基于SIFT特征的图像匹配
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
毕业设计(论文)题目基于SIFT特征的图像匹配
姓名张建华
学号**********
所在学院理学院
专业班级08信计
指导教师吴颖丹
日期2012 年 6 月 2 日
摘要
当今社会已经进入信息时代,随着计算机技术、通信技术和数学的发展,图像信息处理能力和水平也不断提高,相应的也得到更多关注、研究和更加广泛的应用。图像匹配是处理和解决各种图像信息的基础,已经成为虚拟现实和计算机可视化领域的研究热点。一直以来,研究人员对图像匹配技术进行了大量的研究,推出了许多匹配算法,其中特征匹配算法有着较高的精确度和稳定性。SIFT (Scale Invariant Feature Transform)特征匹配算法是Lowe提出来的用于图像特征匹配的算法,是目前特征匹配领域的热点,对图像的旋转,尺度缩放和亮度变换保持不变,对视角变换,仿射变换保持一定程度的稳定。SIFT特征点是图像的一种尺度不变局部特征点,具有独特性好,信息量丰富,多量性,高速性,可扩展性等特点。正是借助于这些特点,使得传统图像配准中的许多诸如前面提到的共性问题得到了很大程度的改善。该算法首先给出了尺度空间的生成方法,检测出极值点;接下来给出了SIFT特征点的提取步骤和精确定位极值点的方法;然后基于特征点邻域像素的梯度和方向生成了关键点的描述向量;最后根据特征向量给出了匹配方法,提取了SIFT的特征点,并其应用于图像匹配。
本文首先简要介绍了图像匹配所需的基础知识,然后详细介绍了SIFT算法的具体流程。通过大量的实验证明SIFT算法具有较强的匹配能力和鲁棒性,是一种较好的图像匹配算法。
关键字:SIFT; 图像匹配; 尺度空间; 极值点; 特征向量
Abstract
Today,society has entered the information age, with the development of computer technology, communications technology and mathematics, the image information processing capabilities and the level is also rising, and also get more attention, research and more widely used. The image matching handle and solve all kinds of image information, has become the research focus of the virtual reality and computer visualization. Researchers has been a lot of research for image matching techniques, the introduction of a number of matching algorithm, which feature matching algorithm has higher accuracy and stability. SIFT(Scale Invariant Feature Transform)feature matching algorithm, proposed by Lowe, is a hot field of feature-matching at present, which remains the same to image rotation, scale zoom and brightness transformations, and also maintains a certain degree of stability on the perspective transformation and affine transformation. SIFT feature points are scale-invariant local points of an image, with the characteristics of good uniqueness, informative, large amounts, high speed, scalability, and so on. In this algorithm, at first method for generating image scale space is presented; at second steps for extracting sift key points and accurate positioning are provided; then vectors for describing key points based on the gradient magnitude and orientation of pixels neighboring to the key points are generated; at last according to the vectors matching algorithm is described.
This paper briefly introduces the basic knowledge required for image matching, and then details the specific process of the SIFT algorithm. Large number of experiments to prove the SIFT algorithm has a strong match and robustness, is a good image matching algorithm.
Key Words:SIFT; Image Registration; Scale space; extreme points; eigenvector.