Organizing Multimedia Information with Maps
Semantic multimedia organization is an open challenge. In this chapter, we present an innovative way of automatically organizing multimedia information to facilitate content-based browsing. It is based on self-organizing maps. The visualization capabiliti
- PDF / 1,256,029 Bytes
- 17 Pages / 595 x 842 pts (A4) Page_size
- 76 Downloads / 198 Views
2 3
LIP6, Universit´e Pierre et Marie Curie – CNRS, Paris, France, [email protected], [email protected] IRISA, Universit´e de Rennes 1, Rennes, France [email protected] IWS, Otto-von-Guericke Universit¨ at, Magdeburg, Germany [email protected]
Summary. Semantic multimedia organization is an open challenge. In this chapter, we present an innovative way of automatically organizing multimedia information to facilitate content-based browsing. It is based on self-organizing maps. The visualization capabilities of the self-organizing map provide an intuitive way of representing the distribution of data as well as the object similarities. The main idea is to visualize similar documents spatially close to each other, while the distance between different documents is bigger. We demonstrate this on the particular case of video information. One key concept is the disregard of the temporal aspect during the clustering. We introduce a novel time bar visualization that reprojects the temporal information. The combination of innovative visualization and interaction methods allows efficient exploration of relevant information in multimedia content.
1 Introduction A huge and ever increasing amount of digital information is created each day. The capacity of the existing manifold storage devices (for instance hard drives, optical disks, flash memories) increases continuously. Multimedia information in digital formats is, on the one hand, found everywhere in our everyday life, in devices such as portable media players, mobile phones, digital cameras. Thus, we already rely on the assistance of desktop search engines like Google Desktop, Beagle, or Spotlight for finding locally stored data. On the other hand, the amount of publicly available information and its boost is even more impressive. Apart from classical media, the recent web 2.0 trend [1] of sharing user-created content is a major contributor. The blog scene as well as community websites like Flickr [2], MySpace [3], or YouTube [4] constantly continue to grow both in terms of users and the sheer amount of data. Facing this amazing amount of information, it has become extremely difficult and time consuming to filter and retrieve the relevant pieces. T. B¨ arecke et al.: Organizing Multimedia Information with Maps, Studies in Computational Intelligence (SCI) 96, 493–509 (2008) c Springer-Verlag Berlin Heidelberg 2008 www.springerlink.com
494
T. B¨ arecke et al.
A big challenge when dealing with multimedia is usually referred to as the Semantic Gap. It arises from the fact that there is a difference between the technical representation and the actual meaning of a given multimedia document. In other words, we cannot index multimedia information like numerical since there is no unique, well-defined semantic for a given document. Ideally, multimedia retrieval should be based on the meaning, but unfortunately, a computer is not able to identify it. Multimedia retrieval systems [5] that provide satisfying interaction possibilities for all types of multimedia in
Data Loading...