Cascade Algorithm of Image Watermarking Based On Multiple Wavelet Components

Council for Innovative Research
www.cirworld.com
International Journal of Computers & Distributed Systems

Volume 1, Issue 2, August, 2012
Cascade Algorithm of Image Watermarking Based on

Multiple Wavelet Components
Amol R. Madane
Dr. Uttam D. Kolekar
Research Scholar,
JJT University, Jhunjhunu,
Rajasthan-333 001, India
Research Guide,
Parshvanath College of Engineering, Thane (W),
Mumbai-400 615, India
ABSTRACT
This paper presents the method of embedding multiple secrete
information (Logos or Messages etc) one by one into the same
host image which is new topic of research. The method
consists of multiple frequency bands which are altered by
putting the different secrete information at different locations.
The tool used to perform the task is discrete wavelet transform
which provides imperceptibility, robustness and security
requirements. Peak signal to noise ratio, mean square error,
correlation coefficients are the parameters used to check the
effectiveness of algorithm. The algorithm is tested by
applying the same on multiple input images with multiple
secrete information. The existing provides more security than
the normal watermarking schemes as host image is having
multiple secrete information for copyright protection.
General Terms
Digital image watermarking, discrete wavelet transforms,
security, algorithms.
Keywords
Addition, Correlation Coefficients, Discrete Wavelet
Transform, Insertion, Inverse Discrete Wavelet Transform,
Mean Square Error, Peak Signal to Noise Ratio.
1. INTRODUCTION
Digital image watermarking is art of embedding i.e. add or
insert the digital secrete information such as logo, text,
message etc into the digital host image. There are various
tools which are used to embed the secrete information by
manipulating the details of the host image. The hiding process
has to be such that the modifications of details are
imperceptible to the normal human eye. Digital watermarking
came to be in great demand when sharing information on the
internet became a usual practice. Sharing files online, you
never know if someone uses them without your consent. To
prevent unauthorized commerce use of your files, you can
publish them to the web in the worst quality or dont publish
anything worthwhile at all. It isnt a good way to solve the
problem of unauthorized use, is it? So, you should look for
more effective ways of copyright protection, such as digital
watermarking.
A digital watermark is a pattern of bits inserted into a digital
file - image, audio or video. Such messages usually carry
copyright information of the file. Digital watermarking takes
its name from watermarking of paper or money. But the main
difference between them is that digital watermarks are
supposed to be invisible or at least not changing the
perception of original file, unlike paper watermarks, which are
supposed to be somewhat visible.
In this paper, we will discuss the cascade algorithm of image
watermarking which is based on different wavelet
components. Section 2 reviews the literature survey on digital
26 | P a g e
www.cirworld.com
image watermarking using DWT. Section 3 will define the

problem. Section 4 discusses the proposed algorithm. Section
5 will give an idea about the results and summary of
discussion.
2. RELATED WORK
Authors [1] proposed the watermarking algorithm by using
the tool like discrete wavelet transform and discrete cosine
transform. Z. Yucel, A.B. Ozguler [2] explained new method
for digital image watermarking based on zero assigned filter
banks and embedded zero tree wavelet (EZW) algorithm.
Authors [3] highlighted the review of recent as well as classic
image copy detection method. Paper [4] proposed a novel
watermarking scheme for an image, in which a logo
watermark is embedding into the multiwavelet domain of the
image using neural network. Mauro Barni, Franco Bartolini,
Alessandro Piva [5] explain the improved wavelet-based
watermarking through pixel wise masking. In [6], authors
highlighted that the invisible digital image watermarking can
carry the significant information. Authors [7] discussed the
algorithm to increase the robustness and protection. In paper
[8], A. Essaouabi, E. Ibnelhaj, F. Regragui explained
watermarking scheme that can embed a watermark to an
arbitrarily shaped object in an image. Authors [9] presented a
method for digital image watermarking using Biorthogonal
wavelet transform. A DWT based algorithm is proposed in
[10] where in both blind and non-blind algorithms are used for
copyright protection. In [11], authors successfully explained
the digital image watermarking technique on digital images
based on discrete wavelet transform by analyzing various
values of PSNRs and MSEs.
3. PROBLEM DEFINITION
In case of image watermarking, the secrete information is
inserted into the host image for copyright protection. This
type of applications can be used for copyright protection. In
all the previously presented algorithms, authors have inserted
the secrete information to perform the system better in terms
of performance. The proposed system gives the new approach
to insert the three different secrete information into the same
host image step by step. This way is beneficial for us to
provide the more security, more authentications.
4. PROPOSED ALGORITHM
4.1. Embedding Algorithm
Cascade Insertion Algorithm for adding the Secrete
Information in the input image,
4.1.1.
a)
b)
Stage 1-
Consider any square gray scale image.

