World Journal of Science and Technology | www.worldjournalofscience.

com | 2011 | 1(8): 32-36

World Journal of Science and Technology 2011, 1(8): 32-36
ISSN: 2231 2587
www.worldjournalofscience.com


DETECTION OF FORGERIES IN DIGITAL COLOR IMAGE
Nair S. Rajlaxmi

and Vijaya C
E&CE Department, S.D.M.College of Engineering and Technology,
Dharwad, Karnataka, India
Corresponding author e-mail:
1
nairrajlaxmi@yahoo.co.in,
2
vijayac26@yahoo.com


Abstract
The digital forgeries are made possible with the advent of low-cost, high-resolution digital cameras and
sophisticated editing softwares. Digital image forgeries are created by copy-pasting a portion of an image onto some
other image. Most digital cameras, employ a single sensor in conjunction with a color filter array (CFA) and then
interpolate the missing color samples to obtain a three channel color image. The resampling operation changes
certain characteristics of the pasted portion, which when detected serves as a clue of tampering. One of the passive
methods is to estimate the probability map (EM algorithm) and take its Fourier Transform Magnitude. In
comparison with this method, we describe two simpler and faster methods to detect the resampling in the color
images. Both the methods are in the frequency domain.

Keywords: Digital Forgery, CFA Interpolation, Resampling, Coloring Image, Transform domain.
Introduction

An image may be defined as a two-
dimensional function, f(x,y) where x and y are the
spatial or plane coordinates and the amplitude of f at
any pair of coordinates (x,y) is called the intensity or
gray level of the image at that point. A digital image is
obtained by spatial domain sampling and quantization
of samples in amplitude domain [1,2]. The unique
stature of photographs as a definitive recording of
events is being diminished due, in part, to the ease
with which digital images can be manipulated and
altered [3]. The advent of low-cost and high-resolution
digital cameras and sophisticated editing softwares
like Adobe Photoshop, GIMP, PaintShop Pro lead to
digital forgeries. In this context, it becomes extremely
important to validate the originality of digital images
[4]. Digital image forgery is the process of altering the
original photographic images by performing some
geometric transformations like resizing, rotating,
scaling, etc. An image is original, when it is acquired
directly from the digital camera and is not changed in
its resolution and size. A forged image is one which is
deliberately altered in its content. The portion of the
image which has been pasted from some other image
is called alien portion [4].

Methods for Forgery Detection

The digital information revolution and issues
concerned with multimedia security have generated
two approaches for forgery detection-active and
passive or blind approaches. The area of active
methods simply can be divided into the data hiding
approach and the digital signature approach. A major
drawback of watermarks is that they must be inserted
either at the time of recording the image, or later by a
person authorized to do so [5].
In contrast to these approaches, passive
techniques for image forensics operate in the
absence of any digital signature or watermark. The
set of image forensics tools can be grouped into five
categories:
(i)Pixel-based techniques that detect statistical
anomalies introduced at the pixel level; it includes
techniques which directly or indirectly analyzes
pixel-level correlations that arise from a specific form
of tampering.
World Journal of Science and Technology | www.worldjournalofscience.com | 2011 | 1(8): 32-36
(ii)Format-based techniques that leverage the
statistical correlations introduced by a specific lossy
compression scheme.
(iii)Camera-based techniques that exploit artifacts
introduced by the camera lens, sensor, or on-chip
post processing.
(iv)Physics based techniques that explicitly model and
detect anomalies in the three-dimensional interaction
between physical objects, light and the camera.
(v)Geometric-based techniques that make
measurements of objects in the world and their
positions relative to the camera. In this paper, we
focus on blind methods using the detection of
inconsistencies in color filter array interpolation.

CFA Interpolation
A digital color image consists of three
channels containing samples from different bands of
the color spectrum, e.g., red, green, and blue. Many
digital cameras are equipped with a single charge
coupled device (CCD) or complementary metal oxide
semiconductor (CMOS) sensor (mainly due to the
cost considerations). These sensors are
monochromatic. The color images are obtained in
conjunction with a color filter array and the most often
used filter is called Bayer filter which gives
information about the intensity of light in red, green,
and blue wavelength regions. The red and



Fig. 1. The top-left portion of a CFA image
obtained from a Bayer array. The red, r2i+1;2j+1 , and
blue, b2i;2j , pixels are sampled on rectilinear lattices,
while the green, g2i+1;2j and g2i;2j+1, pixels are sampled
twice as often on a quincunx lattice. Notice that at
each pixel location only a single color sample is
recorded.
Blue pixels are sampled on rectilinear
lattices, while the green pixels are sampled on a
quincunx lattice, Fig. 1. Since only a single color
sample is recorded at each pixel location, the other
two color samples must be estimated from the
neighboring samples in order to obtain a three-
channel color image. Let S(x, y) denote the CFA
image in Fig. 1, and ) , (
~
y x R , ) , (
~
y x G and
) , (
~
y x B denote the red, green, and blue channels
constructed from S(x; y) as follows:

