The project is supported by Monash Faculty of IT Small Grants (2005 $18,000 & 2006 $19,000). This project challenges current CBIR practices where objects must be segmented from the images during the indexing process in order for them to be queried by the user.
The techniques we propose has the advantage of harnessing the most relevant information identified by the user to provide optimal retrieval results that best match these criteria. It also does not prematurely filter out important information in the image database thus has a greater potential to cater for any sub-image queries provided by the user. Moreover, it eliminates the need for effective image segmentation technique to achieve effective retrieval results in CBIR.