Issues in Visual Querying and Indexing for the Efficient Retrieval-by-content of Arrangements of Point Objects

Abstract

In the context of computer science, an <i>index</i> is a data structure that enables the efficient retrieval of specific items from a collection of data. An <i>image database</i> is a collection of digitized map images, with relevant objects or features represented by <i>symbols</i>. One search method for such a database is to determine the set of images that contain a desired arrangement of symbols. Examples are presented of index data structures to support the <i>position-independent search</i> of an image database, where the absolute positions of the symbols within the image are unimportant, and only the relative spatial relationships of the symbols are significant. By separating the <i>size</i>, <i>shape</i>, and <i>orientation</i> attributes of an arrangement, these index structures support efficient searching that is either size-dependent or size-independent, and either orientation-dependent or orientation-independent. The visual language of an existing retrieval-by-content image database system is extended to allow intuitive control of the additional search flexibility.