Home // MMEDIA 2012, The Fourth International Conferences on Advances in Multimedia // View article

Video Retrieval by Managing Uncertainty in Concept Detection using Dempster-Shafer Theory

Authors:
Kimiaki Shirahama
Kenji Kumabuchi
Kuniaki Uehara

Keywords: Video retrieval; Concept Detection; Uncertainty; Dempster-Shafer Theory; Evidential EM Algorithm

Abstract:
This paper focuses on concept-based video retrieval which examines whether a shot is relevant or irrelevant to a query based on detection results of concepts, like Person, Building and Car. One key problem is uncertainty in concept detection. Even for state-of-the-art methods, it is difficult to accurately detect concepts present in a shot. Relying on such uncertain concept detection results degrades retrieval performance. To overcome this problem, Dempster-Shafer Theory (DST) is used to represent the probability that a concept is possibly present in a shot. By incorporating DST into maximum likelihood estimation, our method estimates the probabilistic distribution of concepts' presences which characterize shots relevant to the query. A preliminary experiment on TRECVID 2009 video data supports the effectiveness of our method.

Pages: 71 to 74

Copyright: Copyright (c) IARIA, 2012

Publication date: April 29, 2012

Published in: conference

ISSN: 2308-4448

ISBN: 978-1-61208-195-3

Location: Chamonix, France

Dates: from April 29, 2012 to May 4, 2012