Similarity Search in Multimedia Databases

A multimedia database is a system able to store and retrieve objects made up of text, images, sounds, animations, voice, video, etc. The wide range of applications for MMDBs leads to a number of different problems with respect to traditional database systems, which only consider textual and numerical attributes. These problems include among other things the querying and retrieving of multimedia objects.

This research field addresses the similarity search in picture and 3D object databases.

• Similarity Search in 3D Databases
  
  In multimedia databases, 3D objects will be also stored in the future. This project deals with finding examples for searched shapes in large databases of 3D models of the computer graphic, such as VRML data.
  

  
  

  
  
• Similarity Search in High Dimensional Feature Vectors
  
  

  The similarity search in high dimensional spaces is an interesting and important problem which is relevant for a wide variety of novel database applications. As recent results show, however, the problem is very difficult to solve, not only regarding the performance issue but the quality for the obtained results, too. As a solution for the problem and within the scope of this research project, a new generalized notion of nearest neighbor search should be developed.
  

  

• Wavelet-based Similarity Search in Picture Databases
  
  In the search in picture databases, the importance of content-based search for similar pictures to a given one increases continuously. For an efficient execution of the similarity search, we try to extract important characteristics from the pictures in order to carry out the search with the help of these characteristics.