Research Journal of Applied Sciences, Engineering and Technology 7(13): 2612-2621, 2014
ISSN: 2040-7459; e-ISSN: 2040-7467
© Maxwell Scientific Organization, 2013
Submitted: May 05, 2013
Accepted: August 23, 2013
Published: April 05, 2014
Fuzzy Watermarking for Colour Images based on Fusion of Discrete Wavelet Transform
(DWT) Technique and Histogram Stretching
1
Karrar Abdulameer Kadhim, 1Dzulkifli Mohamad, 2Tanzila Saba and 2Fatima Nayer
1
Faculty of Computing, University Technology Malaysia, Johor, Malaysia
2
College of Computer and Information Sciences, Prince Sultan University, Riyadh, KSA
Abstract: This study presents a method of scheming colour image by using fuzzy watermarking through Discrete
Wavelet Transform (DWT) and histogram stretching methods. This study aims further at achieving and developing
robustness, imperceptibility and available capacity of cover image for the watermarked image, which can withstand
against various attacks. Digital watermarking techniques are used to protect the copyrights of multimedia data by
embedding secret information in the host media, for example, embedding in images, audios, or videos. Many
watermarking techniques have been proposed in the literature to solve the copyright violation problems, but most of
these techniques failed to satisfy both imperceptibility and robustness requirements. In this research, adaptive RGB
colour image fuzzy watermarking technique is proposed. The host image is converted from RGB to YCbCr colour
space to preserve imperceptibility and robustness, then, Cb component is extracted and partitioned into four
quadrants. While the Histogram Stretching will be used to convert the Watermark image to 4-bits plan to preserve
the capacity for watermarking.
Keywords: 4-bits plan, Discrete Wavelet Transform (DWT) technique, fuzzy watermarking, histogram stretching,
RGB colour image
INTRODUCTION
There is no doubt that using the internet has
become ubiquitous in the 21st century and it is quite true
that everyone feels it is an indispensable part for the
future of business communication. Going by the
systematic digital data attainment and its conversion for
the contemporary purpose is quite simple. Hence,
copyright is an unavoidable issue to protect intellectual
contributions (Rehman and Saba, 2014; Neamah et al.,
2014). Tsai et al. (2000) and Wu (2002) highlighted
recently that, through analyzing the trend of studies in
digital watermark for audio, image or video data; it
could be seen that watermarking techniques are the only
way to prove legal ownership.
Watermarking algorithms can be classified based
on the domain used for watermark embedding. Studies
have shown that two popular techniques; spatial and
transform watermarking techniques exist. Spatial
domain watermarking techniques are useful for higher
data embedding applications (Muhsin et al., 2014,
Rehman et al., 2013). Transform domain watermarking
techniques are suitable in applications where robustness
is of prime concern. These techniques as proposed
include, Discrete Cosine Transform (Chandramouli
et al., 2001) Discrete Fourier Transform (Premaratne,
1999) Discrete Wavelet Transform; Discrete Hadamard
Transform (Anthony et al., 2002); Contourlet
Transform (Jayalakshmi et al., 2006) and Singular
Value Decomposition (Mohan and Kumar, 2008) are
some of the useful transformations for image
processing applications.
With the introduction of the JPEG2000 standard
digital image, watermarking schemes that are derived
from Discrete Wavelet Transform (DWT) are becoming
a robust area attracting lots of attention. Nearly all
watermarking schemes consider Discrete Cosine
Transform known as DCT method of preference. A
summary are available in Cox et al. (2008). Though,
study result shows that DWT has the potentials of
enhancing the strength of watermark against intentional
and unintended attacks. The primary reason would be
that the former JPEG standard depended on DCT and
today using the creation of JPEG 2000, schemes
according to DWT are widely attaining interest.
Though, watermark robustness varies using the
underlying changed algorithms’, provisions must
automatically get to harden a watermark against attacks.
The methods are, e.g. multiple embedding Nasir et al.
(2008) and the use of Error Correction Codes (ECC) are
being used to restore the embedded watermark.
