基于最小路由代价树的大规模显微图像拼接方法  

A Method for Large-scale Microscope Images Mosaicing Based on Minimum Routing Cost Spanning Tree

作　　者：龚咏喜[1] 田原[1] 谢玉波[2,3] 刘瑜[1] 邬伦[1] 

机构地区：[1]北京大学遥感与地理信息系统研究所,北京100871 [2]武汉大学测绘遥感信息工程国家重点实验室,武汉430079 [3]华北计算技术研究所,北京100083

出　　处：《中国图象图形学报》2009年第6期1178-1187,共10页Journal of Image and Graphics

基　　金：国家高技术研究发展计划(863)基金项目(2007AA12Z216;2007AA120502);国家自然科学基金项目(40701134;40771171;40629001)

摘　　要：为了对大规模显微图像进行高质量的拼接,首先提出拼接图的概念及获得高质量全景图像的3个原则,然后采用分块-空间聚类算法配准相邻图像,同时评估配准质量,并计算拼接图的边的权值;最后在此基础上,提出了一种基于最小路由代价生成树的图像拼接方法,该方法通过计算拼接图的最小路由代价生成树来确定所有图像的全局位置,并用来生成全景图像。实验结果表明,该方法可获得高质量的全景图像。In order to build high quality panoramic image, the conception of a mosaicing graph and three rules for high quality panoramic image are presented in the paper. An image registration algorithm based on blocking-spatial clustering is used to calculate the registration position and to evaluate the registration quality of pairs of images to obtain weight of edge in mosaicing graph. Then a method of images mosaicing based on minimum routing cost spanning tree is proposed to calculate global optimum position of every image by constructing the minimum routing cost spanning tree of the mosaicing graph and to create the panoramic image. In the case study, the proposed method demonstrates high quality.

关 键 词：图像拼接 全景图 拼接图 空间聚类 最小路由代价生成树 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]