Share Email Print

Proceedings Paper

Very fast recognition of GIRO check forms
Author(s): Thien Minh Ha; Horst Bunke
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Check forms are used by many people in daily life for money remittance. Surprisingly, the processing of these forms at banks and post offices is only partly automated. In this paper, we deal with a particular kind of form, viz., the GIRO checks used in Switzerland. We describe a fully automatic system which is able to recognize the following items on a GIRO check: the financial institution, the name and address of the receiver, and the account number. The complete recognition of a GIRO check is divided into two phases. In the first phase, the system performs a layout analysis in order to localize regions corresponding to various items on the check. The input gray-level image is first binarized and segmented using the X-Y-tree decomposition algorithm resulting in a list of atomic entities (e.g., individual characters). Each entity is then interpreted as part of an item (e.g., receiver's name), according to the knowledge about possible layouts of a form. All atomic entities belonging to the same item are grouped together and yield the location of that item. In the second phase, the localized items are separately binarized again and submitted to an OCR engine to obtain streams of characters that correspond to the items. We have tested the system on a large number of checks and the results are promising in terms of both computation time and recognition accuracy.

Paper Details

Date Published: 14 April 1993
PDF: 12 pages
Proc. SPIE 1906, Character Recognition Technologies, (14 April 1993); doi: 10.1117/12.143611
Show Author Affiliations
Thien Minh Ha, Univ. of Bern (Switzerland)
Horst Bunke, Univ. of Bern (Switzerland)

Published in SPIE Proceedings Vol. 1906:
Character Recognition Technologies
Donald P. D'Amato, Editor(s)

© SPIE. Terms of Use
Back to Top