基于GLCM特征的改进FCM的SAR图像分割方法  

Modified FCM SAR image segmentation method based on GLCM feature

作　　者：刘健[1] 程英蕾[1] 孙纪达[1] 

机构地区：[1]空军工程大学电讯工程学院,陕西西安710077

出　　处：《计算机工程与设计》2012年第9期3502-3506,共5页Computer Engineering and Design

基　　金：陕西省自然科学基金项目(2010JM8038)

摘　　要：为了克服了较大窗口提取图像边缘处特征值的不足,提出一种基于GLCM特征矩阵的动态滑动窗口算法。针对模糊C均值算法中,聚类中心不容易确定,聚类容易陷入局部最优解的问题,将粒子群优化算法(PSO)引入到聚类算法中,实现全局搜索。应用改进的模糊C均值算法完成了基于SAR纹理特征的图像分割,克服了传统聚类算法仅依赖灰度值进行分割的局限性,也一定程度上克服了斑噪声对SAR图像分割的影响。实验结果表明,该方法应用于SAR图像分割时,取得了很好的分割效果。A dynamic gliding window for computing images＇ GLCM feature matrix is suggested,which has broken through the limitation of images＇ defects happened on the edges in a bigger window.In order to overcome the difficulties for deciding the clustering centers and to avoid getting clustering into the local minimum during the computation by fuzzy C-means（FCM） algorithm,a new FCM method combined with the particle swarm optimization（PSO） algorithm is proposed to segment SAR images.This new method not only gets over the limitation that the traditional clustering algorithms only rely on the information of gray value,but also keeps SAR images from influencing by speckle noise in some extent.The simulation results indicate that this modified method works very well for the SAR images＇ segmentation.

关 键 词：灰度共生矩阵 SAR图像 模糊C均值 粒子群优化 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]