一种采用双势阱策略的小直径图分割方法  

A SMALL DIAMETER GRAPH PARTITION METHOD WITH DOUBLE-WELL STRATEGY

作　　者：汪云飞[1] 毕笃彦[1] 孙毅 孙超[1] 南栋[1] 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038 [2]解放军驻上海胶带股份公司军事代表室,上海200235

出　　处：《计算机应用与软件》2013年第4期275-278,共4页Computer Applications and Software

摘　　要：等周算法用于图像分割时存在迭代效率低的缺陷。提出一种新的双势阱策略,该策略采用度最大和度最小的图结点共同作为接地点,提高了线性方程的求解效率;同时用添加随机边的方法缩短图直径,减少了共轭梯度法的迭代量,进一步提高了算法的收敛速度。仿真实验结果表明,新算法提高等周算法迭代效率达20%以上。Isoperimetric algorithm has the defect of low iterative efficiency when applied in image segmentation,so a new double-well strategy is proposed.In this strategy the image nodes of maximum degree and minimum degree are both used as the ground point,this improves the solution efficiency of linear equation;meanwhile,the method of adding random edges is introduced to shorten the graph diameter,so the numbers of iterations for conjugate gradient method is reduced,as a result the rate of convergence is further improved.Simulating experimental result shows,this new algorithm can efficiently raise the iteration efficiency of isoperimetric algorithm up to 20% or higher.

关 键 词：图像分割 图论 等周算法 双势阱 随机边 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]