基于最小生成树的LAPLACE谱图像匹配算法  

Laplacian Spectrum Image Matching Algorithm Based on Minimum Spanning Tree

作　　者：王年[1] 周梅菊[1] 张江[1] 梁栋[1] 

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

出　　处：《系统仿真学报》2009年第17期5481-5485,共5页Journal of System Simulation

基　　金：国家自然科学基金(10601001;60772121);安徽省自然科学基金(070412065);安徽省教育厅自然科学研究项目(2006KJ030B;2008B024)

摘　　要：提出了一种基于最小生成树的Laplace谱图像匹配算法。首先分别对两幅待匹配图像的特征点集构造完全图,其次寻找最小生成树,然后通过各自的最小生成树构造Laplace矩阵,接下来进行分解并利用分解结果构造匹配矩阵,最后通过匹配矩阵实现两幅图像匹配。实验验证了该算法能够降低匹配的时间复杂度和获得较高的匹配精度。A Laplacian spectrum image matching algorithm based on minimum spanning tree was proposed. Firstly, a complete graph was constructed according to the feature points of two related images respectively, and then minimum spanning tree in each complete graph was searched. Secondly, a Laplacian matrix for minimum spanning tree was constructed respectively. Thirdly, the Laplacian matrices were decomposed respectively and then a matching matrix denoting the matching degree among feature points was constructed by using the results of the decomposition. Finally, the matching feature points of the two images were obtained according to the matching matrices. Experimental results indicate that the given algorithm can decrease matching time complexity and possess the higher matching accuracy.

关 键 词：图论 LAPLACE谱 最小生成树 图像匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]