Current image database systems either treat images as pure byte sequences where the database system does not have any knowledge about the semantics of the image, or they are designed as specialized systems providing sophisticated imaging functionality, but no general data modeling and retrieval capabilities; none of them provides full modeling support for arbitrary pixel data types. Still more important, no system offers enough data independence to query an image in another format than exactly the byte stream as which it has been stored.
In this paper, a small set of language constructs is proposed which facilitates multidimensional discrete data (MDD) management in databases. Its special contribution is the array abstraction together with appropriate operations on arrays of arbitrary size over arbitrary data types. This allows to formulate powerful operations on images in a concise and optimizable way. A MDD storage mechanism is outlined which provides efficient access to such kind of data.
Implementation of the approach presented is currently in progress to enhance the object-oriented database system APRIL with MDD management capabilities.
Keywords multidimensional discrete data - multimedia database system - image database system - image manipulation language - data modeling - data independence - image tiling - spatial index - APRIL
This work has been performed at the Fraunhofer Computer Graphics Research Group in Darmstadt, sponsored by DFN Assoc. (German Research Net Association) under grant no. TK558-VA014