基于捕食者-食饵微粒群优化的二维FCM图像分割方法  

A Two-dimensional FCM Image Segmentation Method Based on Predator-Prey Particle Swarm Optimization

作　　者：周鲜成[1,2] 申群太[2] 李清峰[1] 

机构地区：[1]湖南商学院计算机与电子工程学院,湖南长沙410205 [2]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《信息与控制》2009年第1期75-80,86,共7页Information and Control

基　　金：湖南省自然科学基金资助项目(06JJ50110);湖南省教育厅科学研究项目(05C545).

摘　　要：传统模糊C均值聚类算法进行图像分割时仅利用了像素的灰度信息,没有考虑像素的空间邻域信息,因此抗噪性能差.为了克服传统模糊C均值聚类算法的局限性,提出了一种基于捕食者—食饵微粒群算法的二维模糊C均值聚类图像分割方法.该方法将图像的聚类分割转化为一个优化问题,根据像素的灰度信息和改进二维直方图描述的像素邻域关系特性,建立包含邻域信息的适应度函数,并利用捕食者—食饵微粒群的全局优化能力,通过迭代优化获得最优聚类中心,实现图像分割.仿真结果表明,所提算法不易陷入局部最优,抗噪能力强,聚类正确性高,分割效果好,是一种有效的图像分割算法.The traditional fuzzy C-means （FCM） clustering algorithms only make use of gray information of the pixels, and takes no account of the spatial neighbor information in image segmentation, which leads to poor anti-noise performance. In order to overcome limitations of the traditional FCM algorithms, a two-dimensional fuzzy C-means clustering method based on predator-prey particle swarm optimization is proposed for image segmentation. In this method, the image segmentation is converted into an optimization problem. The fitness function containing neighbor information is set up based on the gray information and the neighbor relations between the pixels described.by the improved two-dimensional histogram. Using the global optimization ability of the predator-prey particle swarm, the image segmentation can be accomplished by iterative optimization to obtain the optimal cluster center. Simulation results show that the proposed method can effectively avoid getting into local optimum. With its strong anti-noise capability, high clustering accuracy and good segmentation effect, the presented method is an effective algorithm for image segmentation.

关 键 词：微粒群算法 捕食者—食饵模型 模糊C均值(FCM)聚类 二维直方图 图像分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]