基于两阶段改进FCM算法的彩色血液图像分割研究  

Study of Color Blood Image Segmentation Based on Two-stage-improved FCM algorithm

作　　者：王彬[1] 陈槐卿[1] 黄华[2] 饶洁[2] 

机构地区：[1]四川大学华西医学中心生物医学工程研究室,成都610041 [2]四川大学电气信息学院通信工程系,成都10065

出　　处：《生物医学工程学杂志》2006年第2期282-286,共5页Journal of Biomedical Engineering

摘　　要：介绍了一种基于FCM算法(Fuzzy c-m eans a lgorithm)的彩色血液细胞图像分割新方法。通过将原始血液显微图像转换为索引图像再对颜色映射表矩阵做模糊聚类来回避直接对像素值聚类,大大压缩了FCM算法的计算数据量。针对彩色血液细胞图像的固有特性将分割过程分为两阶段,包括确定聚类数目和初始聚类中心,并在第二阶段引入一个距离加权矩阵,改变距离度量方式以提高聚类的准确度。解决了FCM算法难以准确收敛到最佳极值的问题,减少了迭代收敛的次数,并且缩短了算法执行时间,实现了对彩色血液细胞图像各组分的正确分割。This paper introduces a new method for color blood cell image segmentation based on FCM algorithm. By transforming the original blood microscopic image to indexed image, and by doing the colormap, a fuzzy apparoach to obviating the direct clustering of image pixel values, the quantity of data processing and analysis is enormously compressed. In accordance to the inherent features of color blood cell image, the segmentation process is divided into two stages： （1）confirming the number of clusters and initial cluster centers; （2） altering the distance measuring method by the distance weighting matrix in order to improve the clustering veracity. In this way, the problem of difficult convergence of FCM algorithm is solved, the iteration time of iterative convergence is reduced, the execution time of algarithm is decreased,and the correct segmentation of the components of color blood cell image is implemented.

关 键 词：血液细胞 彩色图像分割 FCM算法 索引图像 

分 类 号：R318[医药卫生—生物医学工程] R445[医药卫生—基础医学]