一种基于进化策略的图像分割方法  

An Image Segmentation Approach Based on Evolution Strategies

作　　者：潘中良[1] 陈翎[1] 张光昭[2] 

机构地区：[1]华南师范大学物理与电信工程学院电子系,广东广州510006 [2]中山大学电子与通信工程系,广东广州510275

出　　处：《激光与红外》2007年第6期583-586,共4页Laser & Infrared

基　　金：国家自然科学基金(60006002);广东省教育厅自然科学研究项目(02019)资助

摘　　要：文中提出一种基于进化策略求解分割阈值的方法,并在方法中引入了部分个体的交叉和个体的年龄参数,以进一步模拟自然界的进化过程,从而改善了整个方法的计算效率。对使用最大类间的方差准则和最大熵准则的实验结果表明,这种方法能够找到较优的分割阈值,可以方便地实现对图像的分割。The image segmentation is a major aspect in image processing, it will produce influences on the image analysis and object recognition. The thresholding segmentation is one of image segmentation methods whose task is to search for the optimal threshold used by segmentation. In this paper, an image segmentation approach using evolution strategies technique is presented. The evolution strategies use mainly the selection and mutation operations to realize the evolutions of populations. In order to improve the efficiency of the approach, the crossover operations among a part of individuals in population and the age parameter of an individual are introduced into the basic evolution strategies. Some experimental results show the proposed approach can get the near optimal threshold being used to image segmentation.

关 键 词：图像处理 图像分割 进化算法 进化策略 

分 类 号：TN911.73[电子电信—通信与信息系统]