Interactive learning of image visual similarities and semantic categorization

被引:0
作者
Yang, ZJ [1 ]
Kuo, CCJ [1 ]
机构
[1] Univ So Calif, Dept Elect Engn Syst, Integrated Media Syst Ctr, Los Angeles, CA 90089 USA
来源
INTERNET MULTIMEDIA MANAGEMENT SYSTEMS | 2000年 / 4210卷
关键词
image database; content analysis; interactive learning; relevance feedback; image retrieval;
D O I
暂无
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
The "query by example" model has been extensively used to retrieve similar images in content-based image database management. The query is characterized by searching images with feature vectors similar to those of the example based upon either a default or a user-defined similarity metric. However, low level features often encounter a severe performance bottleneck as applied to natural image collections with complicated contents and great perceptual varieties. The feature-based similarity matching approach tends to retrieve many irrelevant images. This is not surprising since images different in semantic meanings but close enough in low level features can be returned as a pertinent result. Such a query process lacks user involvement and therefore results in a gap between features and semantics. In this work, a novel content-based image retrieval scheme that learns image visual similarities and semantic categories from relevance feedback is presented. First, we choose the most suitable low level features to describe images by analyzing image contents and categorize each image by predicting its semantic meanings. During the interactive retrieval process, users are allowed to confirm semantic classification of the query example and evaluate retrieval results with relevance feedback. By processing the feedback information, the system learns both image visual similarities and semantic meanings. In similarity learning, the retrieving results are refined by modifying the similarity metric. Semantic learning is performed by using the decision tree training algorithm. The system continues being updated along the retrieval procedure.
引用
收藏
页码:356 / 367
页数:12
相关论文
共 16 条
  • [1] AKSOY S, 1998, IEEE CVPR 98 WORKSH
  • [2] BHANU B, 1998, IEEE INT C IM PROC
  • [3] BOUET M, 1998, ACM MULTIMEDIA 98
  • [4] HUANG J, 1997, ACM MULT 97 SEATTL W
  • [5] KAPLAN LM, 1995, VISUAL DATA EXPLORAT, V2410, P405
  • [6] LIPSON P, 1997, IEEE C COMP VIS PATT
  • [7] MEDASANI S, 1998, IEEE CVPR 98 WORKSH
  • [8] Interactive learning with a ''society of models''
    Minka, TP
    Picard, RW
    [J]. PATTERN RECOGNITION, 1997, 30 (04) : 565 - 581
  • [9] Content-based image retrieval with relevance feedback in Mars
    Rui, Y
    Huang, TS
    Mehrotra, S
    [J]. INTERNATIONAL CONFERENCE ON IMAGE PROCESSING - PROCEEDINGS, VOL II, 1997, : 815 - 818
  • [10] SQUIRE DM, 1998, SPIE MULTIMEDIA STOR, V3