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.
出处
《电路与系统学报》
CSCD
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)
关键词
图像检索
特征维
相关反馈
语义特征
image retrieval
component of feature
relevance feedback
semantic feature
