Digital watermarking with a new algorithm

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 03 | Mar-2014, Available @ http://www.ijret.org 212
DIGITAL WATERMARKING WITH A NEW ALGORITHM
Afroja Akter1
, Muhammad Ahsan Ullah2
1, 2
Department of Electrical and Electronic Engineering, Chittagong University of Engineering and Technology,
Chittagong, Bangladesh
Abstract
Everyday millions of data need to transmit through a distinct channel for various purposes; as a result there is a certain chance of
third person interruption on that data. In this regards digital watermarking is one of the best solution. This paper proposes a new
embedding algorithm (NEA) of digital watermarking. The algorithm is performed for digital image as data. The performance is
compared for NEA and well established Cox's modified embedding algorithm. The watermarking is based on discrete wavelet
transforms (DWT) and discrete cosine transforms (DCT). The acceptance of the new algorithm is measured by the two requirements
of digital watermarking. One is imperceptibility of the watermarked image, measured by peak signal to noise ratio (PSNR) in dB;
another one is robustness of the mark image, measured by correlation of original mark image and recovering mark image. Here a
512×512 gray scale "Lena" and "Cameraman's" image is taken as host images, and a 128×128 gray scale image is taken as mark
image for 2 level of DWT. The simulation results for different attacking conditions such as salt and pepper attack, additive white
Gaussian noise (AWGN) attack, jpg compression attack, gamma attack, histogram attack, cropping attack, sharpening attack etc.
After different attacks the changing tendency PSNR for both algorithms are similar. But the mean square error (MSE) value of NEA is
always less than Cox’s modified algorithm, which means that after embedding the changes of the host image property lower for NEA
than Cox’s algorithm. From the simulation results it can be said that NEA will be a substitute of modified Cox’s algorithm with better
performance.
Keywords: Digital watermark, DWT, DCT, Cox’s modified algorithm, Lena image, Cameraman image, AWGN, JPG, salt
and pepper attack, PSNR, correlation, MSE.
----------------------------------------------------------------------***------------------------------------------------------------------------
1. INTRODUCTION
Digital watermarking is a scheme where a secret data, it may
an image, audio or video, embedded in a host data such a way
that it remains on it as long as the perceptible quality of the
content in an acceptable level [1]. From the last decade digital
watermarking is best choice for data security. Digital
watermarking may classified as (1) blind or (2) informed. This
paper only focused on informed digital watermarking, where
the mark signal extracted using the host data. Digital
watermarking procedure mainly classified into three steps [14]
 Embedding
 Distribution
 Extraction or Detection
Embedding is the process where a secret data i.e. either an
image or an audio or video is embedded into a host data which
may an image or an audio or video, using a key. Then this
embedded data is broadcast or distributed through a lossless or
lousy channel, which is known as distribution. The third step
is depending on author, when the existence of the mark signal
can be detected using the host data with known key is known
as detection. But when the mark data is recover from the
embedded data using the known host data and key is known as
extraction. The process is shown in figure 1.
Fig 1: Digital watermarking system.
Furthermore digital watermarking is based on two techniques
 Spatial domain based [2]
 Transform domain based
The most common spatial domain based technique is least
significant bit (LSB) [2-4, 13] method. The proposed
technique is based on transform domain. Again the transform
domain based digital watermarking can be classified into
discrete wavelet transforms (DWT) [5-10, 12, 14], discrete
Cosine transforms (DCT) [5, 7, 11, 15]. Multi resolution
properties is the most important characteristics of DWT based
digital watermarking, as for the specified region can be
selected for data embedding. Every image can be divided into

