An open Architecture for enhancing performance of complex OCR applications

Abstract

Taking advantages of the different existing computing environments, infrastructures and resources for running in the optimal and/or in a customized manner any given complex Optical Character Recognition (OCR) software or application constitutes nowadays a challenge. Indeed, we mean by a complex OCR software or application any document digitization and computerization process which includes either plenty heterogeneous documents to process as input and/or several strong and sometimes complex OCR techniques to use in order to achieve good accuracy (recognition) rates. Moreover, the diversity and the very high computing and storage powers provided by such computing environments and infrastructures in one hand and the lack of powerful software and tools allowing their optimal or good utilization in the other hand make the problem a challenge.

 

Consequently, this paper proposes a novel open architecture, which attempts to use properly such environments and infrastructures in order to run at least in a pseudo optimal and/or a customized manner any given complex OCR software or application. Actually, the two most important features, which make the proposed architecture original, are first, its complete independency from the different existing distributed infrastructures that can run any given complex OCR application. Second, its flexibility, which allows any new distributed infrastructure to be considered during the scheduling process of any given complex OCR application since the scheduler, is able to detect automatically and consider any added infrastructure. Our architecture presents several advantages, indeed, it improves drastically the performances of any given complex OCR application, it is platform and software independent in addition to its flexibility as described and explained later.