International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169

Volume: 5 Issue: 12 137 – 141

_______________________________________________________________________________________________

A Novel DWT-CT approach in Digital Watermarking using PSO

Sachin Gupta Priyanka Mitra

Department of Computer Science & Engineering Department of Computer Science & Engineering
Apex Institute ofEngineering & Technology, Jaipur Jaipur Engineering College and Research Centre, Jaipur
[email protected] [email protected]

Abhigya Saxena Brij Kishore

Department of Computer Science & Engineering Department of Computer Science & Engineering
Apex Institute of Engineering & Technology, Jaipur Apex Institute of Engineering & Technology, Jaipur
[email protected] [email protected]

Abstract---The importance of watermarking is dramatically enhanced due to the promising technologies like Internet of Things (IoT), Data
analysis, and automation of identification in many sectors. Due to these reasons, systems are inter-connected through networking and internet
and huge amounts of information is generated, distributed and transmitted over the World Wide Web. Thus authentication of the information is a
challenging task. The algorithm developed for the watermarking needs to be robust against various attack such as salt & peppers, filtering,
compression and cropping etc. This paper focuses on the robustness of the algorithm by using a hybrid approach of two transforms such as
Contourlet, Discrete Wavelet Transform (DWT). Also, the Particle Swarm Optimization (PSO) is used to optimize the embedding strength
factor. The proposed digital watermarking algorithm has been tested against common types of image attacks. Experiment results for the
proposed algorithm gives better performance by using similarity metrics such as NCC (Normalized Cross Correlation value) and PSNR (Peak
Signal to Noise Ratio).

Keywords—watermarking, particle swarm optimization, discrete wavelet transfrom, contourlet transformation, PSNR, NCC.
__________________________________________________*****_________________________________________________

DWT, CLT and PSO is presented in Section 2. The related

I. INTRODUCTION work is discussed in Section 3. The proposed watermarking
In today‟s era, the development of the internet has technique is discussed in Section 4. The experimental results
introduced many new opportunities for the creation and are presented in Section 5 followed by comparison results in
delivery of information in digitized form including text, section 6. Finally Section 7 concluded the proposed research
images, audio, and video. Copyright protection of digital work.
information became essential due to the exponential growth of
the digital media access and editing over networks [1][2]. An
important issue that arises is the protection of rights of all II. PRELIMINARIES
participants. One of the protection mechanisms that attract A. Discrete Wavelet Transformation
many researchers is based on digital watermarking techniques.
Wavelet transforms are based on small waves, called
Digital watermarking is a process of information hiding which
wavelet, of varying frequency and limited duration. The
is used to hide secret information in multimedia data like
wavelet transform decomposes the image into four spatial sub-
digital images, text documents, audios, or video clips. In
bands, i.e. approximation, horizontal, vertical and diagonal.
specific, it acts as a digital signature, giving the image a sense
Hence wavelets reflect the anisotropic properties of HVS more
of ownership or authenticity.
precisely. Magnitude of DWT coefficients is larger in the
In general, a digital watermarking scheme should be
lowest bands (LL) at each level of decomposition and is
imperceptible and robust to attacks such as cropping, rotation,
smaller for other bands (HH, LH, and HL) [7]. One of the
filtering, compression, and others [3] which may cause the
main challenges of the watermarking problem is to achieve a
removal or substitution of the original watermark with another
better tradeoff between robustness and perceptivity.
watermark by the intruders. Commonly used image
Robustness can be achieved by increasing the strength of the
watermarking algorithms are spatial domain based schemes
embedded watermark, but the visible distortion would be
like Least Significant Bits (LSB) [4] and frequency domain
increased as well [11].
transformation schemes like Discrete Fourier Transform
(DFT) [5], Discrete Cosine Transform (DCT) [6], Discrete
Wavelet Transform (DWT) [7] , and Discrete Contourlet
Transform (CT) [8][9][10].
A brief review to understand the background of the
problem is discussed in section 1 and a quick overview about
137
IJRITCC | December 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 12 137 – 141
_______________________________________________________________________________________________