DWT is basically a recent method used in place of
an image but in a new time and frequency scale. The
basic function of DWT is decomposing the input signal
to multi-resolutions. “DWT can be used to decompose
input signal that poses as image into low frequency
Corresponding Author: Karrar Abdulameer Kadhim, Faculty of Computing, University Technology Malaysia, Johor, Malaysia
2612
Res. J. Appl. Sci. Eng. Technol., 7(13): 2612-2621, 2014
(LL) and high frequency (HL, LH and HH). HL here
means the horizontal detail, while LH stands for the
vertical detail and HH for the diagonal part. The lowest
frequency band which serves as the optimal
approximation of the original image, is influenced using
the DWT decomposition progressions technique which
represent the maximum scale and distinguishing
degree” (Santhi et al., 2008). The main aim of this
study work is to come up with a proposed method of
scheming color image by using watermark through
Discrete Wavelet Transform (DWT) and histogram
stretching methods. This study aims further at achieving
and developing robustness, imperceptibility and
available capacity of cover image for the watermarked
color image, which can withstand against various
attacks (Rehman and Saba, 2012; Elarbi-Boudihir et al.,
2011).
PROBLEM STATEMENT
Immediate digital image watermarking is certainly
an urgent requirement for several today’s programs
such as digital cameras and smart phone cameras.
Evaluating two watermarking techniques, transform
domain watermarking approaches require greater
computational complexity over spatial domain
techniques (Kougianos and Mohanty, 2011). This could
further be as a result of forward and inverse changes in
the transform-domain watermarking approaches.
However, it is common to understand that, there are
difficulties with spatial-domain techniques. For
instance, high embedding errors in ISB bit-planes result
in many researchers employing low-order bit-planes for
instance LSB for data hiding (Maity et al., 2007).
However, the lower-order bit-planes techniques does
not contain visually significant information so, the
PROBLEM BACKGROUND
embedded watermark may be simply corrupted or
transformed by unauthorized clients without affecting
It is generally believed that, many studies centers
on visual effects. Abolghasemi et al. (2010) for instance
on the development of watermarking schemes for
recommended a technique using co-occurrence matrix
grayscale images than color images (Liu and Chou,
and bit-plane clipping that could find out the hidden
2007; Saba and Rehman, 2012). Consequently,
data in LSB.
available records show that, information hiding has
Studies have revealed the existence of several
been an important research area in recent years.
diversities of attacks against watermarking methods.
However, the techniques to help address the issue of
For example Basu et al. (2010), experimental
unauthorized copying, tampering and multimedia data
investigations showed that watermarking approaches in
delivery through the internet require urgent attention.
the past were prone to several kinds of malicious
Information hiding techniques currently involves
attacks. Consequently, in order to identify the
merely the steganography and digital watermarking.
ownership of the digital media, they had to be made
Currently,
digital
watermarking
schemes
unavailable to extract the embedded watermark (s).
concerning the data taken into account through getting
Additionally, several attacks against watermarking
rid of might be categorized as blind and non-blind
schemes may be too complex to model (Cox et al.,
approaches. In non-blind watermarking approaches,
2008). Consequently, a universal watermarking
both data for actual host image and understanding
approach that could withstand several kinds of attacks
statistics about watermarked image are known inside
and, concurrently, satisfies the conventional as well as
the amount of watermark recognition and extraction
the embedding capacity needs getting a minimal
(Tao and Eskicioglu, 2004). In contrast, in blind
complexity isn't discovered yet. In this case,
approach finding the watermark and not mention for the
approximation approaches may be used to have the
original image is preferred (Al-Otum and Samara,
ability to discover the possession in the attacked
2010). You find several difficulties regarding the blind
watermarked image with low computational
watermarking approaches. On one hand, high
complexity. The best way to develop a solution that
effectiveness of blind watermarking may also be
could withstand against various kinds of attacks is
proven. Therefore, a completely new technique referred
lacking the knowledge of their exact actions.
to as semi-blind watermarking was introduced. Within
Ibrahim and Kuan (2011) employed watermarking
this kind of watermarking approach, only the original
technique using DWT and encryption canny edges, so
watermark or perhaps the watermarked multimedia
as to include an watermark image include the cover
statistics are known (Tao and Eskicioglu, 2004; Shieh
image. Through the use of this method researcher was
and Athaudage, 2006).
able to get acceptable results for PSNR and NCC after
An issue surrounding using the internet today is
shedding a number of potential attacks. The main
users "stealing" other individual’s images and taking
requirements for watermarking are: Imperceptibility,
advantage of them on the web site without permission.