__________________________________________________________________________________________
two specific regions: (1) low frequency region and (2) high
frequency region, and these regions can be sub-divided in case
of DWT based digital watermarking. DCT divide the image
in three regions: (1) low frequency region, (2) mid-frequency
region and (3) high frequency region. It is easy to select a
distinct region to embed a data in case of DCT based digital
watermarking. This paper based on both of these technique,
thus it can provide better performance.
A 512×512 gray scale image is taken as host image, and a
128×128 gray scale image is taken as mark image. After 2-
level of DWT of host image the mark is embedded on a
selected sub-band. This paper considers the sub-bands from a
high frequency region of DWT, and then selects the
comparative low frequency or high co-efficient sets for DCT.
At lower level of DWT based digital watermarking it is
challenging to embed the mark image into the host image
without hampering its perception level, thus here this situation
is tested.
Acceptance of the proposed algorithm is measured by two
most important requirements of the digital watermarking. One
of them is imperceptibility of the embedded data, measured by
peak signal to noise ratio (PSNR) in dB, and remaining one is
robustness of the recovering data, measured by the correlation
of original mark data and recover data. Greater PSNR is
desired, as it means the perception level of the host image is
little bit interrupted by the mark image. And a better
correlation (maximum 100%, i.e. 1) measured that the mark
image is not interrupted by the attacks in a broadcasting
channel. Another parameter known as mean square error
(MSE) also considering. As less MSE indicates that the mark
image interrupts the host image less. i.e. less MSE is
acceptable. This paper evaluate the technique considering
different attacks like salt and pepper attack, additive white
Gaussian noise (AWGN) attack, jpg compression attack,
gamma attack, histogram attack, cropping, sharpening attack,
rotational attack etc. From simulation results it can be decided
that the new algorithm is accepted with comparing Cox's
modified algorithm even at different attack condition.
This paper is arranged as following sections; in section two
different parameters for digital watermarking are discussed.
The hybrid technique of digital watermarking is discussed in
section three. Simulation results are presented in section four.
2. PERFORMANCE MEASURE
Digital watermarking has certain requirements. To measure
these requirements different image quality metrics are used.
This paper is using peak signal to noise ratio (PSNR) to
measure the amount of visual quality degradation between
original and watermarked image. This criterion is very well
known as imperceptibility of the embedded image [7]. This
paper also measures the robustness of the mark image.
Robustness is measured by correlation of original mark image
and recovered mark image.
2.1 Peak Signal-To-Noise Ratio (PSNR)
To measure the quality of reconstruction of lossy compression
PSNR is frequently used [7, 13]. And it is the ratio of the
maximum possible power of a host image to the power of
corrupting noise that affects the fidelity of its representation. It
can define by the following equation:
2.1
Where MAX represents the maximum possible pixel value in
the host image and MSE is shows the mean square error in
embedded image.
MSE is measured from the host image and noisy
approximation of it [13]. MSE can be defined by the equation
(2.2)
2.2
Where, ),( lkf is host image and ),(' lkf is
watermarked/embedded image.
2.2 Robustness
Robustness measure the tolerance of the technique after
experiencing different attacks, it is measured for the mark
image. I.e. distributing of the embedded image through a
loosy channel how much it experiences the degradation of
mark image is measured by robustness. The measuring tool of
it is correlation of the original mark image and recover mark
image. Correlation of two images can be found from
following equation.
2.3
where w (i,j) and w’(i,j)’ are the original and extracted/recover
watermark image respectively.
3. HYBRID TECHNIQUE
This paper discusses the digital watermarking technique based
on transform domain. Most popular transform domain
techniques included: discrete wavelet transforms (DWT) and
discrete cosine transforms (DCT), separately these techniques
having some merits and demerits, but combining both of these







MSE
MAX
PSNRdB 10log20
 
2
1 1
),('),(
1
 








m
k
n
l
lkflkf
nm
MSE





















i j
avg
i j
avg
i j
avgavg
jiwjiwjiwjiw
jiwjiwjiwjiw
)}),('),('{}),(),({
}),('),('{*)},(),({
Corr.
2
.
2
.
..

