There are currently two interface types for searching and browsing large image collections: keyword-based image retrieval (KBIR), and content-based image retrieval (CBIR). The KBIR system searches images according to the text of keyword annotated on images. This method is simple and relative effective to the query user, however, manpower-costly and likely to result in semantic gap. Because of the inherently complicated characteristics of image information, the CBIR system is of extremely high computing performance demand and hardly scalable due to the centralized architecture of traditional Client/Server network. We argue process of CBIR and similarity measurement of images, propose an index model in image database for CBIR system, and then present a hybrid peer-to-peer (P2P) network model of CBIR system based on 2-layered infrastructure in which the advantages of both distributed computing and centralized managing are integrated. We finally implement a prototype of this CBIR system on the JXTA framework.
[1]
Anil K. Jain,et al.
Image classification for content-based indexing
,
2001,
IEEE Trans. Image Process..
[2]
Rohini K. Srihari,et al.
Intelligent Indexing and Semantic Retrieval of Multimodal Documents
,
2004,
Information Retrieval.
[3]
David Novak,et al.
On scalability of the similarity search in the world of peers
,
2006,
InfoScale '06.
[4]
Ramesh C. Jain.
Visual information management
,
1997,
CACM.
[5]
Pavel Zezula,et al.
A Content-Addressable Network for Similarity Search in Metric Spaces
,
2005,
DBISP2P.