Robustness and capacity. Managed researcher attention
It's impossible to prevent someone from installing
imperceptibility and robustness, but neglected capacity,
images out of your web pages. If you are an artist,
so were some of the extracted watermark pictures after
you'll most likely wish to safeguard your images by
the attack prone to damage because the size of
watermarking them.
watermark image. Therefore, it is necessary to attention
2613 Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
Fig. 1: Embbedding Stage
imperceptiibility and robustness and capaccity
simultaneoously to get th
he best results extracted afterr a
number off potential attaccks.
ensure that the embeedding processs takes into acccount
the main goals of waatermarking, (ii.e., imperceptiibility,
robustnness and capaciity).
Requantization imaage: One of the most impportant
factors in informationn hiding is thee size for the hidden
h
The reesearch tactic of the proposeed watermarkiing
data, which
w
is an acctive factor in the issue of hiding
h
method cann be explained
d in Fig. 1. Thhis approach is to
capacitty and PSNR
R ratio. Reduciing the size of
o the
2614 RESERCH METHODOL
M
OGY
Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
(1)
(a)
(b)
(c)
Fig. 2: Imagge quantization, a: Is the dark im
mage, b: Is the ligght
imagge, c: Is the quan
ntized image
(a)
(b))
Fig. 3: Lenaa image has beeen converted, a: 0-255 range, b: 015 raange
hidden leaads to increassing the efficciency of hidiing
technique (Bauschke, 2003). Because the daata
compressioon is not the main issue off our work, daata
compressioon will not be used in details or as a
professionaal stage. Imaage requantizzation or imaage
normalizattion is generallly used for imaage enhancemeent
by histograam stretching.
In imaage, pixel valu
ues vary from (0-255),
(
while in
some otheer images thesse values are condensed inn a
partial rangge of original range (0-255). Sometimes, the
t
pixel valuees are near zeero value which means imaage
will be in the dark view
w. While imagee view is in ligght
are if pixell values are neaar (255) valuess.
The im
mage requantizzation is used to
t restretch pixxel
values intto the origin
nal range (0--255). Figure 2
illustrate im
mage quantizattion where a: is the dark imagge,
b: is the ligght image and c:
c is the quantiized image.
The prrocess of requaantization can be performed by
next equattion” Histograam Stretching”” (González and
a
Woods, 20008):
n obtained value after strettching,
where, I new is the new
mage pixel value before strettching,
I old iss the original im
Range is the new range that im
mage value will
w be
M is the maxiimum and minnimum
convertted to, Max, Min
values pixel before stretching,
s
startt is the new sttarting
values, which is in ouur case equal too zero.
mage stretchingg is used essenntially for convverting
Im
values from a range to
t another. Thiis research atteends to
utilize image equaliization in ourr work in oppposite
mannerr, i.e. convertinng the values frrom (0-255) raange to
(0-15) range. Valuess of this rangee will need onnly (4)
bits forr representatioon, as shown in Fig. 3, Illuustrate
Lena im
mage has been converted intoo two Ranges which,
w
a: 0-255 range, b: 0-115 range. The 0-15 range is a dark
u it in our work
w
which will
w be
image, which will use
i next step (Im
mage Compression).
useful in
Image compression:: From the passt stage, imagee pixel
values are convertedd from (0-2555) range into (0-15)
range. Thus, convertted values neeed only (4) bitt from
each byyte to represennt them, whichh means a hallf byte
will bee empty for alll image bytess. In this case, each
bytes will
w combinedd into one dim
mensional arraay that
leads too get the half siize for the sam
me image.
Affter the stage for
fo apply histoggram equalizattion to
changee the level for image from (0--255) to (0-15)) level,
as show
wn previously,, now will com
mpress the imaage by
changee the array for this image to one dimennsional
array to
t be ensure compress
c
the image to a half
h
of
originaal image. As shhown in Fig. 4.
CbCr color spaace: The reseaarchers
Converrt RGB to YC
are using YCbCr color
c
space recently
r
becauuse of
h
YCbCrr color space is more robuust and has higher
imperceptibility than RGB color sppace (Khalili, 2003).
2
mponent and CbCr
Y reprresents the luuminance com
represeent blue andd red chromiinance compoonents
respecttively. In otherr hand RGB coolor space reprresents
Fig. 4: Com
mpress image by change to one diimensional arrayy
2615 Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
(a)
Fig. 6: Extracting
E
Y, Cbb, and Cr components
invisibiility to HVS, Fig.
F 6 shows extracting
e
Y, Cb
C and
Cr com
mponents indiividually. Thee purpose of image
partitiooning is findinng the best part
p
for embeedding
waterm
mark image, beefore applyingg DWT on thee part.
