Please use this identifier to cite or link to this item:
Title: Linking face images captured from the optical phenomenon in the wild for forensic science
Authors: Das, Abhijit
Sengupta, Abira
Ferrer, Miguel A. 
Pal, Umapada
Blumenstein, Michael
UNESCO Clasification: 3307 Tecnología electrónica
Keywords: Face recognition
Image capture
Issue Date: 2018
Journal: IEEE International Joint Conference on Biometrics, IJCB 2017
Conference: 2017 IEEE International Joint Conference on Biometrics, IJCB 2017 
Abstract: This paper discusses the possibility of use of some challenging face images scenario captured from optical phenomenon in the wild for forensic purpose towards individual identification. Occluded and under cover face images in surveillance scenario can be collected from its reflection on a surrounding glass or on a smooth wall that is under the coverage of the surveillance camera and such scenario of face images can be linked for forensic purposes. Another similar scenario that can also be used for forensic is the face images of an individual standing behind a transparent glass wall. To investigate the capability of these images for personal identification this study is conducted. This work investigated different types of features employed in the literature to establish individual identification by such degraded face images. Among them, local region based featured worked best. To achieve higher accuracy and better facial features face image were cropped manually along its close bounding box and noise removal was performed (reflection, etc.). In order to experiment we have developed a database considering the above mentioned scenario, which will be publicly available for academic research. Initial investigation substantiates the possibility of using such face images for forensic purpose.
ISBN: 9781538611241
DOI: 10.1109/BTAS.2017.8272770
Source: IEEE International Joint Conference on Biometrics, IJCB 2017,v. 2018-January, p. 781-786
Appears in Collections:Actas de congresos
Show full item record

Page view(s)

checked on Nov 29, 2020

Google ScholarTM




Export metadata

Items in accedaCRIS are protected by copyright, with all rights reserved, unless otherwise indicated.