Conclusion

Most image database systems are built around a multidimensional index that contains the useful data for querying. The index becomes the kernel of the system and can never be dropped or replaced. The DISIMA prototype implements a database approach: data are manipulated through algebraic operators and indexes (if any) are used to speed up the query processing.

DISIMA is a prototype image database system enabling content-based queries. The content of images are represented by salient objects, interesting objects within an image. The single-site implementation is completed and an on-line demonstration can be found at http://db.uwaterloo.ca/ 18#18ddbms/projects/multimedia/index.html. This conference demonstration augments the on-line version with distributed features and a novel indexing facility for multi-precision color-similarity searches.