改进的最小类内绝对差阈值分割及快速算法

Image Thresholding Based on Improved 2-D Minimum Within-Cluster Absolute Difference Method and Its Fast Algorithm

现有的最小类内绝对差阈值分割方法分割结果不够准确及计算效率过低,为此,本文提出了基于递推昆沌粒子群的改进最小类内绝对差阈值分割方法。首先引入了灰度级-梯度直方图以提高分割准确性,然后简化了阈值选取公式并推出了相应的递推算法,最后利用基于改进的Tent混沌粒子群算法寻找最优阈值,提出了以递推方式计算适应度,大大减少了重复计算。实验结果表明:与基于灰度级-平均灰度级最小绝对差穷举算法相比,本文方法剔除了边缘点和噪声点的影响,选取的阈值更为准确,同时,利用群体智能优化搜索过程,运算时间降低了两个数量级;与基于灰度级-梯度最大类间方差及Logistic混沌粒子群递推算法相比,本文方法基于改进的Tent混沌映射,遍历性更高,因此收敛性更好。

In view of the inaccurate segmented result and low computational efficiency of the existing thresholding method based on minimum within-cluster absolute difference,an improved thresholding method based on recursive chaotic particle swarm algorithm is proposed in this paper.To achieve higher segmented accuracy,the gray scale-gradient histogram is introduced firstly,and then the formulae for the proposed method are simplified and the corresponding fast recursive method is presented.After that,the improved Tent map chaotic particle swarm algorithm is used to search for the optimal threshold,and the repeat computations of the fitness function in iteration are reduced significantly using recursive mode.Compared with the exhaustive algorithm based on gray scale-average gray scale histogram and minimum within-cluster absolute difference,experimental results show that the new method obtain more accurate threshold and achieve better segmented result since the effect of the edge points and noise points are removed.And because the swarm intelligence is used to optimize the searching process,the running time is reduced by two orders.Compared with method of maximum between-cluster variance based on gray scale-gradient histogram and logistic map recursive chaotic particle swarm algorithm,the new method is based on improved Tent map,higher convergence precision is obtained because of its higher ergodicity.