NICI > Projects > Eidetic: Intelligent content-based image retrieval
|
|
Project: Eidetic: Intelligent content-based image retrieval
The proposed project, called Eidetic, will be carried out in close cooperation with three other Universities: IPO/TUE, ICT/TUD and IKAT/UM. The main goal of the project is to combine insights in pattern recognition and cognitive ergonomics to yield important progress in content-based image retrieval systems. For this purpose, the current image retrieval system will be used as a basis to combine three different IR techniques: outline-based, feature-based and object-based image retrieval. Furthermore, user-studies will be performed to assess the usability and effectiveness of the system to be (re-)developed.
Main objectives of the project are:
- investigate what constitute the salient objects in image material
- improve on outline-based matching techniques
- develop techniques for the automated detection of outlines
- combine the three image retrieval techniques
- develop the UI for steering the combination of outline-based,feature-based and object-based retrieval and perform usability studies on this UI
- compare two methods of query-by-example techniques, i.e. using thumbnails and using a hierarchical organization of the available image material
- develop a system for describing the layout of objects in an image and use this as a combined query
Project finished.
People that were involved this project
- Dr. E.L. van den Broek
- Dr. L.G. Vuurpijl
Last modification: 10-04-2006