Decompose the input image by using 2D discrete wavelet
transform (DWT). Store the first level approximation
coefficients i.e. ILL1, horizontal coefficient ILH1,

www.cirworld.com
c)
d)
e)
f)
g)
h)
i)

vertical coefficient IHL1, diagonal coefficient IHH1 as

first level watermark key coefficients of input image at
stage 1.
Approximation coefficient of first level of input image is
ILL1 which is further decomposed into new coefficients
i.e. ILL2, horizontal coefficient ILH2, vertical coefficient
IHL2, diagonal coefficient IHH2 as second level
watermark key coefficients of input image using 2D
discrete wavelet transform.
Consider the square gray scale image as first secrete
information.
Decompose the secrete information by using 2D discrete
wavelet transform. Store the first level approximation
coefficients i.e. SLL1, horizontal coefficient SLH1,
vertical coefficient SHL1, diagonal coefficient SHH1 as
first level watermark key coefficients of secrete
information.
Reduce the intensity values of SLL1 part of first secrete
information to half i. e SLL1mod.
Replace (insert) the diagonal part of input image (IHH2)
by SLL1mod part of first secrete information pixel by
pixel.
Perform the two level 2D inverse discrete wavelet
transform (IDWT) of input image by using
approximation coefficients of input image to find the
cascade invisible watermark image at stage 1 using
cascade insertion algorithm. Gray scale invisible
watermark image is obtained, which is perceptually equal
to original input image.
Find the mean square error (MSE), peak signal to noise
ratio (PSNR) and correlation coefficient (CC) in between
the original input image and cascade invisible watermark
image at stage 1 by using the related formulae as these
are the important performance parameters.
4.1.2.
Stage 2-
j)
Consider the cascade invisible watermark image of stage

1 as an input image for the next stage.
k) Decompose the input image by using 2D discrete wavelet
stage 2.
l) Approximation coefficient of first level of input image is
m) Consider the square gray scale image as second secrete
information.
n) Decompose the secrete information by using 2D discrete
information.
o) Reduce the intensity values of SLL1 part of second
secrete information to half i. e SLL1mod.
p) Replace (insert) the horizontal component of input image
(ILH4) by SLL1mod part of second secrete information
pixel by pixel.
q) Perform the two level 2D inverse discrete wavelet
27 | P a g e
www.cirworld.com
r)

ratio (PSNR) and correlation coefficients (CC) in
between the original input image and cascade invisible
watermark image at stage 2 by using the related formulae
as these are the important performance parameters.
4.1.3.
Stage 3-
s)
Consider the cascade invisible watermark image at stage

2 as an input image for the next stage.
t) Decompose the input image by using 2D discrete wavelet
stage 3.
u) Approximation coefficient of first level of input image is
v) Consider the square gray scale image as third secrete
information.
w) Decompose the secrete information by using 2D discrete
information.
x) Reduce the intensity values of SLL1 part of secrete
information to half i. e SLL1mod.
y) Replace (insert) the vertical component of input image
(IHH6) by SLL1mod part of third secrete information
pixel by pixel.
z) Perform the two level 2D inverse discrete wavelet
aa) Find the mean square error (MSE), peak signal to noise
ratio (PSNR) and correlation coefficients (CC) in
between the original input image and cascade invisible
watermark image at stage 3 by using the related formulae
as these are the important performance parameters.
bb) We can make program generalize as per the size is
concern i.e. we can change the size of input image and
secrete information by changing the size parameters in
program. But only essential condition is that both images
should be square images.
4.2.
a)
b)
Proposed Extraction Algorithm
Consider invisible watermark image at stage 3 as input

image.
Decompose the invisible watermark image by using 2D
discrete wavelet transform. Then we will get the first
level approximation coefficients i.e. InverseLL1,
horizontal coefficient InverseLH1, vertical coefficient
InverseHL1, diagonal coefficient InverseHH1 as first
level watermark key coefficients of invisible watermark
image.

www.cirworld.com
c)
d)
e)
f)
g)
h)
i)
j)

