一种基于遗传算法的最小交叉熵阈值选择方法

最小交叉熵阈值法(MCET) 在二级阈值中是有效的, 但在多极阈值的穷尽搜索中却要付出昂贵的时间代价.鉴于此, 提出一种基于遗传算法(GA) 的MCET选择方法: 在执行图像分割(IS) 任务之前, 先将IS 转化为在一定约束条件下待优化的问题; 在寻找待优化问题最优解的计算过程中引入一种回归设计技巧以存储中间结果; 使用这种回归设计技巧, 在一组标准测试图像上利用GA搜索待优化问题的最优解. 实验结果表明, 利用所提出的方法获得的多个阈值非常接近于穷尽搜索获得的结果.

A minimum cross entropy threshold selection method based on genetic algorithm

Although minimum cross entropy thresholding(MCET) is efficient in the case of bilevel thresholding, it encountersexpensive computation when involving multilevel thresholding for exhaustive search on multiple thresholds. Therefore, animproved scheme based on the genetic algorithm is presented for fastening threshold selection in multilevel MCET. Firstly,image segmentation is considered as an optimization problem. Then, this scheme uses a recursive programming technique toreduce the computational complexity of the objective function in multilevel MCET. Finally, a genetic algorithm is proposedto search several near-optimal multilevel thresholds. Simulation results show that the multiple thresholds obtained by usingthe proposed scheme are very close to the optimal ones via exhaustive search on the real images.