图像模糊聚类分割初始聚类中心优化算法研究  

Research on Optimization Algorithm of Initial Clustering Center for Image Segmentation Based on Fuzzy Clustering

作　　者：范浩祥 张小凤[1] FAN Hao-xiang;ZHANG Xiao-feng(School of Physics&Information Technology,Shaanxi Normal University,Xi'an Shanxi 710119,China)

机构地区：[1]陕西师范大学物理学与信息技术学院,陕西西安710119

出　　处：《计算机仿真》2022年第12期378-382,501,共6页Computer Simulation

基　　金：国家自然科学基金(11874252)。

摘　　要：针对模糊C均值聚类使用随机初始化聚类中心的方式可能导致聚类结果陷入局部最优和聚类性能不稳定的问题,提出了利用灰度直方图区域中位点的聚类中心初始化方法。上述算法结合直方图灰度合并谷值寻优确定聚类数量,通过划分灰度级别区域,计算直方图的像素数量中位点来确定待分割图像的初始聚类中心,并将初始化中心应用于模糊C均值算法以实现图像的快速聚类分割。利用Lena图像和BSDS500数据集中的图片对算法的性能进行测试。仿真结果表明,灰度直方图区域中位点法与模糊C均值算法结合,可以用较低的运算量逼近图像的聚类中心区域,提升图像聚类分割的效率。For fuzzy c-means clustering, random initialization of cluster centers may lead to the clustering results falling into local optimization and clustering performance instability, it proposes the gray histogram region median point method. In this algorithm, the number of clusters is determined by combining histogram gray and valley optimization. The initial cluster center of the image to be segmented is determined by dividing the gray level region and calculating the number of pixels in the histogram. The initial cluster center is applied to the fuzzy c-means algorithm to realize the fast clustering segmentation of the image. Lena images and bsds500 images are used to test the performance of the algorithm. The simulation results show that the combination of the gray histogram region median point method and the fuzzy c-means algorithm can approach the clustering center region of the image with a lower amount of computation and improve the efficiency of image clustering segmentation.

关 键 词：聚类分割 灰度直方图 初始聚类中心 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]