Copy-Move Forgery Detection Using Cellular Automata - Cellular Automata in Image Processing and Geometry - page 101

Chapter 6
Copy-Move Forgery Detection Using Cellular
Automata
Dijana Tralic, Paul L. Rosin, Xianfang Sun, and Sonja Grgic
Abstract. Thanks to the availability of many sophisticated image processing tools,
digital image forgery is prevalent nowadays. One of the common methods is copy-
move forgery (CMF), where part of an image is copied to another location in the
same image. Detection of copy-move forgery has been widely researched recently,
and many different solutions have been proposed. This chapter introduces a differ-
ent approach, in which cellular automata (CA) are applied to the task of copy-move
forgery detection (CMFD). The basic idea is to learn, for each overlapping block in
the image, a set of CA rules that represents the intensity changes within that block.
These rules are then used as features for the detection of copied blocks.
A problem arises when applying CA to image processing. If pixel intensities are
used as cell states, then the large range of image intensities leads to a combinato-
rial explosion in the number of possible rules, making it difficult to both learn and
represent rules efficiently. We describe a solution in which a reduced description of
a neighbourhood is accomplished by a proper binary representation of the image
based on local binary patterns (LBPs).
In the case of plain copy-move forgery, a simple 1D CA are sufficient for de-
tection purposes, but any transformation of the copied area (for example, rotation
and scaling) introduces large changes into the binary representation of the image,
resulting in the need for more complicated forms of the CA's neighbourhood. How-
ever, the main issue of CMFD using CA rules is its sensitivity to processing after
the copy-move operation applied to hide traces of the forgery, for example, addition
of noise. Nevertheless, in some cases the CA can effectively cope with such forg-
eries if image pre-processing (for example, simple image filtering) is applied before
forgery detection.
Dijana Tralic · Sonja Grgic
Faculty of Electrical Engineering and Computing, University of Zagreb,
Unska 3, 10000 Zagreb
e-mail: {Dijana.Tralic,Sonja.Grgic}@fer.hr
Paul L. Rosin · Xianfang Sun
School of Computer Science and Informatics, Cardiff University, Cardiff CF24 3AA
e-mail: {Paul.Rosin,Xianfang.Sun}@cs.cardiff.ac.uk