像配准部分是整个图像拼接技术的核心部分，它直接关系到图像拼接算法的成功率和运行速度，因此配准算法的研究是多年来研究的重点。
图像配准的目的是得到宽视野的全景图。在本文中，首先，采用SIFT算法提取原始图像中的局部极值点作为初步的特征点，并剔除不稳定点和边缘响应点；其次，采用K-D树算法，计算空间欧式距离比来进行特征点的初步匹配；最后，采用RANSAC算法进行精匹配，并计算图像间的变换矩阵，以此来实现图像间的配准。
图像融合（Image Fusion）是指将多源信道所采集到的关于同一目标的图像数据经过图像处理和计算机技术等，最大限度的提取各自信道中的有利信息，最后综合成高质量的图像。在本文中，就是把两幅经过刚体变换，仿射变换，透视变换的图像进行融合。变成一幅高质量的全景图。23217
本文的实验是在64位Windows 7操作系统，MATLAB软件上实现的，实验所用测试图由本人在学校用手机拍摄并进行过下采样处理。实验证明，本文算法对日常简单图像可以实现准确快速的拼接。
毕业论文关键词： 图像配准；图像融合；图像拼接；SIFT特征算法；RANSAC算法；K-D算法；图像转换矩阵
Subject ： Multiple Image Stitching Technique Based on The Feature Points
ABSTRACT
Image splicing technology is the number of pieces of overlap image (which may be a different time, different Angle of view or different sensors) into a large seamless high-resolution images of technology. For maximum close to the original image, the distortion as small as possible, with no obvious sutures. Image splicing technology mainly includes two key links of image registration and image fusion for the quality of the image fusion due to its time is not too big, and the effect of several important methods of existing the difference is not much, so relatively mature in general algorithm. Part and image registration is the core part of the image matching technology, it is directly related to the success rate and speed of image matching algorithm, so the registration algorithm research is the focus of research for many years.
The purpose of image registration is to get a panoramic view of the wide field. In this article, first of all, using the SIFT algorithm to extract the local extreme value point of the original image as the preliminary feature points; Secondly, using the K - D tree algorithm, the computational space Euclidean distance ratios of feature points matching; Finally, use RANSAC algorithm to fine matching, and calculate the transformation matrix between images, in order to realize image registration.
Image Fusion, (Image Fusion) refers to the source channel more of the same target collected by the Image data through Image processing and computer technology, such as maximum extraction favorable information in each channel, and finally integrated into high quality images. In this article, is to make two paintings through the rigid body transformation, affine transformation and perspective transformation of image fusion. Into a high quality of the panorama.
Experiment of this article is in a 64 - bit Windows 7 operating system, the MATLAB software, the experiment used test photo taken by my mobile phone at school and for the sampling process. Experiments show that this algorithm for daily simple image can realize fast splicing. 基于特征点的多幅图像拼接技术研究:http://www.751com.cn/tongxin/lunwen_16171.html