CS 68191 Masters Seminar / CS 89191 Doctoral Seminar
Spring 2007
Doctoral Student Presentation
Digital Image Forensics - Detecting Tampered Images
Wei-Hung (Wayne) Cheng
Authentic digital image is a challenging area for image processing
research. In the past, researchers used embedding watermarks to
tackle digital image tampering; the verification point for this method
is extracted watermarks used to claim authenticity or tampered. In
practice, most of the digital images are created without watermarks so
this method failed [1]. In one of the recent approaches, images can
be analyzed by the inconsistency of image quality such as lighting or
image source. A technique has developed for spliced image diction by
object lighting inconsistency [2]. For example, observing the
abnormality in the camera response function, detecting whether two
images came from the same camera, and using pattern noise correlation
to find the camera source of an image [3, 4, and 5] can be useful.
Another approach is based on the statistical view point. This method
involves modeling statistical properties with extracted visual
features from natural images so that it could be used to differentiate
spliced from natural images [6]. For example, used bi-coherence with
other features, used wavelet features, and used geometric features can
be used for modeling statistical properties of natural images [7, 1]
and used to differentiate the spliced digital images.
References
[1] T.-T. Ng, S,-F. Chang, J. Hsu, L. Xie, and M.-P. Tsui,
"Physics-motivated features for distinguishing photographic images and
computer graphics," in ACM Multimedia, 2005.
[2] M.K. Johnson and H.Farid, "Exposing digital forgeries by detecting
inconsistencies in Lighting," in ACM Multimedia and Security Workshop,
2005.
[3] Z. Lin, R. Wang, X. Tang and H.-Y. Shum, "Detecting doctored
images using camera response normality and consistency." In CVPR,
2005, pp. 1087-1092.
[4] M.Kharrazi, H. T. Sencar, and N. D. Memon. "Blind source camera
identification.," in ICIP, 2004, pp. 709-712.
[5] J. Luks, J. Fridrich, and M. Goljan, "Determining digital image
orgin using sensor imperfection," in SPIE, 2005, vol. 5685,
pp. 249-260.
[6] T.-T. Ng, S.-F. Chang, and Q. Sun, "Blind detection of
photomontage using higher order statistics," in ISCAS, 2004.
[7] H.Farid and S. Lyn, "Higher-order wavelet statistics and their
application to digital forensics," in IEEE Workshop on Statistical
analysis in Computer Vision, 2003.