Share Email Print

Proceedings Paper

Biometric watermarking based on face recognition
Author(s): Takami Satonaka
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

We describe biometric watermarking procedure based on object recognition for accurate facial signature authentication. An adaptive metric learning algorithm incorporating watermark and facial signatures is introduced to separate an arbitrary pattern of unknown intruder classes from that of known true-user ones. The verification rule of multiple signatures is formulated to map a facial signature pattern in the overlapping classes to a separable disjoint one. The watermark signature, which is uniquely assigned to each face image, reduces the uncertainty of modeling missing facial signature patterns of the unknown intruder classes. The adaptive metric learning algorithm proposed improves a recognition error rate from 2.4% to 0.07% using the ORL database, which is better than previously reported numbers using the Karhunen-Loeve transform, convolution network and the hidden Marcov model. The face recognition facilitates generation and distribution of the watermark key. The watermarking approach focuses on using salient facial features to make watermark signatures robust to various attacks and transformation. The coarse-to-fine approach is presented to integrate pyramidal face detection, geometry analysis and face segmentation for watermarking. We conclude with an assessment of the strength and weakness of the chosen approach as well as possible improvements of the biometric watermarking system.

Paper Details

Date Published: 29 April 2002
PDF: 11 pages
Proc. SPIE 4675, Security and Watermarking of Multimedia Contents IV, (29 April 2002); doi: 10.1117/12.465325
Show Author Affiliations
Takami Satonaka, Fraunhofer Ctr. for Research in Computer Graphics and Univ. of Rhode Island Rhode Island (United States)

Published in SPIE Proceedings Vol. 4675:
Security and Watermarking of Multimedia Contents IV
Edward J. Delp III; Ping Wah Wong, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?