一种多低层特征结合的CBIR检索方法  

Content-based Image Retrieval with Multiple Low-level Features

作　　者：刘瑞军[1] 顾嘉伟 陈谊[1] 

机构地区：[1]北京工商大学计算机与信息工程学院,北京100048 [2]北京大学信息管理系,北京100871

出　　处：《小型微型计算机系统》2015年第6期1336-1340,共5页Journal of Chinese Computer Systems

基　　金：北京工商大学青年教师科研启动基金项目资助;"十二五"国家科技支撑计划项目(2012BAD29B01-2)资助;国家重点实验室开放基金项目(BUAA-VR-14KF-04)资助

摘　　要：基于内容的图像检索技术(CBIR)的低层视觉特征:颜色、形状、纹理.单一的特征不能满足所有的图像检索需求.多特征结合的图像检索可以使在不同特征上相似的图像同时被检索出来.为了研究图像检索优化方法,本文分析比较了6种低层特征提取算法的优缺点和特性;将各个特征的相似度度量以线性加权的方式予以统一;通过比较使用颜色单特征、形状单特征、纹理单特征的图像检索结果和多特征结合的图像检索结果,得出如下结论:低层特征的图像检索在检索效果上优于使用单特征的图像检索,低层特征的特性之间存在互补和抑制.To Content-Based Image Retrieval （CBIR） ,considering the low-level visual features of an image,users might want to re- trieve images similar on color,shape or texture. It is impossible to meet all these needs with a single feature while a combination of multi features can do the work：retrieving images similar on different features at the same time. In order to know the flaws of single- feature-based image retrieval and optimize the result of multi-feature-based image retrieval, 6 low-level feature extraction algorithms are selected to do the experiment after analysis of advantages and disadvantages of these algorithms and characteristics. Based on these algorithms,the similarity measurements of these features are linear weighted to get unified. Finally, a comparison of single-feature- based image retrieval and multi-feature-based image retrieval reveals the research law.regarding to features image retrieval,the retriev- al result of multi-feature-based image retrieval is better than the image retrieval algorithms using single feature;there are complementa- ry and inhibition between each algorithm with low-level features.

关 键 词：基于内容的图像检索 特征提取 多特征结合 颜色特征 形状特征 纹理特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]