基于SIFT算法的图像特征匹配  

Image Feature Matching Based on SIFT Algorithm

作　　者：周颖[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《现代计算机（中旬刊）》2015年第2期63-68,共6页Modern Computer

基　　金：国家自然科学基金(No.60903118;No.60832011)

摘　　要：SIFT特征匹配算法的原理在于生成特征点的SIFT特征向量,通过对特征向量之间的匹配来实现图像之间的匹配。SIFT特征是一种尺度不变的局部图像特征,阐述生成SIFT特征向量的具体过程,包含尺度空间构建、关键点的检测和精确定位、关键点方向向量的确定和最终SIFT特征描述子的形成等步骤,以及根据形成的特征描述子进行图像的匹配。根据实验结果得出SIFT算法可以有效准确地实现图像之间的匹配。The principle of SIFT feature matching algorithm is to generate SIFT feature vector of the characteristic points, through the matching of feature vector to realize the matching of the images. SIFT feature is a kind of local image characteristics which is invariant to image scale. Expounds the specific process of how SIFT feature vector is generated, including the building of the scale space, the detection of the key points and accurately positioning these points, determines the direction of the feature vector, and finally form the SIFT features vector and according to the vector to realize the image matching. According to the experimental results it is concluded that SIFT algorithm can effectively and accurately realize the matching of images.

关 键 词：图像匹配 SIFT特征匹配 尺度空间 方向向量 特征描述子 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]