Xi(t+1) = Xi(t) + Vi (t+1) (2)

Steps Used PSO Algorithm

 Each particle initial position and initial velocity is
generated randomly.
 The velocity of all particles is updated according to
equation (1) and the position of all particles is
updated according to equation (2).
 If the fitness of the current particle is smaller than its
previous best (pbest) fitness, replace pbest by the
B. Contourlet Transformation update pbest.
Although wavelets are good at representing discontinuities pbest = p(t ) if f(p(t)) > pbest (3)
in one dimension or point singularities, it fails to represent  For each particle, if its fitness is smaller than the best
singularities in higher dimensions. The Discrete Contourlet one (gbest) of all the particles, update gbest.
Transform is a relatively new transform defined in the discrete gbest = g(t) if f(g(t)) > gbest (4)
form by Do et al. [8] to capture the edge information in all Where, f (t) is the objective function to be optimized.
directions. The main feature of this transform is the potential  The process is update till the certain termination
to efficiently handle 2-D singularities, i.e. edges. The CT can conditions are not founded. When the process
be divided into two main steps: Laplacian Pyramid (LP) terminates the pbest and gbest is determined.
decomposition [9] and Directional Filter Banks (DFB)
decomposition [10]. An image is first decomposed into low III. RELATED WORK
pass image and band pass image by LP decomposition. Each The author [13] proposes the application of Discrete
band pass image is further decomposed by DFB step. A DFB Wavelet Transform (DWT) into image watermarking by using
is designed to capture the high frequency content like smooth Particle Swarm Optimization (PSO) which is an evolutionary
contours and directional edges. Multi-resolution and multi- technique with the stochastic, population-based algorithm for
direction decomposition can be obtained by repeating the same solving problem. To protect copyright information of digital
steps mentioned above for the low pass image. images, the original image is decomposed according to two-
In Contourlet, the number of directional sub bands at each dimensional discrete wavelet transform. Subsequently the
level is set to 2n where n is a positive integer number. For preprocessed watermark with an affined scrambling transform
example, if we choose to decompose an image into four levels is optimal embedded into the vertical sub-band (HLm)
using n= (1, 2, 3, 4) then we get 2, 4, 8, and 16 sub bands as coefficients in wavelet domain without compromising the
shown in Figure 2. quality of the image. The scaling factors are trained with the
assistance of PSO.
The author [14] proposed a blind image watermarking
scheme based on wavelet tree quantization. Two largest
coefficients are selected as significant coefficients and the
difference between them is taken as significant difference. The
significant difference with an average significant difference
value and maximum difference coefficients are quantized for
embedding watermark bit.
In the work [15], an image-adaptive watermarking
algorithm based on wavelet transform was proposed. At first, a
digital image used as watermarking was scrambled. Next, the
original image was decomposed by discrete wavelet transform
and in accordance with the characteristics of human visual
system, wavelet decomposition in the low-frequency domain,
Methods which average of adjacent domain instead of single
wavelet decomposition coefficients was used to estimate and
quantitative, watermarking was adaptively embedded in
C. Particle Swarm Optimization wavelet coefficients of low-frequency domain.
It is a population based stochastic optimization model The author [16] have presented blind digital watermarking
which is generated after the social behavior of bird flocks or algorithm in which genetic algorithm is combined with
fish schooling developed by Kennedy and Elberhart [12]. The discrete cosine transform domain. In the proposed algorithm,
real particle swarm optimization method shows every particle firstly the image is divided into 8 x 8 blocks and each block is
like a potential solution of a problem in n – dimensional space. transformed to DCT domain. Now apply GA in each 8 x 8
The modification in particle local best and global best can be DCT block to find the best coefficients for insertion of a
defined as the velocity value. Velocity of each particle is watermark. For embedding the watermark, two DCT
upgraded by using the following equation (1) and the every coefficients, whose difference is large enough and they must
particle updated its position by equation (2) be in the low frequencies are chosen. If the distance is greater,
robustness increases.
Vi (t+1) = a Vi + const1* random * (pbest(t) – Xi(t)) + const2
* random * (gbest(t) – Xi(t)) (1)
138
IJRITCC | December 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 12 137 – 141
_______________________________________________________________________________________________
IV. THE COMBINED OPTIMIZED DWT-CT DIGITAL
WATERMARKING ALGORITHM
In our performance evaluation, we used the combined
DWT-CT transform in which gain factor is optimized by the
PSO. It comprises of two units; embedding algorithm and
extraction algorithm.
I. Watermark Embedding Unit
Before embedding the watermark into the original image
it will be transformed into coefficients by applying DWT and
CT. Original image also transformed into coefficients then
both are applied to the embedding algorithm which is known
as watermarked image, now inverse transforms are applied to
obtain the watermarked image. Selection of the embedding
factor is optimized by PSO. Fig. 3 shows the block diagram of
the watermark embedding algorithm.