One off these parts will
w be chosen for embeddingg. The
part whhich has the highest
h
impercceptibility reprresents
the beest part for embedding thhe watermark. The
experieences for the proposed
p
methhod prove the Cb is
the best part for em
mbedding to be ensuring higher
h
imperceptibility. Figuure 6 illustrate extract Cb partt.
(b)
Fig. 5: RGB
B colour to YCbC
Cr colour image
red, greenn and blue ch
hannels respecctively. Figuree 5
shows connverting host image from RGB
R
to YCbbCr
color space. The main difference
d
betw
ween YCbCr and
a
RGB colorr space is that YCbCr color space represennts
the color as brightness with two moore color signals
which are Cb
C and Cr, wh
hile RGB represents the colorr as
red, green and blue. Nex
xt Eq. 2 showss transformatioons
between RGB
R
color space and YCbCr color
c
space (Chhai
and Bouzerdoum, 2000):
c
image:: Discrete Wavelet
W
Apply DWT for cover
transforrm (DWT) is a mathematical tool for
hierarchhically decompposing an imagge (Hong and Hang,
2006). The watermarkk image is connverted into levvel (0mensional arraay, as
15) annd converted into one dim
mentionned previouslyy, will read thhe cover imagge and
apply DWT
D
algorithm
m to split the image to four bands
to mainntain the imperrceptibility. Thhe select quadrrant of
host im
mage decompposed by usinng discrete wavelet
w
transforrm (DWT), DWT
D
producedd the frequencyy subbands is
i LL, HL, LH and HH bands. The HH bannd will
be seleected to embeedding the waatermark imagge. As
shown in Fig. 7 that illustrates how
w to apply DW
WT on
the hosst image.
dded waterm
mark image in cover im
mage:
Embed
Embeddding is hiding secret image or private labeel or a
video or
o a specific teext inside the cover
c
image without
w
significcant distortionss on the coverr image. Embeedding
providees a sure wayy to determinee the owner of
o the
image, one of the most
m
importannt characteristtics of
(
(2)
m
Embeddding is impercceptibility. Theere is lots of method
to em
mbed; the reesearch will use
u one of them to
C componen
nts: Partitioninng is dividingg a
Extract Cb
embeddding the waterrmark image inn the cover image as
particular image into multiple
m
parts; in general, the
t
will desscribe that.
Affter converting cover image innto four Bandss using
purpose of
o image partitioning is changing the
t
DWT algorithm
a
and selection
s
the HH
H band as shoown in
representattion of image to be more eaase in processiing
Fig. 8 and dependinng on the sizee of the wateermark
and analyyzing. In prroposed technique, the Cb
C
image to ensure impperceptibility annd robustness. After
componentt extracted fro
om host imagge to preserve thhe
2616 Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
Fig. 7: Applly DWT on the host
h image
Fig. 8: Recoonstruction YCbCr
convertingg image wateermark to leevel (0-15) and
a
compressed it to a half from
f
original size
s
by changee it
to one dim
mensional arraay from 0 and 1 as previoussly
mentioned, the next steep is embeddiing a watermaark
image insidde the cover im
mage in HH bannd, equation 3,, is
the formulla for embeddiing which are WI is the outpput
of the watermarking systtem, H be the host image, b is
an approppriate scaling factor (Sulonng et al., 20112;
Hajisami et
e al., 2011; Sin
ngh et al., 20099):
WI = H + b * W
(
(3)
E
Stage
Fig. 9: Extract
space to get final watermarked image form
m. The
followiing Fig. 8 illusttrates reconstruuction process..
Extracct stage: Extrraction stage is to separatte the
waterm
mark image of the cover imaage without caausing
any siggnificant impacct on the waterrmark image and
a the
cover image.
i
The folllowing schem
me illustrates how
h
to
extract the watermarkk image from cover
c
image (Fiig. 9).
where,
WI = Wattermarked imag
ge
H = The value of host image
i
b = The scale factor (0
0.005)
W = The watermark im
mage (Binary)
Wavelet reconstruction
r
n (IDWT): To return the imaage
back from the frequency
y domain to thee spatial domaain,
Inverse Diiscrete Waveleet Transform (IDWT)
(
applieed.
