發信人: ashun (阿順), 信區: DataMining
標  題: MultiMediaMiner: A System Prototype for MultiMedi
發信站: 南京大學小百合站 (Sun Aug 26 17:43:48 2001)

    Osmar R.Za. ane Jiawei Han Ze-Nian Li Sonny H. Chee Jenny Y. Chiang

Intelligent Database Systems Research Laboratory and Vision and Media Laborato
ry

                        School of Computing Science

                          Simon Fraser University

                        Burnaby, BC, Canada V5A 1S6

           E-mail: fzaiane, han, li, schee, ychiangg@cs.sfu.ca


[Abstract]


Multimedia data mining is the mining of high-level multimedia information and 
knowledge from large multimedia databases. A multimedia data mining system pro
totype,MultiMediaMiner , has been designed and developed. It includes the cons
truction of a multimedia data cube which facilitates multiple dimensional anal
ysis of multimedia data,primarily based on visual content, and the mining of m
ultiple kinds of knowledge, including summarization, comparison, classication,
 association , andclustering.


1 [Introduction]


There has been rapid progress in the eld of data mining and data warehousing r
esearch, but nothing substantial in mining multimedia. Many data mining and

data warehousing systems have been developed for mining knowledge in relationa
l databases and data warehouses [4]. Multimedia has been the major focus for


many researchers around the world. Many techniques for representing, storing, 
indexing, and retrieving multimedia data have been proposed. However, rare are
 the researchers who ventured in the multimedia data mining eld. Most of the s
tudies done are conned to the data ltering step of the KDD process. 


In [3], Czyzewski shows how KDD methods can be used to analyze audio data and 
remove noise from old recordings. Chien et al. in [2] use knowledge-based AI t
echniques to assist image processing in a large image database generated from


the Galileo mission. Others use Multimedia to complement data mining systems. 
Bhandari et al., for instance,Research is supported in part by the Natural Sci
ences and Engineering Research Council of Canada, the Canadian Network

of Centres of Excellence (IRIS:HMI-5 and TL:NCE5.2), and MPR Teltech Ltd.

marries a data mining application with multimedia resources. His application d
oes not claim to mine a multimedia database, but uses video clips to support t
he knowledge discovered from a numerical database [1].


Multimedia data mining is a subfield of data mining that deals with the extrac
tion of implicit knowledge,multimedia data relationships, or other patterns no
t explicitly stored in multimedia databases.Recent advances in the research on
multimedia databases [7, 8, 5] enable creation of large multimedia

databases which can be queried in an eective way.


These advances, in combination with the research into multimedia database and 
advances in data mining in relational databases [4], created a possibility for
 the creation of multimedia data mining systems.The current MultiMediaMiner sy
stem includes four data mining modules for mining knowledge in image

and video databases: characterization, comparison,classication, and associatio
n. Additional modules are in the design and development stage.

A more detailed description of the MultiMediaMiner system is presented in Sect
ion 2. A summary and adiscussion of our on-going research are in Section 3.
--
業精于勤荒于嬉，行成于思毀于隨。 —— 韓愈
臨淵羨魚不如退而結網。           —— 班固
勿以惡小而為之，勿以善小而不為。 —— 劉備

※ 來源:．南京大學小百合站 http://bbs.nju.edu.cn [FROM: 202.119.80.20]