Issue No. 01 - January/February (2003 vol. 18)
Katerina Pastra , University of Sheffield
Horacio Saggion , University of Sheffield
Yorick Wilks , University of Sheffield
<p>Current approaches in automatic text-based image indexing fail to capture semantic information that is expressed in image captions, but such indexing is important for high precision and recall in subsequent image retrieval. The indexing prototype developed in the Scene of Crime Information System project tries to go beyond keyword-based approaches and methods that extract syntactic relations from captions. The authors suggest a more effective and efficient solution: using advanced natural language processing techniques to extract relational facts consisting of a "pragmatic relation" and the entities it relates (triples of the form "argument1 RELATION argument2"). They implement this approach for indexing crime scene photographs.</p>
image indexing, image retrieval, text-based image indexing, relational fact, crime investigation technology
K. Pastra, H. Saggion and Y. Wilks, "Intelligent Indexing of Crime Scene Photographs," in IEEE Intelligent Systems, vol. 18, no. , pp. 55-61, 2003.