=
=
otherwise
y x ifS y x S
y x R
r y x
, 0
) , ( ), , (
) , (
,
~
(1)

=
=
otherwise
y x ifS y x S
y x G
g
y x
, 0
) , ( ), , (
) , (
,
~
(2)

=
=
otherwise
y x ifS y x S
y x B
b y x
, 0
) , ( ), , (
) , (
,
~
, (3)

where (x,y) span an integer lattice. A complete
color image, with channels R(x,y), G(x,y), and B(x,y)
needs to be estimated. These channels take on the
non-zero values of ) , (
~
y x R , ) , (
~
y x G and
) , (
~
y x B and replace the zeros with estimates from
neighboring samples. The estimation of the missing
color samples is referred to as CFA interpolation or
demosaicking. We will briefly describe two CFA
interpolation techniques for the Bayer array that are
used in our studies.

Bilinear Interpolation
Bicubic Interpolation

The simplest methods for demosaicking are
kernel-based interpolation methods that act on each
channel independently. These methods can be
efficiently implemented as linear filtering operations
on each color channel:


= =
=
N
N v
r
N
N u
v y u x R v u h y x R ) , ( ) , ( ) , (
~
(4)

= =
=
N
N v
g
N
N u
v y u x G v u h y x G ) , ( ) , ( ) , (
~
(5)

= =
=
N
N v
b
N
N u
v y u x B v u h y x B ) , ( ) , ( ) , (
~
(6)
World Journal of Science and Technology | www.worldjournalofscience.com | 2011 | 1(8): 32-36
where ) , (
~
y x R , ) , (
~
y x G and
) , (
~
y x B are defined in Equations (1)-(3), and hr(.),
hg(.), hb(.) are linear filters of size (2N + 1) x (2N + 1).
Different forms of interpolation (bilinear and
bicubic) differ in the form of the interpolation fillter
used. For the Bayer array, the bilinear and bicubic
filters for the red and blue channels are separable
[5,6].
Bilinear Interpolation
The 1-D Bilinear Interpolation filter is given by:
] 2 / 1 1 2 / 1 [ =
l
h (7)
and its 2-D filter is given by
l
T
l
h h . The bilinear filters
for the green channel, take the form:

(
(
(

=
0 1 0
1 4 1
0 1 0
4
1
l
h
(8)
Bicubic Interpolation
The 1-D Bicubic Interpolation filter is given by:
] 16 / 1 0 16 / 9 1 16 / 9 0 16 / 1 [ =
c
h (9)
and its 2-D filter is given by
c
T
c
h h . The bicubic filters
for the green channel, take the form:


(
(
(
(
(
(
(
(
(

=
0 0 0 1 0 0 0
0 0 9 0 9 0 0
0 9 0 81 0 9 0
1 0 81 256 81 0 1
0 9 0 81 0 9 0
0 0 9 0 9 0 0
0 0 0 1 0 0 0
4
1
c
h
(10)

Proposed Method

In this section, we present two techniques in the
frequency domain.

Forgery detection using DFT followed by High
Pass Filtering
Whenever an image undergoes forgery, the
spectra of the image get periodically repeated. In
order to retain only the original spectra and remove all
the other copies of it, the image is low pass filtered. If
a resampled portion is pasted on another image there
is an inconsistency in the high frequency content of
the over all image. The high pass filtering will bring
out the differences. A straight-forward technique to
illustrate this is to compute the 8 x 8 block DFT of the
entire image, retain only a few high frequency
coefficients, and then reconstruct back the image.
The 2D DFT for 8x8 block is given by,

v j
i j
u j
e e j i p v u F
4
7
0
7
0
4
) , ( ) , (


= =

= (11)
The presence of high frequencies are clearly
made out in the regions that have undergone
forgeries, compared to the non-resampled regions in
the image.
Forgery detection using DCT followed by High
Pass Filtering
The image is divided into 8x8 image
subblocks and DCT is applied, instead of FT, for each
subblock individually. Hence, we simplify the general
2D-DCT in terms of 8x8 sizes. The equation for 2D
8x8 DCT is given as:

4
C(u)C(v)
v) F(u, =

)
16
1)v (2j
cos(
7
0 i
7
0 j
)
16
1)u (2i
j)cos( p(i,
+

=

=
+
(12)

where
p(i,j) = pixel level at the location (i,j)
F(u,v) = DCT coefficient at the frequency indices (u,v)
The 2D ideal High Pass Filter is defined as,

>

=
cutoff v u F
cutoff v u F
v u H
) , ( , 1
) , ( , 0
) , ( (13)

The magnitude of the DFT and DCT clearly
marks out the alien portion, which has been pasted
on it.

