基于改进加权图转换的图像匹配算法  

Algorithm of image matching based on improved weighted graph transformation

作　　者：闫占凯 刘志波 张官亮[2] 朱友清[3] 张蓓 

机构地区：[1]武警河北省总队司令部通信站,石家庄050051 [2]武警乌鲁木齐指挥学院,乌鲁木齐830049 [3]国防科学技术大学电子科学与工程学院,长沙410073

出　　处：《计算机应用研究》2014年第4期1256-1259,共4页Application Research of Computers

摘　　要：对提出的基于马氏距离的点匹配方法进行了理论分析与实验验证,针对马氏距离及加权图转换匹配方法的不足,将马氏距离融入到加权图转换匹配算法中,提出了一种新的稳健的图像匹配策略——基于马氏距离加权图转换的图像匹配算法。该算法利用图中的点及其K-近邻点的马氏距离中值和角度距离建立权重矩阵,根据不断更新得到的权值更新图,逐个剔除出格点,获得更加精确的匹配结果。仿真数据和真实图像实验对比结果表明该方法的可行性和鲁棒性。This paper proposed a point matching algorithm based on Mahalanobis distance and analyzed its performance. Due to the limitation of Mahalanobis distance and weighted graph transformation, it embedded the similarity evaluated by mahalanobis distance into WGTM algorithm. It proposed a novel and robust image matching strategy, the algorithm based on weighted graph transformation using Mahalanobis distance. It built the weight matrix by using the median of mahalanohis distance and angular dis- tances between edges that connect a feature point to its K-nearest neighbors in the graph. It updated the graph according to the weight ever-updating. It obtained then precise results of the point pairs matching through iteratively eliminating the outliers. Ex- perimental results on synthetic data and real-world data have demonstrated that the proposed algorithm is effective and robust.

关 键 词：图像匹配 马氏距离 加权图转换 K-近邻 角度距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]