All new frequency
f
sub-band coefficiients of selectted
quadrant will
w be reconstrructed using ID
DWT. Both hiigh
and low paass filter will inverse selecteed quadrant form
frequency to the spattial domain with the sam
me
decomposiition level. Th
his process will
w generate one
o
watermarkked quadrant.
b component:: In this stage, after
Apply DWT for Cb
mage will connvert RGB coolor to
readingg the cover im
YCbCrr like do it in embedding staage and then extract
e
Cb andd apply DWT to
t extract wateermark image pieces
as will shown that inn details. Applyy DWT algoritthm to
split thhe image to foour bands. Thee select quadrrant of
host im
mage decompoosed by usingg Discrete Wavelet
W
Transfoorm (DWT), DWT
D
producedd the frequencyy subbands is
i LL, HL, LH
H and HH bannds. Then seleect the
HH subb-band to get watermark
w
piecces.
Reconstru
uction YCbCrr: Finally, the quadrants
q
will be
combined again to get YCbCr
Y
color space watermarkked
image, theen YCbCr willl be converteed to RGB collor
s
illustratees how
Extracct watermark image: This stage
to extrract the pieces of watermark image from thhe HH
2617 Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
sub-band of
o cover image. After read the
t watermarkked
image andd apply DWT to division intto four sub-baand
then select the HH sub
b-band, will get
g the pieces of
ng et al., 2012;; Hajisami et al.,
a
watermarkk image (Sulon
2011; Singgh et al., 2009) by the followiing formula 6.
W = (W
WI – H)/b
(44)
where, m n is the im
mage size of watermark
w
andd host
image, I (i, j) is the Host
H
image piixel value, K (i,
( j) is
the Waatermark imagee pixel value, MSE
M is Mean Square
S
Error foor (I, K).
Robusttness measurring: To meaasure the sim
milarity
betweeen original watermark
w
im
mage and exttracted
waterm
mark correlationn used as the foollowing Eq.7:
where,
W = The Watermark im
mage (Binary)
WI = The Watermarked image
H = The Value of Hostt image
b = The Scale factor (0
0.005)
(7)
Inverse convert
c
imag
ge: After get the pieces of
watermarkk image from the
t HH2 sub-bband will inverrse
the requanntization to co
onvert the leveel for watermaark
image from
m (0-15) levell to original leevel (0-255) and
a
ensure to enhance
e
the im
mage after convvert it. Equatioons
5, illustratte the convertt watermark im
mage to (0-2555)
level (Gonnzález and Woo
ods, 2008):
(
(5)
where,
Xold
Xnew
R
Start
v
before coonvert level
= The original value
= The current vaalue, after convvert the level
= The current raange (15)
= The byte will be start for thee image (0)
r
n (IDWT): To return the imaage
Wavelet reconstruction
back from the frequency
y domain to thee spatial domaain,
(
applieed.
Inverse Diiscrete Waveleet Transform (IDWT)
All new frequency
f
sub-band coefficiients of selectted
quadrant will
w be reconstrructed using ID
DWT. Both hiigh
and low paass filter will inverse
i
selecteed quadrant form
frequency to the spattial domain with the sam
me
his process will
w generate one
o
decomposiition level. Th
watermarkked quadrant.
uring: Peak Signal
S
to Noise
Impercepttibility measu
Ratio (PSN
NR) used to measure the quality betweeen
original and
a
watermark
ked image, inncreasing PSN
NR
related to the quality of
o reconstructted watermarkked
P
formulaa:
image. Thee following is PSNR
(
(6)
W i j are the pixel values at the
where, W i j and W'
positionn (i, j) of the original and the exttracted
waterm
mark by that 1 ≤ (i, j) ≤ 32, resspectively.
RESULT
TS AND DISCU
USSION
In this section results are presennted of the prooposed
waterm
marking based on Discrete Wavelet Trannsform
(DWT)). The evaluatioon of these resuults presented in this
chapterr by using sttandard imagees which are Lena,
Baboonn, pepper, Tiffaany and Airplaane as a host im
mages.
Thhe results will get
g it from embbedding a wateermark
image in
i standard RG
GB colour imaages which are Lena,
Baboonn, pepper, Tiffa
fany and Airplaane images. Thhe host
image are transform
med into frequuency domain using
Discrette Wavelet Traansform (DWT
T), DWT resullted to
four suub-bands frequencies which are
a LL, HL, LH
L and
HH. Onne watermark image has beeen embedded inn subband (H
HH) of host im
mages.
