Conclusion

Future Work

The theoretical foundations of the part indicate several opportunities for further related research, e.g.:

• •

  Development of an intuitive CBIR user interface

• •

  Support for spatial information integrated into the query language handle local objects

• •

  Integration of an efficient image segmentation algorithm

• •

  Implementation of powerful fv, such as shape matching

• •

  Focus on higher scalability, e.g. stronger use of index structures and heuristics

• •

  Application of other efficient learning techniques to build query descriptors

• •

  Use of query descriptors to replace keywords in CBIR queries or to label images with unknown content

• •

  Improvement of the Boolean ranking mechanisms, e.g. by applying Bayesian combination rules