Two Data Organizations for Storing Symbolic Images in a Relational Database System
Aya Soffer, Hanan Samet · 1999
A method is presented for integrating images into the framework of a conventional database management system (DBMS). It is applicable to a class of images termed symbolic images in which the set of objects that may appear are known a priori. The geometric shapes of the objects are relatively primitive and they convey symbolic information. Both the pattern recognition and indexing aspects of the problem are addressed. The emphasis is on extracting both contextual and spatial information from the raw images. A logical image representation that preserves this information is defined. Methods for storing and indexing logical images as tuples in a relation are presented. Indices are constructed for both the contextual and the spatial data, thereby enabling efficient retrieval of images based on contextual as well as spatial specifications. Two different data organizations (integrated and partitioned) for storing logical images in relational tables are proposed. They differ in the way that the logical images are stored. Sample queries and execution plans to respond to these queries are described for both organizations. Analytical cost analyses of these execution plans are given. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.