Capture device identification from digital images using Kullback-Leibler divergence

It is proposed a forensic method for the capture device identification from digital images, which requires two elements: i) a digital image subject to controversy named disputed image and ii) a set of eligible capture devices with which the disputed image could have been shot. In order to define a d...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications Vol. 80; no. 13; pp. 19513 - 19538
Main Authors: Quintanar-Reséndiz, Ana L., Rodríguez-Santos, Francisco, Pichardo-Méndez, Josué L., Delgado-Gutiérrez, Guillermo, Ramírez, Omar Jiménez, Vázquez-Medina, Rubén
Format: Journal Article
Language:English
Published: New York Springer US 01.05.2021
Springer Nature B.V
Subjects:
Case studies
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Digital imaging
Fingerprints
Forensic sciences
Multimedia Information Systems
Nonuniformity
Probability distribution
Signal processing
Special Purpose and Application-Based Systems
Photographic cameras (07.68.+m)
Statistical mechanics (05.20.-y)
Forensic science (89.20.Mn)
Image processing algorithms (07.05.Pj)
Statistics (02.50.-r)
ISSN:1380-7501, 1573-7721
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:It is proposed a forensic method for the capture device identification from digital images, which requires two elements: i) a digital image subject to controversy named disputed image and ii) a set of eligible capture devices with which the disputed image could have been shot. In order to define a device statistical fingerprint , a set of reference digital images is produced for each eligible capture device. The device statistical fingerprint is estimated averaging the statistical distribution of the photo response non-uniformity (PRNU) signal extracted from each set of reference digital images. Then, a comparison based on Kullback-Leibler divergence (KLD) is performed between the statistical fingerprint for each capture device and the statistical distribution of the PRNU signal extracted from the disputed image . Considering that KLD is a non-symmetric measure, the capture device, for which the smallest KLD has been estimated, will be chosen such as the one that shot the disputed image. The effectiveness of the proposed method was estimated by using a case study, which includes eight eligible capture devices, each of which shot thirty reference images and twenty disputed images. Then, the performance of the proposed method was like the performance of the methods that use peak-to-correlation energy as the discrimination criterion when they were applied to the case study. Finally, the proposed method offers two advantages; it reduces the processing time when the PRNU signal is extracted from digital image and it avoids the aberration produced by the lens into the PRNU signal.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-021-10653-1