基于图的谱系数夹角的特征点匹配  

Feature Points Matching Based on Angle between Spectral Coefficient of Images

作　　者：沈玲[1] 王年[2] 

机构地区：[1]安徽新华学院,安徽合肥230088 [2]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《计算机技术与发展》2015年第12期68-71,共4页Computer Technology and Development

基　　金：安徽省自然科学项目(KJ2012Z133)

摘　　要：提出一种用图的谱系数夹角特征描述图像几何结构的特征点匹配算法。首先通过对两幅待匹配的图像分别构造高斯权Laplace矩阵,并进行奇异值分解(Singular Value Decomposition,SVD)以获得其特征向量,然后由特征向量计算各分量间夹角的余弦并构造对称矩阵,最后对其进行奇异值分解。根据分解结果构造出两幅匹配图像特征点间能够反映其匹配程度的关系矩阵,从而由该关系矩阵实现特征点间的匹配。分别对模拟、真实及合成数据图像进行实验对比,说明了文中算法的有效性和可行性。An algorithm for images features points matching by representing geometric structure of images based on the angle between spectral coefficient vectors was proposed. The algorithm defined Gaussian-weighted Laplacian matrices for the feature points of two ima- ges respectively, obtaind the eigenvectors based on the result which was gotten by the singular value decomposition on the two matrices. Gained a symmetric matrix with the cosine value of the angle between the weight. Then with the result of the decomposition of the sym- metric matrix, get a relationship matrix which denoted the matching degree among feature points. Finally, the algorithm obtained feature points matching of the two images with the relationship matrixes. Experiments on analog images, real-world images and synthetic data demonstrate the effectiveness and feasibility of the approach.

关 键 词：图 谱系数夹角 LAPLACE矩阵 特征点 匹配 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]