基于空间直方图的免疫图像聚类算法研究  

Research of immune image clustering algorithm based on space histogram

作　　者：李海芳[1] 王莉[2] 

机构地区：[1]太原理工大学计算机与软件学院,山西太原030024 [2]大同大学数学与计算机科学学院,山西大同037009

出　　处：《计算机工程与设计》2008年第11期2932-2935,共4页Computer Engineering and Design

基　　金：山西省自然科学基金项目(2006011030、2007011050)

摘　　要：随着图像检索系统的发展,现有的各种图像特征提取方法已不能很好地满足用户的要求,另外,合理组织和管理图像数据库已渐渐成为用户检索的关键所在。空间直方图度量法融合了图像颜色特征和颜色空间分布信息对图像进行特征提取,并得到了对应的特征向量,该方法捕获了颜色的空间布局信息,比较完整地从颜色角度描述了一幅图像;利用免疫聚类算法实现图像聚类,主要是通过模拟抗体捕获抗原的机制,对已提取的图像特征向量进行了聚类分析,实验结果表明,免疫聚类较传统k均值聚类有簇内紧凑性高和簇间独立性高的优点,同时也进一步提高了图像检索系统的效率。With the development of image retrieval system, the various existing image feature extraction methods isn＇t unable to effectively meet the user＇s requirements, in addition, reasonable organizing and managing of image database has been a key of users＇ retrieving. Space histogram measurement is proposed, this method has fused color feature and color space distributing information to extract image features, and relevant feature vectors is gained, this method captures color space layout information and describes an image integrally from the perspective of color; immune clustering algorithm is used into image clustering, above image feature vectors is clustered by simulating the principle of antibody capturing antigen. The experiment indicates that immune clustering algorithm has some merits such as strong compactness in each cluster and independence between clusters, which also can improve the retrieval efficiency.

关 键 词：空间颜色直方图 图像分块 人工免疫 图像聚类 情感语义 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]