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.%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.
展开▼