__________________________________________________________________________________________
can provides better performance. In this regard this paper
evaluates the new algorithm based on hybrid technique. In this
paper this hybrid digital watermarking technique is discussed
in two steps:
 Digital watermarking embedding technique.
 Digital watermarking extracting technique.
Both steps are discussed as follows:
3.1 Digital Watermarking Embedding Technique:
Using DWT a 512×512 host image (here gray scale "Lena" or
"Cameraman") is sub-divided into 4-subbands. Figure 3
represents the 1st level of DWT decomposition of host image.
Where LL, HL, LH, HH are 4 sub-bands. Here LL represents
the high scale low frequency co-efficient set, HL represents
the horizontal details of low scale high frequency set, LH
represents the vertical details of low scale high frequency set,
HH represents the diagonal details low scale high frequency
set of the host image.
Fig 2: Flow chart of the embedding procedure.
Then using 2 dimensional of DWT on HL of the host image,
2nd level decompose image can be found as shown in figure
4.
After the 2nd level decomposition of the host image then the
comparative low frequency components denote as HLLL2 is
select for 2 dimensional DCT. It is common practice to select
the mid frequency region for embedding purposes during DCT
based digital watermarking for robustness, but having chance
to degrade the perceptibility of the host image. To trade of the
imperceptibility and robustness this paper focused on the low
frequency region after 2 dimensional DCT of the host image.
Then a 128×128 gray scale image as a mark is embedded on
the selected region after enrolled a key and DCT on it. With
scaling factor 0.9 the mark image is embedded into the host
image using the new hybrid algorithm and Cox's modified
algorithm as in equation 3.1 and equation 3.2 respectively
Fig 3: 1st level DWT decomposition of "Lena" image.
Fig 4: 2nd level DWT decomposition of "Lena" image.
Fig 5: DCT region of an image

__________________________________________________________________________________________
.
22
)(),(
)(),(
),(),('
mfnmf
mfnmf
wnmfnmf
avg
avg


  3.1
wnmfnmf  ),(),(' 3.2
Where f(m,n) is the host image, favg.(m) is mean value of the
host image, w is original mark image.
Here from Cox's modified algorithm it has been shown that
the mark image totally independent on host image, thus the
imperceptibility is degraded sharply when embedded or
watermarked image is experienced different attacks from
without attack condition. But from the new algorithm it is
noticed that the mark image is multiplied with the absolute
deviation of host image. Thus here degradation rate is slow.
3.2 Digital Watermarking Extraction Technique:
The extraction technique is depicts in figure 6. To extract the
mark image from the embedded or watermarked image the
original 512×512 image is used, hence this is an informed
digital watermarking technique.
Fig 6: Flowchart of extracting procedure
First of all the watermarked image is decomposed into 2-level
of DWT, then the extraction location is selected. Then DCT is
done on the selected sub-bands. Similarly the host image is
first decomposing into 2-level DWT and then DCT on the
selected sub-band. Then the mark image is extracted using the
new algorithm and Cox's modified algorithm respectively as
bellows in equation 3.3 and equation 3. 4:
)](),([
)(),()],(),('[
'
22
mfnmf
mfnmfnmfnmf
w
avg
avg




3.3

),(),('
'
nmfnmf
w

 3.4
Where w’ is recover image. Then IDCT is done on the
extracted watermark and the recover mark image is found
enrolled the secret key.
4. RESULTS
To evaluate the performance of the new algorithm and
comparative result, here 512×512 Lena image and a 512×512
Cameraman image are used while a 128×128 gray scale image
is used as mark image. For simulation considering scaling
factor is 0.9.
Fig 7: Host Lena image
Fig 8: Host Cameraman image.