V. EXPERIMENTAL RESULTS
In order to easily compare with other watermarking
techniques in the literature, a standard 256 x 256 grey scale
Lena image is used as the host image. The watermark image is
made from a binary loge image of size 128 x 128 pixels. In
contourlet decomposition, both LP decomposition and DFB
decomposition with „pkva‟ filters is used. The first two
pyramidal levels are chosen and the number of directional sub
bands for each level is set to 2, 4 and 8. In wavelet
decomposition, „haar‟ filters are used because of their high
efficiency. Testing the robustness is done by applying a
number of different attacks which are JPEG, median filtering,
average filtering, salt & pepper noise, cropping, image
sharpening, and histogram equalization.
II. Watermark Extraction Unit
Watermark extraction process deals with the retrieval of
the watermark in the presence of the original image. The Types of Lena Image
objective of the watermark extraction algorithm is to obtain attacks Watermarked Image Extracted Logo
the reliable an estimate of the original watermark from the
watermarked image. The extraction process is opposite of the
watermark embedding process. Fig. 4 shows the block Cropping
diagram of the watermark extraction algorithm.

JPEG

Sharpening

Rotation

139
IJRITCC | December 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 12 137 – 141
_______________________________________________________________________________________________
Average
Filtering

Median
Filtering

Gaussian
Figure 7 Comparison of NCC values without attacks
Table II PSNR Parameter values with attacks
Types of Attacks Existing Proposed
Salt & Method[16] Method
Pepper Salt and Pepper Noise 24.868 34.3960
Gaussian Noise 19.491 40.0398
Median Filtering 28.853 33.2168
Averaging Filtering 28.229 30.9815
Figure 5. Several attacks on “Lena” image and their
Sharpening 25.434 34.5457
corresponding extracted watermarks
JPEG Compression (50%) 26.579 35.1935
VI. COMPARISON RESULT JPEG Compression (70%) 33.051 39.3148
Here the results of the proposed method are compared with
Sonam et al. [16] method for Lena image. As it is shown in
the table, when the image has been corrupted by JPEG
compression, the proposed method provides more robust
watermark extraction when the quality factor is low. The
proposed method shows that the image remains imperceptible
(because PSNR> 30 dB) while the watermark survives from
attacks. The comparison is given in the table below:

Table I Parameter values without attacks

Error Existing Method [16] Proposed Method

Metrics Mandrill Leena Mandrill Leena
NCC 0.998 0.998 1.0000 1

PSNR 29.220 36.301 40.4277 40.0398

Figure 8 Comparison of PSNR values after attacks

Table III NCC Parameter values with attacks
Types of Attacks Existing Proposed
Method[16] Method
Salt and Pepper Noise 0.987 0.9953
Gaussian Noise 0.959 1
Median Filtering 0.997 1
Averaging Filtering 0.997 1
Sharpening 0.980 0.9956
Figure 6 Comparison of PSNR values without attacks JPEG Compression (50%) 0.989 0.9949
JPEG Compression (70%) 0.996 0.9980

