Indexing by recognition approach for printed Arabic documents images

摘要

We propose in this paper an indexing by recognition approach applied to Arabic printed documents images archives. Indeed, our goal is to restore the set of the images susceptible to contain the key word or all the key words introduced by the user as a request. To fulfill this goal, we started with the creation of the prototypes of Arabic printed characters taken in their different shapes; thereafter we have implemented a module of analysis of the user's textual request as well as a procedure of synthesis of sub words models. Finally, we developed a connected components detection procedure which constitutes the preprocessing phase of the recognition stage.