Perspective projection for variance pose face recognition from camera calibration

M. M. Fakhir, W. L. Woo, J. A. Chambers, S. S. Dlay

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

Variance pose is an important research topic in face recognition. The alteration of distance parameters across variance pose face features is a challenging. We provide a solution for this problem using perspective projection for variance pose face recognition. Our method infers intrinsic camera parameters of the image which enable the projection of the image plane into 3D. After this, face box tracking and centre of eyes detection can be identified using our novel technique to verify the virtual face feature measurements. The coordinate system of the perspective projection for face tracking allows the holistic dimensions for the face to be fixed in different orientations. The training of frontal images and the rest of the poses on FERET database determine the distance from the centre of eyes to the corner of box face. The recognition system compares the gallery of images against different poses. The system initially utilises information on position of both eyes then focuses principally on closest eye in order to gather data with greater reliability. Differentiation between the distances and position of the right and left eyes is a unique feature of our work with our algorithm outperforming other state of the art algorithms thus enabling stable measurement in variance pose for each individual.

Original languageEnglish
Title of host publicationOptics, Photonics and Digital Technologies for Imaging Applications IV
EditorsPeter Schelkens, Touradj Ebrahimi, Gabriel Cristóbal, Frédéric Truchetet, Pasi Saarikko
PublisherSPIE
Volume9896
ISBN (Electronic)9781510601413
DOIs
Publication statusPublished - 29 Apr 2016
EventOptics, Photonics and Digital Technologies for Imaging Applications IV - Brussels, Belgium
Duration: 5 Apr 20166 Apr 2016

Publication series

NameProceedings of SPIE
PublisherSPIE
Volume9896
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceOptics, Photonics and Digital Technologies for Imaging Applications IV
CountryBelgium
CityBrussels
Period5/04/166/04/16

Fingerprint

Dive into the research topics of 'Perspective projection for variance pose face recognition from camera calibration'. Together they form a unique fingerprint.

Cite this