基于Max-tree的连通区域标记新算法  

Connected Component Labeling Algorithm Based on Max-tree

作　　者：章德伟[1] 蒲晓蓉[1] 章毅[1] 

机构地区：[1]电子科技大学计算机科学与工程学院计算智能实验室,四川成都610054

出　　处：《计算机应用研究》2006年第8期168-170,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60471055);电子科技大学青年科研基金资助项目(L08010601JX04030)

摘　　要：采用灰度图像创建Max-tree的基本思想,提出一种新的二值图像连通区域标记算法。该算法主要采用8-邻域搜索及排序队列方式实现,通过一次扫描二值图像即可完成连通区域标记。提出一种新的8-邻域搜索策略,可以将邻域搜索次数由八次减少到平均四次以下,从而提高了系统效率。此外,还给出一种排序队列的快速实现方法,并将其应用到标记算法中。而且,该算法的运行时间仅与待标记图像的大小有关,与连通区数目和图像内容无关。该算法已应用于海藻图像识别,实验结果表明该算法是快速、高效的。Based on the theory of Max-tree, a new connected component labeling algorithm is proposed in this paper. The proposed algorithm can label the connected components by scanning the binary image on time. By improving the traditional algorithm of 8-Neighbor Searching, the average Neighbor Searching times can be reduced from eight to four. An efficient implementation of ordered queue is also introduced in this paper. The experiments on diatom images identification have been done to test the performance of the proposed algorithm, which show that the new algorithm has better performance than the traditio-nal ones.

关 键 词：Max-tree 连通区域标记 8-邻域搜索 排序队列 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]