基于内容的图像检索技术研究  

A Survey on Content-Based Image Retrieval Technology

作　　者：祝晓斌[1] 刘亚奇[1] 蔡强[1] 曹健[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

出　　处：《计算机仿真》2015年第5期1-4,85,共5页Computer Simulation

基　　金：北京市属高等学校科学技术与研究生教育创新工程建设项目(PXM2014_014213_000043);北京市属高等学校教师队伍建设--青年英才计划资助项目(YETP1452)

摘　　要：随着数字图像在多媒体领域的广泛应用,对基于内容的图像检索技术的需求也不断增加。基于内容的图像检索技术总体上可以分为两部分:图像特征提取、图像特征的索引与匹配。图像特征提取主要解决如何在数学上有效地描述一幅图像。文中分别介绍了颜色、形状和纹理特征提取算法近年来的研究成果。图像特征索引与匹配,主要解决如何根据特征描述判断图像间的相似程度,并准确、快速列出图像库中与检索图像相似的图像,分别介绍了相似度测量方法、聚类与分类技术、相关反馈技术三类技术的主要研究成果。最后对基于内容的图像检索技术的研究难点进行了讨论,对未来可能的研究方向进行了展望。With the widely application of digital image in multimedia domain, the requirement of content-based image retrieval （CBIR） is rising. The content-based image retrieval can be broadly divided into two parts, namely image feature extraction, and the image indexing and matching. Feature extraction targets at effectively describing an image. And we elaborate the sate-of-the-art algorithms in color, shape and texture feature extraction in recent years. As to image feature indexing and matching, it aims at evaluating the similarity of images based on their extrac- ted feature descriptors, and retrieving a series of similar images with query from candidate dataset efficiently and ac- curately. For feature indexing and matching, we review the state-of-the-art algorithms in similarity measures, clus- tering and classification methods and relevance feedback technologies. Finally, the key problems in content-based image retrieval are discussed, and the promising future research directions are prospected.

关 键 词：基于内容的图像检索 特征提取 相似度测量 聚类与分类 相关反馈 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]