A novel illumination normalization method based on local relation map

This paper presents a novel illumination normalization method to address the issue of illumination invariant face recognition. The proposed method applies a Difference of Gaussians (DoG) filter in the logarithm domain of the images to reduce the effects caused by the shadows. After that, a local rel...

Full description

Saved in:
Bibliographic Details
Main Authors: Lian, Zhichao, Er, Meng Joo, Li, Juekun
Other Authors: School of Electrical and Electronic Engineering
Format: Conference or Workshop Item
Language:English
Published: 2013
Subjects:
Online Access:https://hdl.handle.net/10356/101917
http://hdl.handle.net/10220/12788
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-101917
record_format dspace
spelling sg-ntu-dr.10356-1019172020-03-07T13:24:50Z A novel illumination normalization method based on local relation map Lian, Zhichao Er, Meng Joo Li, Juekun School of Electrical and Electronic Engineering IEEE Conference on Industrial Electronics and Applications (7th : 2012 : Singapore) DRNTU::Engineering::Electrical and electronic engineering This paper presents a novel illumination normalization method to address the issue of illumination invariant face recognition. The proposed method applies a Difference of Gaussians (DoG) filter in the logarithm domain of the images to reduce the effects caused by the shadows. After that, a local relation map (LRM) is extracted as illumination invariant features for further recognition task. The proposed method outperforms the existing normalization approaches significantly based on the experimental results in the Yale B and Extended Yale B database. Moreover, the proposed method does not involve any prior information or modeling step and takes a low computational loan. Therefore it can be easily implemented in a real-time face recognition system. 2013-08-01T04:32:04Z 2019-12-06T20:46:37Z 2013-08-01T04:32:04Z 2019-12-06T20:46:37Z 2011 2011 Conference Paper https://hdl.handle.net/10356/101917 http://hdl.handle.net/10220/12788 10.1109/ICIEA.2012.6360731 en
institution Nanyang Technological University
building NTU Library
country Singapore
collection DR-NTU
language English
topic DRNTU::Engineering::Electrical and electronic engineering
spellingShingle DRNTU::Engineering::Electrical and electronic engineering
Lian, Zhichao
Er, Meng Joo
Li, Juekun
A novel illumination normalization method based on local relation map
description This paper presents a novel illumination normalization method to address the issue of illumination invariant face recognition. The proposed method applies a Difference of Gaussians (DoG) filter in the logarithm domain of the images to reduce the effects caused by the shadows. After that, a local relation map (LRM) is extracted as illumination invariant features for further recognition task. The proposed method outperforms the existing normalization approaches significantly based on the experimental results in the Yale B and Extended Yale B database. Moreover, the proposed method does not involve any prior information or modeling step and takes a low computational loan. Therefore it can be easily implemented in a real-time face recognition system.
author2 School of Electrical and Electronic Engineering
author_facet School of Electrical and Electronic Engineering
Lian, Zhichao
Er, Meng Joo
Li, Juekun
format Conference or Workshop Item
author Lian, Zhichao
Er, Meng Joo
Li, Juekun
author_sort Lian, Zhichao
title A novel illumination normalization method based on local relation map
title_short A novel illumination normalization method based on local relation map
title_full A novel illumination normalization method based on local relation map
title_fullStr A novel illumination normalization method based on local relation map
title_full_unstemmed A novel illumination normalization method based on local relation map
title_sort novel illumination normalization method based on local relation map
publishDate 2013
url https://hdl.handle.net/10356/101917
http://hdl.handle.net/10220/12788
_version_ 1681046415676014592