Simulations and Results

We have considered an 256 x 256 color
image for testing. Shown in Fig.2 (i) is a RGB forged
color image and (ii)is only red matrix forged color
image, both using the bilinear interpolation, (iii) is a
RGB forged color image and (iv)is only red matrix
forged color image, both using the bicubic
interpolation. The stool in the forged image is cropped
World Journal of Science and Technology | www.worldjournalofscience.com | 2011 | 1(8): 32-36
from another image and resampled by a factor of 3
and then pasted to the original image. We present the
results of applying the techniques presented in
section IV on this image.








Fig 2. (i) Forged color image (RGB),
(ii)Forged color image(Red) with bilinear
interpolation, (iii) Forged color
image(RGB) and (iv)Forged color
image(Red) with bicubic interpolation

With Bilinear interpolation, Fig.3 shows the
results of applying DFT and DCT magnitude on each
8x8 block on the entire color image, follwed by high
pass filtering. We notice that forged color images
shows the presence of high frequency components at
the forged portions.

Bilinear HPF





DFT







DCT

Fig 3. Results of (i)applying 8x8 block
DFT and DCT and (ii) followed by HPF

With Bicubic Interpolation, Fig.4 shows the
results of applying DFT and DCT magnitude on each
8x8 block of the entire image and (ii) show the results
with HPF. It can be noticed that the forged portions
gets highlighted.

Bicubic HPF



DFT








DCT


Fig 4. Results of (i)applying 8x8 block
DFT and DCTand (ii) followed by HPF

Fig5, shows the results of Bilinear and
Bicubic Interpolation on the forged image in Fig.2(ii).
It can be visually noticed that Red channel which is
retained, is marked with application of DFT and DCT.
The HPF is same for both the interpolation
algorithms.

Bilinear Interpolation Bicubic Interpolation

HPF



DFT




DCT

Fig 5. Results of applying 8x8 block DFT
and DCT followed by HPF on the forged
color images

Conclusion

The passive digital image forgery detection,
aims at verifying the authenticity of digital images
without any digital signature on the original images.
Most of the digital cameras employ a single sensor
with a color filter array (CFA), in order to obtain a
World Journal of Science and Technology | www.worldjournalofscience.com | 2011 | 1(8): 32-36
three channel color image, the missing samples are
interpolated. The specific correlations are introduced
by interpolation that are destroyed when tampering
with an image. Therefore, the presence or lack of
correlations produced by CFA interpolation can be
used to authenticate an image, or detect it as a
forgery.
We have shown, for two different CFA
interpolation algorithms and have presented two
techniques to detect inconsistencies in images. The
forged images are created by applying geometric
transformations that introduces resampling in forged
portions. The presence of high frequency will
distinguish the regions that have undergone
resampling. It can be noticed that the results for
Bilinear and Bicubic interpolation holds the same.
In comparison with EM algorithm, we
describe a simpler and faster method to detect the
resamples by directly taking the magnitude of the
DFT and DCT of the 8x8 block of the entire images,
by skipping the step of estimating the probability map.
It is based on the fact that the dynamic range of
frequencies is more in the forged image due to
resampling than in the original image. The images are
scaled for clear view. It can be visually observed that
with the DFT, better results are obtained in
comparison with DCT. The weakness of our methods
is that it is applicable only for images with almost
plain background. This method may be employed in
security check points wherein fast response is
essential. Results are presented for variety of forgery
images.







References

1. Rafael C. Gonzalez, Richard E. Woods,Digital
Image Processing, 2nd ed., Pearson Education
(Singapore) Pte. Ltd.

2. Uday Kumar. S and Vijaya. C, Digital Signal
Processing, Elite Publishers.

3. Alin C. Popescu and H. Farid,(2005).
Exposing digital forgeries by detecting traces of
re-sampling, IEEE Transactions on Signal
Processing, 53(2):758767.

4. Prasad and K. R. Ramakrishnan, (2006). On
resampling detection and its application to image
tampering, in Proc. IEEE International
Conference on Multimedia and Exposition, pages
13251328, Toronto, Canada.

5. A. Popescu and H. Farid,(2005).Exposing
digital forgeries in color filter array interpolated
images, IEEE Transactions on Signal
Processing, 53(10):39483959.

6. Babak Mahdian and Stanislav Saic. Blind
methods for detecting image fakery

7. Anil K. Jain, Fundamentals of Digital Image
Processing.

8. S.V.Narasimhan, S. Veena, (2008). Signal
Processing, Principles and Implementation,
Revised Edition.

9. P.P.Vaidyanathan, Multirate Systems and Filter
Banks, Prentice Hall, Inc.