Share Email Print

Proceedings Paper

Handwritten character recognition using background analysis
Author(s): Guido Tascini; Paolo Puliti; Primo Zingaretti
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

The paper describes a low-cost handwritten character recognizer. It is constituted by three modules: the `acquisition' module, the `binarization' module, and the `core' module. The core module can be logically partitioned into six steps: character dilation, character circumscription, region and `profile' analysis, `cut' analysis, decision tree descent, and result validation. Firstly, it reduces the resolution of the binarized regions and detects the minimum rectangle (MR) which encloses the character; the MR partitions the background into regions that surround the character or are enclosed by it, and allows it to define features as `profiles' and `cuts;' a `profile' is the set of vertical or horizontal minimum distances between a side of the MR and the character itself; a `cut' is a vertical or horizontal image segment delimited by the MR. Then, the core module classifies the character by descending along the decision tree on the basis of the analysis of regions around the character, in particular of the `profiles' and `cuts,' and without using context information. Finally, it recognizes the character or reactivates the core module by analyzing validation test results. The recognizer is largely insensible to character discontinuity and is able to detect Arabic numerals and English alphabet capital letters. The recognition rate of a 32 X 32 pixel character is of about 97% after the first iteration, and of over 98% after the second iteration.

Paper Details

Date Published: 14 April 1993
PDF: 8 pages
Proc. SPIE 1906, Character Recognition Technologies, (14 April 1993); doi: 10.1117/12.143613
Show Author Affiliations
Guido Tascini, Univ. di Ancona (Italy)
Paolo Puliti, Univ. di Ancona (Italy)
Primo Zingaretti, Univ. di Ancona (Italy)

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

© SPIE. Terms of Use
Back to Top