Image Processing Reference
In-Depth Information
6.1
Introduction
Nowadays analogue images are completely replaced by digital images thanks to
the simplicity of their acquisition, processing and storage. Many sophisticated dig-
ital image processing tools allow modification (tampering/manipulation) of digital
image content to be done without any visible traces, leading to many forged im-
ages used in everyday life. Different approaches for revealing the manipulation of
the content of digital image have been developed recently, with the aim to ensure
the credibility of digital images. However, there is still no methodology to verify the
integrity of digital images in an automatic manner. Therefore, digital image foren-
sics [7] is an emerging research field that includes methods for determining the
authenticity and the origin of digital images.
Image authentication methods can generally be classified into two main cate-
gories [6]: active and passive methods. Active methods involve embedding of some
information into an image in the archiving process, such as digital watermarks [13]
and signatures. Tampering of images usually destroys or modifies that embedded in-
formation, so any kind of content manipulation can be easily detected by analyzing
information extracted from image. One active approach for digital image forensics
is described in Chapter 7, where cellular automata are used as a tool for generating
digital signatures.
Passive methods on the other hand involve checking the integrity of an image by
analysis of image statistics and properties (for example, sensor noise [8], illumina-
tion conditions [10], etc.). Detection methods in this category are usually aimed at
solving some special kind of tampering attempts so there is no unique technique for
detection of all types of forgeries.
One common type of forgery attack is a copy-move forgery (CMF) [20] in which
part of an image is copied and moved to a new location. CMF is often used because
of the simplicity of performing that type of forgery. Detection of CMF can be done
using different techniques, but the most basic method is based on dividing the image
into overlapping blocks and calculating some set of features from every block. Those
features are then used to determine similarity between blocks in image. Different
feature sets were proposed for this purpose, such as average or intensity [14] or
frequency coefficients [9], etc.
In this chapter, a new approach for copy-move forgery detection based on cellu-
lar automata (CA) is presented [23]. The basic idea is to use cellular automata to
learn a set of rules for each sub-image block in an image. Those rules can serve as
features for determining the similarity of blocks. A problem that arises when us-
ing CA with grayscale images is the large number of possible rules, which leads
to an even larger number of possible combinations of those rules. A reduced rep-
resentation can be accomplished by binary representation of images in a proper
manner using techniques based on local binary patterns (LBP's) [16]. The problem
of plain copy-move detection can be solved by applying simple 1D CA, but detec-
tion of different types of CMFs (for example, rotation or scaling of copied area) re-
quires defining more complicated types of neighbours for CA. Also, post-processing
Search WWH ::
Custom Search