Authors: Anna Bartkowiak
Addresses: University of Wroclaw, Institute of Computer Science, Joliot Curie 15, 50–383 Wroclaw, Poland; Wroclaw High School of Applied Informatics, Wejherowska 28, 54–239 Wroclaw, Poland
Abstract: We look for outliers in graphical data containing n = 6977 faces or non-faces images from Seung|s collection. Our concern is: what kind of outliers may be found in such graphical data. To obtain the global geometrical characteristics, the Pseudo Grand Tour and Kohonens|s self-organising maps are applied. We define as outliers those images which reproduce themselves badly from K principal components, with K denoting intrinsic dimension. The concept of mild and severe outliers, and own and alien principal components is also introduced.
Keywords: biometrics; faces; non-faces; outliers; anomaly; images; distributional properties; intrinsic dimensionality; PCA; principal component analysis; reconstruction; lower dimensions; graphical visualisation; pseudo grand tour; self-organising maps.
International Journal of Biometrics, 2010 Vol.2 No.3, pp.203 - 221
Published online: 01 Jun 2010 *Full-text access for editors Access for subscribers Purchase this article Comment on this article