一种改进的基于标准离差的相关反馈图像检索方法  

An Improved Method in Relevance Feedback-Based Image Retrieval System Using Standard Deviation

作　　者：闫玲博[1] 魏维[1] 王艳 

机构地区：[1]西安通信学院 [2]61938部队

出　　处：《微计算机应用》2008年第12期98-102,共5页Microcomputer Applications

摘　　要：MARS系统是采用相关反馈方式进行图像检索的一个范型。本文在分析该系统所采用权值调整算法基础上,考虑负例图像因素引入了一个参数,并提出了一种新的改进算法。通过在由1000幅图像构成的Corel数据库中的试验,改进后算法在大部分图像类型的检索上性能有明显提升。MARS （Multimedia Analysis Retrieval System） is a representative implementation of the content-based image retrieval system employing relevance feedback. With analysis of the strategies of weight adjustment in MARS, this paper proposes an improved methed in the development of the relevance feedback. Taking part in the influence of negative instances, a parameter due to standard deviation is introduced in the proposed method. According to the test with the Corel database of 1000 images, significant improvement has been achieved in most types of the image retrievals.

关 键 词：基于内容的图像检索(CBIR) 相关反馈 MARS系统 标准离差 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]