结合邻域信息的改进抑制式FCM图像分割方法  

Improved Suppressed FCM Algorithm for Image Segmentation Based on Neighborhood Information

作　　者：兰红[1,2] 闵乐泉[1,3] 

机构地区：[1]北京科技大学自动化学院,北京100083 [2]江西理工大学信息工程学院,江西赣州341000 [3]北京科技大学数理学院,北京100083

出　　处：《电视技术》2013年第17期17-21,共5页Video Engineering

基　　金：国家自然科学基金项目(61074192);北京科技大学冶金工程研究院基金项目(YJ2010-019);江西省教育厅重点项目(赣教技字[12770]号)

摘　　要：针对模糊C均值聚类(FCM)算法在含噪声图像分割中存在的运算开销大、收敛速度慢等不足,在抑制式RC-FCM算法的基础上提出一个结合邻域信息的改进抑制式FCM算法,算法通过构建图像灰度及邻域信息的二维直方图设定聚类样本空间和初始聚类,根据样本到聚类中心的最大和最小距离动态设定抑制因子,再通过改进的最小抑制FCM算法实现图像分割。实验表明改进算法不仅能够提高FCM算法的聚类速度,而且提高了对噪声图像的分割精度。Aim at the shortcomings of traditional Fuzzy C-Means clustering（FCM） algorithm in computing time and convergence speed for noise image segmentation ,an improved suppressed FCM algorithm is proposed in this paper. The novel algorithm is based on RC-FCM algorithm and combined with neighborhood information of the image. Firstly,it initializes sample space and cluster centers with two-dimention histogram,which is constructed by the image grayscales and their neighbours＇ pixel values. Then ,it dynamically sets the suppressing factor according to the maximum and minimum distance from the sample data to each cluster center. Finally,it realizes image segmentation with improved minimum suppressed FCM algorithm. Experimental results prove that the proposed algorithm is not only effective to improve the speed of clustering,but also be helpful to achieve better segmentation results for noise images.

关 键 词：抑制式FCM算法 邻域信息 抑制因子 图像分割 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]