140
IJRITCC | December 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 12 137 – 141
_______________________________________________________________________________________________
[11] C.I. Podilchuk, W.J. Zeng, "Image-adaptive watermarking using
visual models", IEEE Trans. on Selected Areas in
Communications, vol. 16, pp. 525–539, 1998.
[12] J. Kennedy, R.C. Eberhart, “Particle Swarm Optimization”, in
Proceedings of the IEEE International Conference of Neural
Networks (Perth, Australia), IEEE Service Center, Piscataway,
NJ, 1995, pp. 1942-1948.
[13] Hai Tao, JasniMohamad Zain, “A wavelet-based particle swarm
optimization algorithm for digital image watermarking”,
Integrated Computer-Aided Engineering 19 (2012) 81–91.
[14] J. George, S. Varma and M. Chatterjee, "Color image
watermarking using DWT-SVD and Arnold transform," Annual
IEEE India Conference (INDICON), 2014, pp. 1-6
[15] Tao Liu,” A kind of DWT Domain Image Digital Watermarking
Algorithm Research” China Advanced Materials Research Vols.
Figure 9 Comparison of NCC values after attacks 1044-1045 2014 pp 991-994.
[16] Sonam Tyagi, Harsh Vikram Singh, Raghav
Agarwal,”Image watermarking using genetic algorithm in DCT
VII. CONCLUSION domain” IEEE International Conference on Inventive Systems
In this paper, PSO based DWT-CT domain watermarking and Control (ICISC) Year: 2017 Pages: 1 - 6
algorithm has been proposed. Numerous tests were done on
watermarked image. The quality of watermarked image is
good in terms of perceptibility and the proposed method is
shown to be robust to all attacks which are performed.
Compared to results reported at [16], proposed method gives
improved NCC and show superiority over existing methods.
Further work on the proposed approach is in progress to
optimize parameters by using hybrid optimization
technique.Further work can be extended to implement the
algorithm on other formats of document rather than text.
REFRENCES
[1] F. Hartung, and M. Kutter, “Multimedia Watermarking
Techniques”, Proceedings of the IEEE, Special Issue on
Identification and Protection of Multimedia Information 87, pp
1079-1107, July 1999.
[2] V. M. Potdar, S. Han, and E. Chang, “A Survey of Digital Image
Watermarking Techniques”, 3rd International Conference on
Industrial Informatics, 2005
[3] F. Hartung, J.K. Su, and B. Girod, “Spread Spectrum
Watermarking: Malicious Attacks and Counter-Attacks,” Proc.
of SPIE Vol. 3657: Security and Watermarking of Multimedia
Contents, San Jose, CA, USA, Jan. 1999
[4] Macq B.M., Quisquater J.J., “Cryptology for digital TV
broadcasting”, Proceedings of IEEE, ISSN: 0018-9219, vol. 83,
pp. 944-957, June 1995.
[5] Ruanaidh J. J. K. O., Dowling W. J., Borland F. M., (1996),
"Phase watermarking of digital images," in Proc. IEEE Int.
Conf: Image Processing, pp. 239-242, Sept. 16-19.
[6] Zhao, J., and Koch, E. (1995, August) “Embedding robust labels
into images for copyright protection” Proceeding of the
International Congress on Intellectual Property Rights for
Specialized Information, Knowledge and New Technologies,
Vienna, pp. 1–10
[7] M. Barni, F. Bartolini, A. Piva, "Improved wavelet-based
watermarking through pixel-wise masking", Image Process.
IEEE Trans. Image Process. 10 (5), 2001, pp. 783–791.
[8] M. N. Do and M. Vetterli, “The contourlet transform: An efficient
directional multiresolution image representation,” IEEE Trans.
on Image Processing, vol. 14, No. 12, pp. 2091–2106, December
2005.
[9] P. J. Burt and E. H. Adelson. The laplacian pyramid as a compact
image codes. IEEE Trans. on Communications, 31:532– 540,
Apr. 1983
[10] R. H. Bamberger and M. J. T. Smith, “A filter bank for the
directional decomposition of images: Theory and design,” IEEE
Trans. on Signal Processing 40, pp. 882–893, April 1992.
141
IJRITCC | December 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________