结合空间金字塔局部特征的谱匹配算法

摘要

In order to improve the matching accuracy of image matching algorithm based on spectral decomposition, the image local feature description vector is obtained by improving Center-Symmetric Local Binary Pattern(CS-LBP) and introducing the spatial pyramid method of scale division. Then the adjacent matrix is reconstructed according to the similarity between description vectors,and the match-ing relation between the image feature points is acquired through the spectral decomposition. The con-trast experiment results show that the matching algorithm has high accuracy.%为提高基于谱分解的图像匹配算法的匹配精度，通过对中心对称局部二值模式（CS-LBP）进行修正，并引入空间金字塔尺度划分方法，获取图像局部特征描述向量；并以此向量之间相似性作为度量方式，重构邻接矩阵，通过谱分解获取特征点匹配关系。通过对比实验，结果表明，该算法匹配精度较高。