Approximation coefficient of first level is InverseLL1

which is further decomposed into new coefficients i.e.
InverseLL2, horizontal coefficient InverseLH2, vertical
coefficient InverseHL2, diagonal coefficient InverseHH2
as second level watermark key coefficients of invisible
watermark image.
Consider the InverseHH2 part of cascade invisible
watermark image as secrete information and multiply by
the scaling factor to get first extracted secrete
information NewSLL2.
transform (IDWT) of first secrete information by using
approximation coefficients of secrete information to find
the first secrete information having size 256X256 using
insertion algorithm. Gray scale secrete information is
obtained, which is perceptually equal to first secrete
information.
Consider the InverseLH2 part of cascade invisible
the scaling factor to get second extracted secrete
transform (IDWT) of second secrete information by
using approximation coefficients of secrete information
to find the second secrete information having size
256X256 using insertion algorithm. Gray scale secrete
information is obtained, which is perceptually equal to
second secrete information.
Consider the InverseHL2 part of cascade invisible
the scaling factor to get third extracted secrete
transform (IDWT) of third secrete information by using
approximation coefficients of secrete information to find
the third secrete information having size 256X256 using
insertion algorithm. Gray scale secrete information is
obtained, which is perceptually equal to third secrete
information.
ratio (PSNR) and correlation coefficient (CC) in between
the original secrete information and extracted secrete
information by using the related formulae as these are the
important performance parameters.
5. PERFORMANCE PARAMETERS
The performance parameters used for comparing the results
are Peak Signal to Noise Ratio (PSNR), Mean Square Error
(MSE), and Correlation Coefficients (CC). The
imperceptibility of a watermark is measured by the
watermarked image quality in terms of Peak-Signal-to- Noise
Ratio (PSNR) (in dB) [9]. The robustness performance of
watermark extraction is evaluated by normalized correlation
coefficient, r, of the extracted watermark A and the original
watermark B.
Where A and B respectively, the normalized original and

extracted watermark by subtracting its corresponding means
value. The magnitude range of r is [0, 1], and the unity holds
if the extracted image perfectly matches the original one. The
difference between host image and watermark image or
watermark logo and extracted watermark logo can be found
28 | P a g e
www.cirworld.com
by using the factor mean square error. It is better to have the

value less as possible.
6. RESULTS AND DISCUSSION

The existing algorithm has been tested on different types of
host images. The three different types of secrete information
used for experimentation are shown in figure 1.
(a)
(b)
(c)
Fig 1: Secrete information (256X256)
(a) 1, (b) 2, (c) 3
The different input images used for experimentation purpose
are shown in figure 2.
(a)
(b)
(c)
(d)
(e)
(f)

www.cirworld.com
(g)
(i)

(a)
(b)
(c)
(d)
(h)
(j)
Figure 3: (a) Input Image, (b) Cascade invisible

watermark image at stage 1, (c) Cascade invisible
watermark image at stage 2, (d) Cascade invisible
watermark image at stage 3 i.e. final cascade invisible
watermark image
Table 1: Performance parameters for original image and
invisible watermark image
Input Image
(k)
(l)
Amol
Baby
Boat
Brain
(m)
(n)
Bridge
Fig 2: Different Input Image (512X512) (a) Amol, (b)

Baby, (c) Boat, (d) Brain, (e) Bridge, (f) Crowd, (g) Finger,
(h) Lax, (i) Lena, (j) Pancard, (k) Peppers, (l) Text, (m)
Woman1, (n) Woman2
The secrete information shown in Figure 1 are added/inserted
in different host images (as shown in Figure 2) as per the
algorithm. The output images at different stages are shown in
figure 3. If we compare the image input image with the
invisible watermark image at all the stages, both the image
looks like as same image. The secrete information is
perceptually invisible in watermark image i.e. invisible
watermark image.
Table 1 will give the different values of performance
parameters like peak signal to noise ratio (PSNR), mean
square error (MSE), correlation coefficient (CC) between
original image and watermark image at different levels for
different input images.
29 | P a g e
www.cirworld.com
Crowd
Finger
Lax
Lena
Pancard
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Original Image and Watermark

