|Efficient Object Recognition Based Image Annotation (Pixtract)|
With the widespread use of digital cameras, cheaper and cheaper storage devices and the increasing digitalisation of art collections and library archives the number of digital images is constantly rising. Along with this fact the desire for a fast retrieval of relevant documents at a later time is also emerging. This requires efficient search strategies and indexing techniques on the one hand and metadata enrichment of the images on the other hand. Because of the huge amount of documents the manual annotation of images is impossible. In recent years several new promising automatic object recognition concepts were proposed for this reason which are more or less generic oriented.
The objective of this project is to implement an automatic content-based image annotation. The concrete aim is to achieve a translation from image contents to textual descriptions. Therefore the feature-based search is cut off from the text-based search. The former is used to create image annotations in combination with new object recognition concepts, which efficiency has to be improved through data organisation, index structures and access paths. Upon the assigned annotations a traditional text-based search can be implemented with established text indexing methods. The application of multimedia data-mining has to be analysed too for dealing with these massive amounts of data. The crucial point is to design a management structure that allows future extensions or restructuring and also lists sufficient restrictions for efficient annotation of images based on their content.
Prof. i. R. Dr. Klaus Meyer-Wegener
Annotation; Image Description; Image Retrieval; Search
Duration: 1.1.2007 - 31.8.2012
|Nagy, Robert ; Meyer-Wegener, Klaus: Towards Extensible Automatic Image Annotation with the Bag-of-Words Approach. In: Huet, Benoit ; Chua, Tat-Seng ; Hauptmann, Alexander (Ed.) : Proceedings of the International Workshop on Very-Large-Scale Multimedia Corpus, Mining and Retrieval (ACM MM VLS-MCMR 2010 Firenze, Italy 25-29th October 2010). Vol. 1, 1. Edition New York, NY, USA : ACM, 2010, pp 43-48. - ISBN 978-1-4503-0166-4|
|Nagy, Robert ; Dicker, Anders ; Meyer-Wegener, Klaus: Definition and Evaluation of the NEOCR Dataset for Natural-Image Text Recognition. Erlangen : Friedrich-Alexander-Universität Erlangen-Nürnberg. 2011 (CS-2011-07). - Internal report. 31 pages (Department Informatik Technical Reports) ISSN 2191-5008|
|Nagy, Robert ; Dicker, Anders ; Meyer-Wegener, Klaus: NEOCR: A Configurable Dataset for Natural Image Text Recognition. In: Iwamura, Masakazu ; Shafait, Faisal (Ed.) : Camera-Based Document Analysis and Recognition (CBDAR, 4th Int. Workshop in conjunction with ICDAR 2011 Beijing, China 22.09.2011). 2011, pp 53-58. |