With computer network popularization entirely,multimedia information retrieval grows rapidly in the last few years. The QBIC system from IBM Almaden Research Center have initiated a new field in image information query. Images may be queried by color, gray scale, texture and position features. Query requiring would be represent by graphic mode, for example, selecting color out of color pallet and selecting texture out of sample images. Users can filter and select visual information rapidly and simply by this approach. We have researched Multimedia Information REtrieval System(MIRES) under supported by "863" advanced technology project.

Key Technique                                                               top

1. Content based image information retrieval using color, shape, texture and so on.

2. Chinese information full text retrieval.

3. Content establishing and acquiring.

4. Information storage management.

5. Client retrieval of platform independent .

Product Function                                                            top

    The MIRES client allows users to:(1) submit a sample image to the image server;(2) submit semantic keywords to the image server;(3) both of above two method;(4) for (1), users may setup every feature weightiness; (5) for result images set, users can use interactive feedback function.

    The server receives retrieval request from client, analyze it, select query tactic, startup retrieval engine to retrieval destination, then server returns ordinal image result to the client. Moreover, the server performs managing retrieval engine, managing image database, etc..

    MIRES has finer interface and encapsulation to other software. 

    Network mode, to say, MIRES can be distributed at several computers. 

Product speciality                                                          top

    Without client setup program but a  browser like IE supporting Java.
    Internet oriented online service.
    Multi-method combined query, interactive correlative feedback optimized query.
    MIRES adopts multi-server architecture.
    Large content image database.

To the users                                                                top

    MIRES's user are two types. The one is developer and researcher on content-based image retrieval. MIRES provides them a fine development platform base CORBA and allows them to add image retrieval algorithm conveniently.

    Client: Java,IE.
    Web server: Java,CORBA,Platform support JSP/Servlet,OS may be Windows,UNIX,Linux,etc..
    Application Server: C++ development environment,SQL Server, Windows 2000 server.

Application extend                                                         top 

(1) Law retrieval system of Beijing Library

(2) Chinese traditional medicine plant database in Science Database

(3) Retrieval system of Journal of Computer

(4) Computer term retrieval

(5) Remote sensing image retrieval

Member                                                                      top

    Hu hong    Li hui    Shi zhiping    Li qingyong

Homepage

Key Laboratory of Intelligent Information Processing 
Institute of Computing Technology
Chinese Academy of Sciences
Last modified on Oct.5,2002