Thhe scale factoor (b) is useed to increasse the
robustnness of waterm
marking and (bb = 0.005) hass been
chosen for it. The sizze of the host images
i
is (5122×512)
byte annd the size for watermark image
i
is (128×
×128).
The cooming figures illustrate the steps of embeedding
result implementation
i
n before the attacks for seelected
images.
Thhere are a lot of
o researchers prefer use YC
CbCr ,
becausee the YCbCrr give more robust and higher
h
imperceptibility in embedding
e
stage from RGB
B color
space (Khalili,
(
2003)), Y is the lum
minescent compponent
represeents and CbCrr represents coolor componentts blue
and reed, respectiveely. While RGB color space
represeents the Red, Green
G
and Blue,, respectively.
Apply DWT on CB components: The high freqquency
componnents are usuaally used for watermarking since
the hum
man eye is leess sensitive to changes in edges
(Kashyyap and Sinhaa, 2012). We will
w apply DW
WT to
split the Cb componeent to four sub--band which arre LL,
HL, LH
H and HH bandds, then chose the
t HH sub-baand for
imbeddding the waterm
mark image. The
T Fig. 10 illuustrate
the resuult after apply DWT
D
on Cb paart for Lena im
mage.
Thhere are a lot of researcherss prefer use YCbCr,
Y
becausee the YCbCrr give more robust and higher
h
2618 Res. J. Apppl. Sci. Eng. Technol.,
T
7(13)): 2612-2621, 2014
2
To be all the byte for new imagge use four bit and
useless four as we shhown in Fig. 11 illustrate the result
after chhange the leveel to (0-15) thee image will be
b like
dark.
Affter requantizattion watermarkk image, now itt has 4
bits in each byte forr the new wattermark imagee. The
mark values intto one
techniqque will changge the waterm
dimenssional array as shown in the Fig.
F 12.
Fig. 10: Appply DWT on Cb component
Fig. 11: Reqquantization imaage
f
RGB color
imperceptiibility in embeedding stage from
space (Khaalili, 2003). Y is the luminescent componeent
represents and CbCr rep
presents color components
c
bllue
and red, respectively. While RGB color spaace
represents the Red, Green
n and Blue, resspectively.
Requantizzation and compress wattermark imagge:
One of thee most importan
nt factors in innformation hidiing
is the size for the hidden
n data, which iss an active factor
in the isssue of hiding
g capacity annd PSNR rattio.
Reducing the size of thee hidden is leaads to increasiing
the efficciency of hiding techhnique. Imaage
Requantizaation is geenerally useed for imaage
enhancemeent by histograam stretching. But in propossed
methode will
w use the histtogram stretchiing to change the
t
level for watermark
w
imag
ge from (0-2555) to (0-15) levvel,
that for com
mpress the imaage to half of the original sizze.
a
This section
s
Testingg and evaluation before attack:
will teest and evaluaate the results for the prooposed
methodd in this researrch depending on two criteriaa such
as, PSN
NR and NCC. In the begginng, the researcch will
test thhe watermarkked image which
w
results from
embeddding the waterrmark image inn Cb componeent for
the covver image depending on PSNR
R.
Ass mentioned in chapter three,, the (PSNR) used
u
to
measurre the quality between
b
originnal and waterm
marked
image, where, the increasing
i
in PSNR lead to
t the
quality of reconstructted watermarkeed image. Therrefore,
s
this reesearch will measure the PSNR for several
waterm
marked images such as, Lenna, Baboon, Pepper,
P
Tiffanyy and Airpllane. Figure 12 explainn the
waterm
marked imagees after appplying the embed
e
techniqque. Table 1 shows the PSNR
P
ratio foor the
waterm
marked images as shows in Fig. 13.
Waterm
marked imagge after attaccks: The folllowing
lines will
w explain the
t
result afteer applying various
v
attacks for three imaages which aree Lena and Baboon
B
mmon attacks are Gaussian noise,
respecttively. The com
Salt andd Peppers, etc.. The followingg sentences illuustrate
results after attack.
Fig. 12: Waatermark image in one dimensionnal array
(a)
(b)
(c)
(d)
Fig. 13: Thee watermarked images, (a) Lenaa, (b) Baboon, (c) Pepper, (d) Tifffany, (e) Airplanne
2619 (e)
Res. J. Appl. Sci. Eng. Technol., 7(13): 2612-2621, 2014
Table 1: Explain the PSNR ratio for watermarked image
Abolghasemi, M., H. Aghaeinia, K. Faez and M.A.
