Reliable Computing 4: 103–104, 1998.
1998 Kluwer Academic Publishers. Printed in the Netherlands.
Intervals Help to Design an Imaging System
EDM International (Electronic Data Management), Juarez, Chihuhua, Mexico;
Current address: Department of Computer Science, University of Texas at El Paso, El Paso, TX
79968, USA, e-mail: email@example.com
(Received: 15 April 1997; accepted: 21 May 1997)
Abstract. We present a method for calculating the number of stations needed to process a given
number of documents in an imaging system. We use interval computations to take into consideration
variants in throughput (caused, among others, by machine malfunction, human intervention, network
response time for reading and/or writing ﬁles, and potentially other factors).
Image processing systems. An imaging system digitizes (scans) documents (hand-
written or typed) and extracts the information from the resulting electronic images.
In a typical imaging system, image processing consists of the following stages
Optical Character Recognition (OCR)
OCR Correction → Key from Image
The arrows show the ﬂow of documents through the system.
Each process is very computation-intensive; as a result, a workstation can, at
best, handle only one process. Moreover, usually, several workstations are needed
for each process.
Design of image processing systems. There exist ready software and hardware for
each process, and there exist links that connect each process to the other. The main
design problem is, therefore, to decide how many workstation we need for each
Traditional design methods. Traditional methods for choosing the number of
workstations N (see, e.g.,  and references therein) are based on the assumption
G. Gonzalez implemented this system while employed at EDM International. Conﬁdential
information property of EDM International was omitted from this paper. The simpliﬁed diagram of
the system represents a generic imaging system for forms processing, as depicted in magazines and
publications of the trade.