Indexing pictorial documents by their content: a survey of current techniques

To efficiently solve the image retrieval problem in an image database, many techniques have been devised addressing the requirements of different applications. Present information management systems exploit these techniques to allow querying of pictorial documents archives. To do this, the first step is to extract relevant keys from the image content itself, so that images are not treated as illustrative appendices of textual documents. Even more important, the search process is not restricted to external attributes such as the storing date or the archive number. The kind of characteristics used to build the database index depends on the users' information needs for a given cognitive environment. Currently, the focus is on image color statistics and patterns (textures), on shapes and their spatial relations, or on a combination of these. This paper reviews indexing methods that are significant examples of techniques presently used.