Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Filtering Methods for Similarity-Based Multimedia Retrieval

Summary: Filtering Methods for Similarity-Based
Multimedia Retrieval
Vassilis Athitsos, Jonathan Alon, Stan Sclaroff, and George Kollios
Computer Science Department, Boston University, Boston MA 02215, USA
{athitsos, jalon, sclaroff, gkollios}@cs.bu.edu
Abstract. A common problem in multimedia databases is retrieving the
most similar matches to a query object. Finding those matches can be
too slow to be practical, especially in domains where comparing multi-
media objects involves computationally expensive similarity (or distance)
measures. Filter-and-refine retrieval is a framework for addressing this
problem: the filter step quickly filters out most database objects, and the
refine step identifies the best matches among the remaining candidates.
This paper describes two filtering methods, that work by constructing ef-
ficient approximations of computationally expensive similarity measures.
The first method can be applied to arbitrary domains, and the second
method explicitly targets domains where measuring similarity includes
an alignment process. The benefits of these two filtering methods are
illustrated in experiments with databases from different domains, i.e.,
hand images, gesture videos, and online digit recognition for hand-held


Source: Athitsos, Vassilis - Department of Computer Science and Engineering, University of Texas at Arlington


Collections: Computer Technologies and Information Sciences