Adaptive shadow removal algorithm for face images

Zhen Zeng, Rumin Zhang, Jianwen Chen, Liaoyuan Zeng, Wenyi Wang, Sean McGrath

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

Abstract

In the real world, illumination is an inevitable factor in face recognition. It has been proved that illumination variations are more significant than inherent variations between persons. This paper proposes an adaptive image processing method, which can not only suppress the effect of light in face recognition, but also remove shadow caused by illumination. In this paper, first, adaptive illumination preprocessing is performed to make the image have appropriate brightness. Then, the shadows boundaries of the image is extracted and binarized to obtain the shadow boundaries mask. Finally, the high-quality face image without shadows is reconstructed based on the mask of shadows boundaries and the face image after the illumination preprocessing. Experiments on the CMU-PIE dataset have shown that our method can achieve both good visual effects and a significant improvement in face recognition accuracy.

Original languageEnglish
Title of host publication2018 12th International Conference on Sensing Technology, ICST 2018
PublisherIEEE Computer Society
Pages227-231
Number of pages5
ISBN (Electronic)9781538651476
DOIs
Publication statusPublished - 2 Jul 2018
Event12th International Conference on Sensing Technology, ICST 2018 - Limerick, Ireland
Duration: 4 Dec 20186 Dec 2018

Publication series

NameProceedings of the International Conference on Sensing Technology, ICST
Volume2018-December
ISSN (Print)2156-8065
ISSN (Electronic)2156-8073

Conference

Conference12th International Conference on Sensing Technology, ICST 2018
Country/TerritoryIreland
CityLimerick
Period4/12/186/12/18

Keywords

  • adaptive
  • face recognition
  • illumination
  • lighting normalization
  • shadow

Fingerprint

Dive into the research topics of 'Adaptive shadow removal algorithm for face images'. Together they form a unique fingerprint.

Cite this