基于随机权重粒子群和K-均值聚类的图像分割  

An Image Segmentation Algorithm Based on Random Weight Particle Swarm Optimization and K-means Clustering

作　　者：李海洋[1,2] 文永革[1] 何红洲[1] 李柏林[3] 

机构地区：[1]绵阳师范学院数学与计算机科学学院,四川绵阳621000 [2]北京邮电大学计算机学院,北京100876 [3]西南交通大学机械工程学院,四川成都610031

出　　处：《图学学报》2014年第5期755-761,共7页Journal of Graphics

基　　金：四川省科技厅资助项目(2012JYZ013);四川省教育厅资助项目(12ZB070);绵阳师范学院资助项目(2013A12)

摘　　要：K-均值聚类具有简单、快速的特点,因此被广泛应用于图像分割领域。但K-均值聚类容易陷入局部最优,影响图像分割效果。针对K-均值的缺点,提出一种基于随机权重粒子群优化(RWPSO)和K-均值聚类的图像分割算法RWPSOK。在算法运行初期,利用随机权重粒子群优化的全局搜索能力,避免算法陷入局部最优;在算法运行后期,利用K-均值聚类的局部搜索能力,实现算法快速收敛。实验表明:RWPSOK算法能有效地克服K-均值聚类易陷入局部最优的缺点,图像分割效果得到了明显改善;与传统粒子群与K-均值聚类混合算法(PSOK)相比,RWPSOK算法具有更好的分割效果和更高的分割效率。K-means clustering is widely used in image segmentation due to its simplicity and rapidity.However,it is easy to fall into local optimum,leading to poor image segmentation results.In order to overcome this disadvantage of K-means,this article proposes a mixed image segmentation algorithm based on random weight particle swarm optimization （RWPSO） and K-means clustering.In the early stages of the algorithm running,it can avoid falling into local optimal using the global search capability of RWPSO.In the later stages of the algorithm running,it can achieve fast convergence using the local search capability of the K-means clustering.Experimental results show that RWPSOK algorithm can effectively overcome the weak global search capability drawback of the K-means clustering.It can significantly improve the image segmentation results.Compared with traditional particle swarm K-means clustering algorithm （PSOK）,RWPSOK algorithm has better segmentation results and higher efficiency.

关 键 词：随机权重 粒子群优化 K-均值聚类 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]