Share Email Print
cover

Proceedings Paper

Image quality guided approach for adaptive modelling of biometric intra-class variations
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

The high intra-class variability of acquired biometric data can be attributed to several factors such as quality of acquisition sensor (e.g. thermal), environmental (e.g. lighting), behavioural (e.g. change face pose). Such large fuzziness of biometric data can cause a big difference between an acquired and stored biometric data that will eventually lead to reduced performance. Many systems store multiple templates in order to account for such variations in the biometric data during enrolment stage. The number and typicality of these templates are the most important factors that affect system performance than other factors. In this paper, a novel offline approach is proposed for systematic modelling of intra-class variability and typicality in biometric data by regularly selecting new templates from a set of available biometric images. Our proposed technique is a two stage algorithm whereby in the first stage image samples are clustered in terms of their image quality profile vectors, rather than their biometric feature vectors, and in the second stage a per cluster template is selected from a small number of samples in each clusters to create an ultimate template sets. These experiments have been conducted on five face image databases and their results will demonstrate the effectiveness of proposed quality guided approach.

Paper Details

Date Published: 28 April 2010
PDF: 10 pages
Proc. SPIE 7708, Mobile Multimedia/Image Processing, Security, and Applications 2010, 77080L (28 April 2010); doi: 10.1117/12.850592
Show Author Affiliations
Ali J. Abboud, Univ. of Buckingham (United Kingdom)
Sabah A. Jassim, Univ. of Buckingham (United Kingdom)


Published in SPIE Proceedings Vol. 7708:
Mobile Multimedia/Image Processing, Security, and Applications 2010
Sos S. Agaian; Sabah A. Jassim, Editor(s)

© SPIE. Terms of Use
Back to Top