Presentation 2004/9/3
Face Image Retrieval using 3D Data with Robustness against Pose and Illumination Variations
Masahiko Hamanaka, Jose Ferreiro, Rui Ishiyama, Shizuo Sakamoto,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This report proposes a new face recognition scheme called 3D(query)-2D(enroll) recognition. It can retrieve facial images from an inputted 3D data with robustness against pose and illumination variations. When using the previously proposed Geodesic Illumination Basis method to match a 2D image and a 3D data for 2D(query)-3D(enroll) recognition, it is found that its scores are drastically changed dependent on images. This report shows the scores become stable by introducing score normalization, and not only verification rates but also identification rates can be improved. In the experiments using facial image database of 1000 individuals, the proposed method achieved a first-choice success ratio of over 90% with an HER of about 2%.
Keyword(in Japanese) (See Japanese page)
Keyword(in English)
Paper # PRMU2004-59
Date of Issue

Conference Information
Committee PRMU
Conference Date 2004/9/3(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Face Image Retrieval using 3D Data with Robustness against Pose and Illumination Variations
Sub Title (in English)
Keyword(1)
1st Author's Name Masahiko Hamanaka
1st Author's Affiliation Media and Information Research Laboratories, NEC Corporation()
2nd Author's Name Jose Ferreiro
2nd Author's Affiliation Swiss Federal Institute of Technology (EPFL)
3rd Author's Name Rui Ishiyama
3rd Author's Affiliation Media and Information Research Laboratories, NEC Corporation
4th Author's Name Shizuo Sakamoto
4th Author's Affiliation Media and Information Research Laboratories, NEC Corporation
Date 2004/9/3
Paper # PRMU2004-59
Volume (vol) vol.104
Number (no) 290
Page pp.pp.-
#Pages 8
Date of Issue