Mehrabi, 2010. Detection of LSB±1 steganography
Images
based on co-occurrence matrix and bit plane
Lena
clipping. J. Electro. Imag., 19(1).
Baboon
Pepper
Anthony, T.S.H., J. Shen, S.H. Tan and A.C. Kot, 2002.
Tiffany
Digital image-in-image watermarking for copyright
Airplane
protection of satellite images using the fast
hadamard transform. Proceeding of the IEEE
Table 2: Summary of results
International Geoscience and Remote Sensing
Attacks
Measurement
Lena
Baboon
Symposium, 6: 3311-3313.
Gaussian noise
PSNR
21.120
20.930
NCC
0.9920
0.9951
Basu, D., A. Sinharay and S. Barat, 2010. Bit plane
PSNR
26.030
25.980
Salt and pepper
index based fragile watermarking scheme for
NCC
0.9982
0.9978
authenticating color image. Proceedings of the First
Sharpening
PSNR
23.430
20.710
International Conference on Integrated Intelligent
NCC
0.9018
0.9341
PSNR
27.850
22.730
Median filter
Computing (ICIIC '10), pp: 136-139.
NCC
0.9927
0.9971
Bauschke,
H.H., 2003. Recompression of JPEG images
Rotation
PSNR
21.910
18.720
by requantization. IEEE T. Image Process., 12(7):
NCC
0.9985
0.9928
843-849.
JPEG compression
PSNR
32.920
29.980
NCC
0.9982
0.9982
Chai, D. and A. Bouzerdoum, 2000. A bayesian
approach to skin color classification in YCbCr
In the Table 2, will explain the summary of the
color space. Proceedings of the TENCON 2000,
results of some a potential attacks which are: Gaussian
Kuala Lumpur, 2: 421-424.
Noise, Salt and Pepper, Sharpening, Median Filter,
Chandramouli, R.R., G. Benjamin and R. Collin, 2001.
Rotation and JPEG compression attacks with two
A multiple description framework for oblivious
measurements which are PSNR and NCC.
watermarking. Proceedings of the Security,
Watermarking and Multimedia, 4314: 585-593.
CONCLUSION
Cox, I.J., M.L. Miller, J.A. Bloom, J. Fridrich and T.
Kalker, 2008. Digital Watermarking and
This study has presented watermarking approach
Steganography. 2nd Edn., Morgan Kaufmann
embedded with discrete wavelet transform and high
Publishers.
frequency sub-band (HH1). This fusion is effective for
Elarbi-Boudihir, M., A. Rehman and T. Saba, 2011.
watermarking as human eye is not able to detect edge
Video motion perception using optimized Gabor
changes (Kashyap and Sinha, 2012). The high
filter. Int. J. Phys. Sci., 6(12): 2799-2806.
frequency areas are more imperceptibility than low and
González, R.C. and R.E. Woods, 2008. Digital Image
middle frequency areas and less sensitive to the human
Processing. Prentice Hall, NY.
visual system, but it is low robustness. That is the main
Hajisami, A., A. Rahmati and M.B. Zadeh, 2011.
idea of this thesis, which is how to embed in this subWatermarking based on independent component
band without low robustness for embedded.
analysis in spatial domain. Proceedings of the
The problem solved by concerning (robustness)
UKSim 13th International Conference on
and finds the best quadrant for embedding. Histogram
Modelling and Simulation, pp: 299-303.
stretching equation used to compress the watermark
Hong, W. and M. Hang, 2006. Robust digital
image before embedding in the quadrant of the cover
watermarking scheme for copy right protection.
image, also the benefit for using histogram stretching
IEEE T. Signal Proces., l2: 1-8.
equation to ensure enhanced the watermark image after
Ibrahim, R. and T.S. Kuan, 2011. Steganography
extracting stage. The Histogram stretching equation
Algorithm to Hide Secret Message inside an Image
applies for watermark image to compress the image
Computer Technology and
Application, 2:
before embedding to get more robustness and enhanced
102-108.
the image after extract it. In addition to that watermark
Jayalakshmi, M., S.N. Merchant and U.B. Desai, 2006.
image convert it to one dimensional array and spread
Digital watermarking in contourlet domain.
over of cover image. This strategy helped to gain
Proceedings of the 18th International conference
imperceptibility, robustness and capacity at the same
on Pattern Recognition, 3: 861-864.
time. The problem solved by concerning (HVS) to find
Kashyap, N. and G.R. Sinha, 2012. Image
the best quadrant for embedding.
watermarking using 3-level discrete wavelet
transform (DWT). I. J. Modern Edu. Comput. Sci.,
REFERENCES
3: 50-56.
