Share Email Print

Proceedings Paper

Automatic signature verification based on the wavelet descriptor
Author(s): Zijun Yang; C.-C. Jay Kuo
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

To extract relevant features hidden in signatures is an important step in signature verification system. A `good' feature makes it easier and more efficient to discriminate forgery signatures from genuine ones. In this research, a new technique employing the wavelet feature for automatic signature verification is proposed. Signatures can be characterized by their position information either in the time domain with a graphic tablet or in the space domain with the off-line tracing algorithm. An efficient tracing and representation algorithm for off-line signatures is considered here. First, the dynamic position function of static signatures can be recovered by a new tracing algorithm. Then, we develop a hierarchical feature extraction method to represent signatures by using the planar wavelet descriptor. The wavelet descriptor has many nice properties such as multiresolution, invariance, uniqueness, stability, and spatial localization. With the wavelet descriptor, stable, efficient and representative features are extracted from signatures. It has been demonstrated with experimental results that features described by the wavelet descriptor have a better performance than Fourier descriptors in the signature verification application. Finally, we perform extensive system evaluation, and show that the wavelet descriptor technique provides significant advantages in feature extraction and information reduction for signature verification.

Paper Details

Date Published: 1 October 1998
PDF: 12 pages
Proc. SPIE 3460, Applications of Digital Image Processing XXI, (1 October 1998); doi: 10.1117/12.323217
Show Author Affiliations
Zijun Yang, Univ. of Southern California (United States)
C.-C. Jay Kuo, Univ. of Southern California (United States)

Published in SPIE Proceedings Vol. 3460:
Applications of Digital Image Processing XXI
Andrew G. Tescher, Editor(s)

© SPIE. Terms of Use
Back to Top