A Compact and Complete AFMT Invariant with Application to Face Recognition.

Yu, H. and Bennamoun, M., 2005. A Compact and Complete AFMT Invariant with Application to Face Recognition. In: ACIDCA-ICMI'2005: Proceedings of the 2nd International Conference on Machine Intelligence 2005. ACIDCA-ICMI'2005.

Full text available as:

[img]
Preview
PDF - Published Version
516kB

Abstract

In this paper, we present a complete set of hybrid similarity invariants under the Analytical Fourier-Mellin Transform (AFMT) framework, and apply it to invariant face recognition. Because the magnitude and phase spectra are not processed separately, this invariant descriptor is complete. In order to simplify the invariant feature data for recognition and discrimination, a 2D-PCA approach is introduced into this complete invariant descriptor. The experimental results indicate that the presented invariant descriptor is complete and similarityinvariant. Its compact representation through the 2D-PCA preserves the essential structure of an object. Furthermore, we apply this compact form into ORL, Yale and BioID face databases for experimental verification, and achieve the desired results.

Item Type:Book Section
Additional Information:5-7 November 2005, Tozeur, Tunisia.
Subjects:Generalities > Computer Science and Informatics
Group:Media School > National Centre for Computer Animation
ID Code:14750
Deposited By:Mr Hongchuan Yu
Deposited On:21 May 2010 12:49
Last Modified:07 Mar 2013 15:29

Document Downloads

More statistics for this item...
Repository Staff Only -
BU Staff Only -
Help Guide - Editing Your Items in BURO