基于图论Gomory_Hu算法的快速图像分割.pdfVIP

第25卷第 9期 计 算 机 应 用 研 究 Vo.l 25 No. 9 2008年9月 ApplicationResearch ofComputers Sep. 2008 GomoryHu* 1 1, 2 1 1 , , , (1.西北工业大学应用数学系, 西安 710072; 2. 中国科学院遥感应用研究所,遥感科学国家重点实验室, 北京 100101) : omoryHu算法是图论中的经典算法, 用于寻找图的最小流割等价树, 具有最优解, 但是该算法很难处 理较大的图像,而且倾向于分割出孤立点集为此, 给出了孤立点的判定 法, 并提出一种基于 omoryHu算法 的图像分割 法该算法首先通过快速聚类减少图中顶点数目, 然后构造新的赋权图, 并应用 omoryHu算法 对图进行最优划分, 得到分割结果提出的算法对多幅自然图像进行了分割实验,平均分割时间在 3 s内实验 结果证明了算法的有效性和快速性 : 图像分割; omoryHu算法; 聚类; 图论 : TP391 : A : 2008) Fast image segmentation based on omoryHu algorithm 1 1, 2 1 1 LIU Bingtao, TIAN Zheng , ZHOU Q iangfeng , LI X iaobin ( 1. Dep t. of App liedM athematics, N orthw estern P olytechnical University, X i an 710072, Ch ina; 2. N ationalK ey aboratory of R em ote Sensing Science, Institute of R emote Sensing App lications, Ch ineseA cademy of Sciences, B eij ing 100101, China) Ab stract: omoryHu algorithm is a classical algorithm for finding them inmi um flow and cut equivalent tree of the graph in graph theory, ithad the optmi al solution. But itwas very difficult to dealw ith large mi ages, and itwas biased toward finding small components. In order to solve this problem, thispaper presented a noveldefinition of isolated node, and proposed a no vel algorithm based on mi proved omoryHu algorithm. This algorithm first used the fast clustering algorithm to reduce the number of vertices, then used the mi proved omoryHu algorithm in thenew graph and achieve theoptmi al solution. Then ap plied thenew algorithm to a number of natural mi ages. The expermi entalresultsdemonstrate that the algorithm is effe