Image
PSNR
MSE
CC
78.2167 9.7275e-004 0.9977
78.2167 9.7275e-004 0.9934
78.2167 9.7275e-004 0.9887
78.2508 9.7275e-004 0.9957
78.2508 9.7275e-004 0.9904
78.2508 9.7275e-004 0.9855
78.0090 9.7275e-004 0.9942
78.0090 9.7275e-004 0.9799
78.0090 9.7275e-004 0.9670
78.2508 9.7275e-004 0.9940
78.2508 9.7275e-004 0.9700
78.2508 9.7275e-004 0.9452
78.2508 9.7275e-004 0.9900
78.2508 9.7275e-004 0.9652
78.2508 9.7275e-004 0.9476
78.2508 9.7275e-004 0.9905
78.2508 9.7275e-004 0.9690
78.2508 9.7275e-004 0.9489
78.2508 9.7275e-004 0.9904
78.2508 9.7275e-004 0.9337
78.2508 9.7275e-004 0.9050
77.6880 9.7275e-004 0.9750
77.6880 9.7275e-004 0.9385
77.6880 9.7275e-004 0.8849
78.2508 9.7275e-004 0.9918
78.2508 9.7275e-004 0.9805
78.2508 9.7275e-004 0.9610
78.2508 9.7275e-004 0.9842
78.2508 9.7275e-004 0.9580
78.2508 9.7275e-004 0.9049

www.cirworld.com

Peppers
Diagonal
77.2405 9.7275e-004 0.9930
Horizontal 77.2405 9.7275e-004 0.9797
Vertical
77.2405 9.7275e-004 0.9646
Text
Diagonal
77.2167 9.7275e-004 0.9879
Horizontal 77.2167 9.7275e-004 0.9338
Vertical
77.2405 9.7275e-004 0.8968
Woman
Diagonal
76.1839 9.7275e-004 0.9873
1
Horizontal 76.1839 9.7275e-004 0.9695
Vertical
76.1839 9.7275e-004 0.9447
Woman
Diagonal
78.2167 9.7275e-004 0.9952
2
Horizontal 78.2167 9.7275e-004 0.9890
Vertical
78.2167 9.7275e-004 0.9826
Then, the secrete information extracted from invisible
watermark image with the help of extraction algorithm. Table
2 gives the values of performance parameters of secrete
information and the extracted secrete information.
Table 2: Performance parameters for secrete information
and the extracted secrete information
Input Image
Amol
Baby
Boat
Brain
Bridge
Crowd
Finger
Lax
Lena
Pancard
Peppers
Text
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Original Secrete Information and

Extracted Secrete Information
PSNR
MSE
CC
72.2302
0.0039
0.9428
72.2302
0.0039
0.9606
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
30 | P a g e
www.cirworld.com
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.9834
0.9277
0.9449
0.9918
0.9996
0.9995
0.9996
0.9655
0.9853
0.9937
0.9887
0.9939
0.9993
0.9968
0.9911
0.9987
0.9990
0.9992
0.9995
0.9994
0.9995
0.9995
0.9972
0.9985
0.9993
0.9644
0.9820
0.9867
0.9991
0.9985
0.9994
0.9968
0.9989
Woman
1
Woman
2
Vertical
Diagonal
Horizontal
Vertical
Diagonal
Horizontal
Vertical
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
72.2302
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.0039
0.9987
0.9920
0.9968
0.9995
0.9850
0.9941
0.9989
The existing algorithm is robust enough to show the best

