一种基于遗传算法的最优阈值图像分割算法

为了提高图像分割效率,提出一种基于遗传算法的最优阈值搜索方法OTSGA.OTSGA算法对图像的灰度级进行二进制编码,生成初始种群,求出每个个体的二维最大熵,然后根据设定的寻优准则进行相应的遗传操作以搜索阈值最优解.为了避免在求解过程中出现早熟现象,OTSGA算法将交叉操作得到的个体群与上一代种群混合,得到新的种群进行遗传操作,避免了个别个体在遗传运算的最初迭代时就在种群中占据主导地位,导致求解过程的过早收敛.实验结果表明,OTSGA最优阈值搜索方法不仅降低了运算开销,而且获得了满意的图像分割效果.

Optimal Threshold Image Segmentation Method Based on Genetic Algorithms

In order to improve the efficiency of image segmentation,OTSGA(optimal threshold selection based on genetic algorithm) is proposed as an optimal threshold searching method.OTSGA algorithm encodes gray levels of image to binary code,generates initial population,and calculates the maximum 2D entropy of every individual.Then,according to optimal theorem,it executes genetic operations to get optimal threshold.Specially,in the genetic operations,OTSGA method commixes individuals with their parents' generation for getting new population so that to decrease the probability of premature converge which is caused by individual as the leading part of population in early iterative.The experimental result shows that OTSGA algorithm can not only obviously save the running time,but also improve the quality of image segmentation.