Khalili, M., 2003. A comparison between digital image
Al-Otum, H.M. and N.A. Samara, 2010. A robust blind
color image watermarking based on wavelet-tree
watermarking in tow different color spaces using
bit host difference selection. Signal Process., 90(8):
DWT2. National Academy of Science of Armenia
2498-2512.
Yerevan, Armenia.
2620 PSNR
(db)
54.73
54.46
54.72
54.69
52.27
Res. J. Appl. Sci. Eng. Technol., 7(13): 2612-2621, 2014
Kougianos, E. and S.P. Mohanty, 2011. Real-time
perceptual watermarking architectures for video
broadcasting. J. Syst. Software, 84(5): 724-738.
Liu, A.C. and C.H. Chou, 2007. Robustness comparison
of color image watermarking schemes in uniform
and non-uniform color spaces. Int. J. Comput. Sci.
Netw. Secur., 7(7): 239-248.
Maity, S.P., M.K. Kundu and T.S. Das, 2007. Robust
SS watermarking with improved capacity. Pattern
Recogn. Lett., 28(3).
Mohan, B.C. and S.S. Kumar, 2008. A robust image
watermarking scheme using singular value
decomposition. J. Multimedia, 3(1): 7-15.
Muhsin, Z.F., A. Rehman, A. Altameem, T. Saba and
M. Uddin, 2014. Improved quadtree image
segmentation approach to region information.
Imaging Sci. J., 62(1): 56-62.
Nasir, I., Y. Weng and J. Jiang, 2008. Novel multiple
spatial watermarking technique in color images.
Proceeding of the 5th International Conference on
Information Technology: New Generations, pp:
777-782.
Neamah, K., D. Mohamad, T. Saba and A. Rehman,
2014. Discriminative features mining for offline
handwritten signature verification. 3D Res., 5(3).
Premaratne, P., 1999. A novel watermark embedding
and detection scheme for images in DFT domain.
Proceedings of IEE 7th International Conference
on Image Processing and Applications, 2: 780-783.
Rehman, A. and T. Saba, 2012. Neural network for
document image preprocessing: State of the art.
Artif. Intell. Rev., DOI: 10.1007/s10462-0129337-z.
Rehman, A. and T. Saba, 2014. Features extraction for
soccer video semantic analysis: Current
achievements and remaining issues. Artif. Intell.
Rev., 41(3): 451-461.
Rehman, A., S. Alqahtani, A. Altameem and T. Saba,
2013. Virtual machine security challenges: Case
studies. Int. J. Mach. Learn. Cybern., DOI:
10.1007/s13042-013-0166-4.
Saba, T. and A. Rehman, 2012. Effects of artificially
intelligent tools on pattern recognition. Int. J.
Mach. Learn. Cybern., 4(2): 155-162.
Santhi, V., N. Rekha and S. Tharini, 2008. A hybrid
block based watermarking algorithm using DWTDCT-SVD techniques for color images. Proceeding
of the International Conference on Computing
Communication and Networking, pp: 1-7.
Shieh, W. and C. Athaudage, 2006. Coherent optical
orthogonal frequency division multiplexing.
Electron. Lett., 42: 587-589.
Singh, J., P. Garg and A.N. De, 2009. Audio
watermarking using spectral modifications. Int. J.
Inform. Commun. Eng., 5(4): 297-301.
Sulong, G.B., H. Hasan, A. Selamat, M. Ibrahim and
Saparudin, 2012. A new color image watermarking
technique using hybrid domain. Int. J. Comput. Sci.
Issues, 6(1): 109-114.
Tao, P. and A.M. Eskicioglu, 2004. A robust multiple
watermarking scheme in the discrete wavelet
transform
domain.
Internet
multimedia
management systems. Proceedings of the SPIE,
Bellingham, WA, 5601: 133-144.
Tsai, M.J., K.Y. Yu and Y.Z. Chen, 2000. Joint wavelet
and
spatial
transformation
for
digital
watermarking. IEEE T. Consum. Electron., 46(1):
241-245.
Wu, C.W., 2002. One the design of content-based
multimedia authentication systems. IEEE T.
Multimedia, 4(1): 385-393.
2621