results for all types of different input images.
If we observe the values of performance parameters for
watermark images (table 1) and secrete information (table 2),
Lena image gives best results in all the respects. Hence, we
selected the same image for future research work.
After adding the secrete information at Diagonal, Horizontal,
Vertical locations separately, then we are getting best result
after adding the secrete information at diagonal (i.e. values of
PSNR, MSE, CC) and worst results after adding the secrete
information at vertical location (i.e. values of PSNR, MSE,
CC). So, we chose the sequence of adding the secrete
information as diagonal, horizontal, vertical.
If we compare the results of insertion and addition, then
overall observation is as follow,
In case of Watermark images, Addition algorithm gives
better result than insertion algorithm (by comparing the values
of PSNR, MSE, and CC).
On the other hand, insertion algorithm gives better result than
the addition algorithm (by comparing the values of PSNR,
MSE, and CC) in case of secrete information.
7. CONCLUSION
We have developed the application for the demonstration of
cascade invisible digital image watermarking with a given set
of secrete information. The application provides the best value
of mean square error (MSE), peak signal to noise ratio
(PSNR) and normalized correlation coefficient (CC) for the
cascade algorithm that determines the robustness of the
secrete information in digital image. The addition algorithm
gives better results than insertion algorithm in case of
watermark image. On the other side, insertion algorithm gives
better result than addition algorithm in case of secretes
information. The developed algorithm not only detect but also
extract the secrete information whenever required. As we have
inserted/added three different secrete information, the
algorithm is healthy enough to provide the maximum
probability of authentication and copyright protection.
8. REFERENCES
[1]. Jumma Dr. Bassim Abdulbaki, Aladdin Arwa 2010,
Image Watermarking using DWT_DCT, Engineering
and Technology Journals, pp. 1-14, Vol. 28, No. 23,
Nov 2010.
[2]. Yucel Z., Ozguler A.B. 2005, An image watermarking
algorithm via zero assigned filter banks, Proceedings of
Fifth IEEE International Symposium on Signal
Processing and Information Technology, pp. 363-368,
Dec 2005.
[3]. Wan Y. H., Yuan Q. L., S. M. He Ji, L. M., Wang Y. L.
2008, A survey of the image copy detection, IEEE
Conference on Cybernetics and Intelligent Systems, pp.
738-743, Sep 2008.
[4]. Zhang Jun, Wang Nengchao, Xiong Feng 2002, Hiding
a logo watermark into multiwavelet domain using neural
network, Proceedings of 14th IEEE International

www.cirworld.com
[5].
[6].
[7].
[8].
[9].

Conference on Tools with Artificial Intelligence, pp.

477-482, Nov 2002.
Barni Mauro, Bartolini Franco, Piva Alessandro 2001,
"Improved Wavelet-Based Watermarking Through PixelWise Masking", IEEE Transaction on Image Processing,
pp. 783-791, Vol. 10, No. 5, May 2001.
Premaratne Prashan, DeSilva Liyanage, Burnett Ian
2006, "Low Frequency Component-Based Watermarking
Scheme Using 2D Datamatrix", International Journal of
Information Technology, pp. 1-12, Vol. 12, No. 7,
November 2006.
Sharkas Maha, ElShafie Dahlia, Hamdy Nadder 2007, "A
Dual Digital-Image Watermarking Technique", Journal
of World Academy of Science, Engineering and
Technology, pp. 136-139, Vol. 5, No. 1, May 2007.
Essaouabi A., Ibnelhaj E., Regragui F. 2009, Digital
Image Watermarking for Arbitrarily Shaped Objects
Based on SA-DWT, International Journal of Computer
Science Issues, pp. 1-8, Vol. 5, No. 1, October 2009.
Hajjara Suhad, Abdallah Moussa, Hudaib Amjad 2009,
Digital Image Watermarking Using Localized
Biorthogonal Wavelets, European Journal of Scientific
Research, pp. 594-608, Vol. 26, No. 4, February 2009.
31 | P a g e
www.cirworld.com
[10]. Tripathi Shikha, Ramesh Nishanth, Bernito, Neeraj K J

2010, A DWT based Dual Image Watermarking
Technique for Authenticity and Watermark Protection,
International Journal of Signal and Image Processing, pp.
33-45, Vol. 1, No. 2, December 2010.
[11]. Singh Akhil Pratap, Mishra Agya 2010, Wavelet Based
Watermarking on Digital Image, Indian Journal of
Computer Science and Engineering, pp. 86-91, Vol. 1,
No. 2, August-September 2010.
[12]. Kumar Amitesh, Santhi V. 2011, "A Review on
Geometric Invariant Digital Image Watermarking
Techniques", International Journal of Computer
Applications, pp. 09758887, Vol. 12, No. 9, January
2011.
[13]. Zhang Fan, Zhang Hongbin 2004. Image Digital
Watermarking Capacity and Reliability Analysis in
Wavelet Domain, IEEE Proceedings of 47th Midwest
symposium on Circuits and Systems, pp. 101-104, Jul
2004.
[14]. Madane Amol R., Watermark Image Recognition from
Local Scale-Invariant Feature, 11th International
Conference on Computer Modeling and Simulation
(UKSim 2009), pp 258-263, Mar 2009.

Cascade Algorithm of Image Watermarking Based On Multiple Wavelet Components

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Cascade Algorithm of Image Watermarking Based On Multiple Wavelet Components

Hochgeladen von

Copyright:

Verfügbare Formate

Council for Innovative Research

International Journal of Computers & Distributed Systems