Publication detail

Automatic Hair Color De-identification

PŘINOSIL, J. KŘUPKA, A. ŘÍHA, K. DUTTA, M. SINGH, A.

Original Title

Automatic Hair Color De-identification

English Title

Automatic Hair Color De-identification

Type

conference paper

Language

en

Original Abstract

A process of de-identification used for privacy protection in multimedia content should be applied not only for primary biometric traits (face, voice) but for soft biometric traits as well. This paper deals with a proposal of the automatic hair color de-identification method working with video records with frontal facial view. The proposed method allows mutual conversion of hair color between five basic colors (black, brown, blonde, red, grey/white). The method involves image hair area segmentation, which is based on motion estimation, face and facial features detecting and skin segmentation. Further basic hair color recognition is performed to estimate appropriate algorithm for modification of the hair color achieving real-looking de-identified images.

English abstract

A process of de-identification used for privacy protection in multimedia content should be applied not only for primary biometric traits (face, voice) but for soft biometric traits as well. This paper deals with a proposal of the automatic hair color de-identification method working with video records with frontal facial view. The proposed method allows mutual conversion of hair color between five basic colors (black, brown, blonde, red, grey/white). The method involves image hair area segmentation, which is based on motion estimation, face and facial features detecting and skin segmentation. Further basic hair color recognition is performed to estimate appropriate algorithm for modification of the hair color achieving real-looking de-identified images.

Keywords

soft biometrics; hair detection; hair color recognition; de-identification

RIV year

2015

Released

08.10.2015

ISBN

978-1-4673-7909-0

Book

ICGCIoT - International CConference of Green Computing and Internet of Things 2015

Pages from

732

Pages to

736

Pages count

5

URL

Documents

BibTex


@inproceedings{BUT117888,
  author="Jiří {Přinosil} and Aleš {Křupka} and Kamil {Říha} and Malay Kishore {Dutta} and Anushikha {Singh}",
  title="Automatic Hair Color De-identification",
  annote="A process of de-identification used for privacy protection in multimedia content should be applied not only for primary biometric traits (face, voice) but for soft biometric traits as well. This paper deals with a proposal of the automatic hair color de-identification method working with video records with frontal facial view. The proposed method allows mutual conversion of hair color between five basic colors (black, brown, blonde, red, grey/white). The method involves image hair area segmentation, which is based on motion estimation, face and facial features detecting and skin segmentation. Further basic hair color recognition is performed to estimate appropriate algorithm for modification of the hair color achieving real-looking de-identified images.",
  booktitle="ICGCIoT - International CConference of Green Computing and Internet of Things 2015",
  chapter="117888",
  doi="10.1109/ICGCIoT.2015.7380559",
  howpublished="electronic, physical medium",
  year="2015",
  month="october",
  pages="732--736",
  type="conference paper"
}