__________________________________________________________________________________________
Table 1: PSNR, MSE and Correlation Values for Lena image
Table 2: PSNR, MSE and Correlation Values for Cameraman image
New algorithm Cox's modified algorithm
Attack PSNR(dB) MSE Correlation PSNR(dB) MSE Correlation
No attack 28.13 99.97 1.00 28.03 102.15 1.00
Salt and pepper 29.01 81.5195 0.72 28.95 82.67 0.87
AWGN 28.77 86.21 0.58 28.72 87.19 0.80
Jpg compression 29.04 81.09 0.95 28.97 82.33 0.98
Gamma
Correction
25.11 200.06 0.55 25.07 202.16 0.66
Histogram 29.49 73.09 0.57 29.45 73.67 0.81
Cropping 29.08 80.21 0.89 29.02 99.56 0.97
Sharpening 32.92 34.92 0.69 35.89 16.73 0.89
Speckle Noise 28.25 97.28 0.18 28.22 97.89 0.31
Median filter 29.14 79.10 0.53 29.22 77.65 0.79
Scaling 29.17 78.65 0.10 29.94 65.80 0.14
Fig 9: Watermark image.
Figure 7 is Lena image and figure 8 is Cameraman image. In
this paper both of these image are consider as host image. And
figure 9 represents the watermark image. Table 1 shows
different results for both new algorithm and Cox's modified
algorithm for Lena image, whereas table 2 represents results for
Cameraman image. From both table it can be said that the
PSNR value of the watermarked image is slightly degraded
from the original image at different attacks whereas this is
sharply degraded for Cox's modified algorithm. Here salt and
pepper attack, additive white Gaussian noise (AWGN) attack,
jpg compression attack, gamma attack, histogram attack,
cropping attack, sharpening attacks are consider for
simulation. For salt and pepper attack the PSNR value for the
Cox's modified algorithm is increased. This phenomenon is
New algorithm Cox's modified algorithm
Attack PSNR(dB) MSE Correlation PSNR(dB) MSE Correlation
No attack 27.87 98.08 1.00 27.69 102.14 1.00
Salt and pepper 28.59 82.97 0.85 27.73 101.08 0.91
AWGN 28.12 92.47 0.77 27.65 103.11 0.85
Jpg compression 28.62 82.33 0.99 27.75 100.63 0.99
Gamma Correction 23.73 253.9 0.85 23.72 255.00 0.88
Histogram 27.62 103.64 0.86 27.60 104.10 0.88
Cropping 28.68 81.22 0.96 27.80 99.56 0.97
Sharpening 27.80 99.53 0.83 27.76 100.40 0.84
Speckle Noise 27.13 116.03 0.28 26.94 121.19 0.37
Median filter 29.41 68.81 0.81 28.41 86.41 0.85
Scaling 31.26 44.81 0.29 30.92 48.48 0.34

