摘要
近年来,基于划分的聚类算法被广泛应用于数据和图像聚类中。针对应用最为广泛的k-均值算法在图像聚类中存在的聚类速度慢、效果差等问题,提出一种仿射传播算法应用于图像聚类中。提取图像中颜色、形状和纹理等特征向量,利用仿射传播算法对综合特征向量模型进行聚类,最后将仿射传播算法和k-均值算法对MIT图像的聚类作了对比分析。仿真实验表明,仿射传播算法在速度和聚类效果上均优于已有的k-均值算法,在准确性和实时性方面均能达到较好的效果。
Recently,the clustering methods based on partitioning are widely used in the field of data and image clustering.For the most widely k-means being slow and poor effect,this paper presented an improved algorithm based on affinity propagation,which was applied to image clustering.First,it presented the method combing color,shape and texture feature for efficient image retrieval.Then,on the basis of comprehensive characteristic model,it introduced a novel clustering method based on affinity propagation algorithm,Finally,it compared the results of affinity propagation with k-means in the MIT image database.The simulation experimental results show that the proposed method is superior to the traditional k-means clustering algorithm in the speed and effect of clustering.In addition,it is effective in exactness and real-time property.
出处
《计算机应用研究》
CSCD
北大核心
2012年第10期3980-3982,共3页
Application Research of Computers
基金
陕西省教育厅科技立项项目(2010JK847)
西北大学研究生重点课程项目基金资助项目(09YKC21)
关键词
仿射传播算法
图像聚类
相似度距离
灰度共生矩阵
affinity propagation algorithm
image clustering
similarity metrics
gray-scale co-occurrence matrix