Distance-based Similarity Models for Content-based Multimedia Retrieval
Concomitant with the digital information age, an increasing amount of multimedia data is generated, processed, and finally stored in very large multimedia data collections. The expansion of the internet and the spread of mobile devices allow users the utilization of multimedia data everywhere. Multimedia data collections tend to grow continuously and are thus no longer manually manageable by humans. As a result, multimedia retrieval approaches that allow efficient information access to massive multimedia data collections become immensely important. These approaches support users in searching multimedia data collections in a content-based way based on a similarity model. A similarity model defines the similarity between multimedia data objects and is the core of each multimedia retrieval approach.
This thesis investigates distance-based similarity models in the scope of content-based multimedia retrieval. After an introduction to content-based multimedia retrieval, the first part deals with the fundamentals of modeling and comparing contents of multimedia data. This is complemented by an explanation of different query types and query processing approaches. A novel distance-based similarity model, namely the Signature Quadratic Form Distance, is developed in the second part of this thesis. The theoretical and empirical properties are investigated and an extension of this model to continuous feature representations is proposed. Finally, different techniques for efficient similarity query processing are studied and evaluated.
|Published in:||Dissertation, Fakultät für Mathematik, Informatik und Naturwissenschaften, RWTH Aachen University.|
Tag der mündlichen Prüfung: 16.07.2013
|Forschungsgebiet:||Exploration of Multimedia Databases|