A system for automatic Chinese business card recognition

We present a system for automatic Chinese business card recognition. Chinese business card is characterized by its complex layout - Chinese-English mixed character sets and diverse typesetting. Our goal is to extract meaningful information item from the scanned image, such as "holder-name", "title", "address", etc. The automatic recognition task is divided into three stages: 1) document geometrical layout analysis; 2) bilingual character recognition; 3) text understanding and information extraction. The final results of this system are stored in database for later use. The whole system has been tested with card images scanned at 400 dpi and results are reported.