__________________________________________________________________________________________
also noticed for the NEA. But MSE value of the Cox's modified
algorithm is greater than the NEA, which means that the host
image is less hampered for NEA than the Cox’s algorithm.
Similar criterion is noticed for other attacks like AWGN, Jpg
compression, cropping, median filter and scaling attacks. But in
all cases the PSNR of the new embedding algorithm (NEA) is
higher than the Cox’s modified algorithm. In case of
correlation the NEA and Cox’s modified algorithm having
almost similar value. Thus considering all conditions this paper
appreciates the new algorithm for digital watermarking.
5. CONCLUSIONS
An informed digital watermarking based on DWT-DCT
technique with a new algorithm is presents in this paper. This
paper also shows the comparative results of Cox's modified
algorithm and new algorithm for Lena and Cameraman image.
The PSNR changing criterion is similar for both new
embedding algorithm (NEA) and Cox’s modified algorithm.
Although in all cases PSNR of the NEA is higher than the
Cox’s algorithm. The MSE value indicates that the NEA can
embed the mark image into the host image with less
degradation than the Cox’s algorithm. From all simulation
results this paper appreciates the performance of new
embedding algorithm (NEA) and it can be said that the NEA
will be a substitute of modified Cox’s algorithm with better
performance.
REFERENCES
[1]. Jonathan K. Su, Frank Hartung, Bernd Girod,” Digital
Watermarking of Text, Image, and Video Documents”,
Telecommunications Laboratory University of Erlangen-
Nuremberg Erlangen, Germany.
[2]. Ibrahim Nasir, Ying Weng, Jianmin Jiang," A New Robust
Watermarking Scheme for Color Image in Spatial Domain",
School of Informatics, University of Bradford, UK
[3]. R.G.van Schyndel, A.Z.Tirkel, C.F.Osborne,”A DIGITAL
WATERMARK ” ,Department of Physics, Monash University,
Clayton, 3168, Australia. Scientific Technology, 21 Walstab St,
E. Brighton, 3187, Australia
[4]. Mr. Rohith.S1, Dr. K.N.hari bhat2,” A Simple Robust
Digital Image Watermarking against Salt and Pepper Noise
using Repetition Codes”,Nagarjuna College of Engineering and
Technology, Bengaluru, Karnataka, India.
[5]. Shital Gupta, Dr Sanjeev Jain, "A Robust Algorithm of
Digital Image Watermarking Based on Discrete Wavelet
Transform", Department of Computer Science and
Engineering, LNCT, Bhopal.
[6]. Nidhi Bisla, Prachi Chaudhary, "Comparative Study of
DWT and DWT-SVD Image Watermarking Techniques",
International Journal of Advanced Research in Computer
Science and Software Engineering", Volume 3, Issue 6, June
2013
[7]. M.A. Mohamed and A.M. El-Mohandes, "Hybrid DCT-
DWT Watermarking and IDEA Encryption of Internet
Contents", International Journal of Computer Science Issues,
Vol.9, Issue 1, No 2, January2012
[8]. Emir Ganic, Ahmet M. Eskicioglu,"Robust DWT-SVD
Domain Image Watermarking: Embedding Data in All
Frequencies".
[9]. Gaurav Bhatnagar, Balasubramanian Raman," A new
robust reference watermarking scheme based on DWT-SVD",
Department of Mathematics, Indian Institute of Technology
Roorkee, Roorkee - 247667, India
[10]. Praful Saxena, Shanon Garg and Arpita Srivastava,"
DWT-SVD Semi-Blind Image Watermarking Using High
Frequency Band", 2nd International Conference on Computer
Science and Information Technology (ICCSIT'2012) Singapore
April 28-29, 2012.
[11]. Mr.D.V.N.Koteswara Rao,Y.Madhuri, S.V.Rajendra
Kumar ,Y.V.Suresh Babu, " Robust Image Watermarking using
DCT & Wavelet Packet Denoising ", International Journal of
Scientific & Engineering Research Volume 3, Issue5, May-
2012.
[12]. Corina Nafornita and Alexandru Isar, "Application of
discrete wavelet transform in watermarking",, Politehnica”
University of Timisoara, Romania.
[13]. Koushik Pal, G. Ghosh1 and M. Bhattacharya, " A Novel
Digital Image Watermarking Scheme for Data Security Using
Bit Replacement and Majority Algorithm Technique", Institute
of Radio Physics and Electronics, University of Calcutta,
Kolkata Indian .
[14]. Ms. T Mita Kumari, ”Image Adaptive Watermarking
Using Wavelet Transform”, Department of Electrical
Engineering, National Institute of Technology, Rourkela
769008, 2007.
[15]. Abrar Ahmed Syed, "Digital Watermarking", The
University of Texas, Arlington.
BIOGRAPHIES
Afroja Akter, Graduated in Electrical and
Electronic Engineering, From Chittagong
University of Engineering and Technology,
Chittagong, Bangladesh in September 2013.
Her research interests are in image
processing, VLSI, and Renewable energy.
Muhammad Ahsan Ullah received B.Sc. in
Engineering degree in Electrical and
Electronic Engineering from Chittagong
University of Engineering and Technology,
Bangladesh in 2002, M.E. degree in
Electronic and Information Engineering
from Kyung Hee University, Republic of
Korea in 2007 and PhD degree in Electrical Engineering at
Nagaoka University of Technology at 2011. He is interested in
Signal processing, Coding Theory, threshold decoding, MC-
CDMA, OFDMA, MIMO, and Image Processing.

Digital watermarking with a new algorithm

More Related Content

What's hot (16)

Viewers also liked (20)

Similar to Digital watermarking with a new algorithm (20)

More from eSAT Journals (20)

Recently uploaded (20)

Digital watermarking with a new algorithm