基于空间关系的子区域最佳匹配算法  

Optimal Matching Based on Spatial Relationship in Local Regions

作　　者：徐丹[1] 章毓晋[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《电子与信息学报》2005年第8期1177-1181,共5页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(60172025)资助课题

摘　　要：该文提出了一种基于空间关系的图像检索算法——子区域最佳匹配法。这种算法首先根据符号投影模型,将图像内目标抽象为带有多个属性的图标。然后,利用搜索邻节点的等角度区域最近原则、划分子区域的中心点法、图论中的广度优先算法进行图标匹配。最后,根据所定义的局部区域相似距离及图间距离来计算图像与图像之间的相似度,进而完成基于图像内部空间关系的检索。该文以这种算法为核心,构建了一个篮球站位分析演示系统,并用实际数据验证了所提出的基于空间关系的图像检索方法的有效性。A new algorithm named optimal matching based on spatial relationship in local regions for content-based image retrieval is proposed in this paper. The algorithm first employs the symbol mapping model to reduce a symbolic image to a l-D representation, and then uses the principle of nearest distance within equal angle region for searching nearest nodes, takes the center point for dividing images into sub-regions, and applies the breadth-first traversal algorithm in graph theory to perform graph matching. Finally, it uses the defined similar distances between local regions and among different graphs to compute the similarity and to retrieve images based on inner spatial relationship. Based on these algorithms, a prototype system for analyzing the position of players in basketball game is constructed. The effectiveness of image retrieval based on spatial relationship is verified in this system by using some real data.

关 键 词：基于内容的图像检索 空间关系 子区域划分 遍历 用户定制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]