In 1994, the U.S. Postal Service (USPS) delivered 177 billion pieces o
f mail to 125 million delivery addresses. To assist in processing this
large volume of mail, USPS and the Center for Document Analysis and R
ecognition (CEDAR) have been exploring automation methods for processi
ng letter mail with handwritten addresses. The CEDAR system for locati
ng and reading destination address blocks on letter mail pieces operat
es at real-time rates. The system architecture uses a variety of pipel
ining and multiprocessor techniques to achieve the required processing
speed acid minimize latency. Image processing operations are performe
d using specialized hardware operating at 103 ns/pixel. This hardware
includes CCITT Group 3 (FAX) decompression, image enhancement, and con
version of the image representation from pixel-based representation to
symbolic representation. This symbolic representation of the image ca
n be processed in real time using software running on general-purpose
processors. This software performs the address block location (ABL), h
andwritten address interpretation (HWAI), and data base querying neede
d to determine the delivery point of the mail piece. The current syste
m can correctly locate and produce a five-digit ZIP code with an accur
acy of 66% on a test set of 1566 handwritten mail pieces. A unique del
ivery point (e.g., ZIP+4 code) can be determined on these same mail pi
eces 26% of the time. (C) 1996 John Wiley & Sons, Inc.