一种自适应提取最优特征维的相关反馈算法  

A Relevance Feedback Algorithm Using Adaptive Optimal Feature Component Extraction

作　　者：朱旭娟[1] 李晓华[1] 沈兰荪[1] 

机构地区：[1]北京工业大学信号与信息处理研究室,北京100022

出　　处：《电路与系统学报》2004年第1期36-40,共5页Journal of Circuits and Systems

基　　金：国家自然科学基金资助项目(60172045)

摘　　要：本文提出一种新的相关反馈算法,该算法依据用户的反馈信息自适应选取用户最感兴趣的特征维用于图像检索,并结合正负反馈图像集的预处理,图像检索精确度得到较大提高。算法在500幅和4500幅两个图像库中做了实验,通过与Rui Y特征内相关反馈算法的比较,验证了算法的高效性。Relevance feedback is an important technique in Content-based Image Retrieval (CBIR) developed in the past few years. A new relevance feedback algorithm is proposed in this paper. Based on users preference feedback, features in which the user is especially interested will be chosen as the attributes in image retrieval. In the aid of positive and negative feedback preprocessing, the precision of image retrieval will be greatly improved. Experiments have been done based on two image banks containing 500 and 4500 images respectively. Comparison concerning experimental retrieval results with Rui Ys feature intra-relative feedback algorithm proves the efficiency of t proposed algorithm higher than other algorithms.

关 键 词：图像检索 特征维 相关反馈 语义特征 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]