Query by shots: retrieving meaningful events using multiple queries and rough set theory

  • Authors:
  • Kimiaki Shirahama;Kuniaki Uehara

  • Affiliations:
  • Kobe University, Kobe, Japan;Kobe University, Kobe, Japan

  • Venue:
  • Proceedings of the 9th International Workshop on Multimedia Data Mining: held in conjunction with the ACM SIGKDD 2008
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

In videos, even if events have the same semantic content (e.g. conversation, battle, chase and so on), they are presented in different ways. So, these events have significantly different properties. For example, one battle event is characterized by a large sound volume (e.g. gunshot), while another one is characterized by a large amount of motion (e.g. fighting action). Thus, events with the same semantic content cannot be internally defined by a single model. In this paper, we propose an example-based event retrieval method which uses multiple queries to externally define events with the same semantic content. Specifically, we apply rough set theory to multiple queries. Thereby, we can extract subsets of events, which are characterized by different combinations of low-level features. Then, by unifying extracted subsets, we conceptualize the semantic content. The experimental results indicate a possibility of our method based on multiple queries and rough set theory.