A robust combination of dwt and chaotic function for image watermarking

International Journal of Current Trends in Engineering & Technology
ISSN: 2395-3152
Volume: 02, Issue: 02 (MAR-APR, 2016)
252
A Robust Combination of DWT and Chaotic Function for Image
Watermarking
Rajkumari Lalwani, S. R. Yadav, Avinash Sharma
Computer Science & engineering Department
Millennium Institute of Technology, Bhopal
1
rajkumarilalwai@gmail.com, 2
techmillenniumk.yadav@gmail.com, 3
avinashavi07@rediffmail.com
Abstract— with the increase in the digital media transfer
and modification of image is very easy. This independency
generate proprietorship problem of the user. So this paper
focuses on this problem of increasing the robustness of the
image against various attacks. Here new approach of
protecting watermark of the image is done by DCT
technique at selected blocks of chaotic function work
effectively for fulfilling the requirement. Experiment is
done on standard images and under compression attack.
Results shows that proposed work is better as compare to
previous existing research.
Keywords-- Color Format, Digital Watermarking,
Frequency domain, LSB.
I. INTRODUCTION
As digital world is growing drastically people are moving
towards different services provide by it. Some of this
service is social network, online market. But this technology
gives rise to new problem of piracy or in other words
proprietary get easily stolen. So to overcome these different
techniques are used for preserving the proprietary of the
owner. One of such digital approach is watermarking which
is a subsection of hiding information that is used to put
some information in the original image which will specify
the originality of the digital data like photographs, digital
music, or digital video [1, 2, 4]. One of the basic causes of
the copyright issue is the ease available of the internet and
some software that can modify the content as per the user
requirement. Watermark is broadly divide into two
categories first is visible watermarking and other is invisible
watermarking. Here watermark information seen by naked
eyes is considered as visible watermarking as shown in fig.
1. While in case of invisible watermark data is not visible by
naked eyes as shown in fig. 2, although watermark data is
present in the original data. Data may be of any digital
information like text file, image, video file, etc.
Requirement of invisible watermark is popular in
photography, movies, etc. So putting the signature in the
image for validating the data is done. Although invisible
embedding in carrier image is complex and challenging task
but different techniques are working in this field. Proposed
work will increase the hiding capacity and preserve carrier
as well as watermark information.
Figure 1 Visible watermark in image data
Figure 2 Visible watermarks in image data
II. RELATED WORK
J. Wu, and J.Xie [22] propose an adaptive watermarking
technique in DCT domain using HVS model and fuzzy c-
means technique (FCM). In this method FCM technique is
used to classify non-overlapping 88×original blocks into
categories: one is suitable for watermarking with high
imperceptibility and robustness and the other is unsuitable.
Watermark is inserted in DCT mid-frequency coefficients of
selected blocks. W. Zhang et el. [23] propose an adaptive
digital watermarking approach. In this method FCM
technique is used to determine the watermark strength of
each image pixel, and then watermark is inserted adaptively
to the N largest magnitude non-dc DCT coefficients of the
host image. The both the method performs better against
additive noise, compression and cropping etc. Yifei Pu. et
el. [24] proposes a public adaptive watermark algorithm m
for color images based on principal components analysis of

ISSN: 2395-3152
253
generalized Hebb. The algorithm is based on principal
component analysis of generalized Hebb adaptive algorithm
in Artificial Neural Network and to do adaptive quantitative
coding for principal component coefficients according to the
proportion of marginal or textural information of the
watermark image. In addition, it adaptively adjusts the
embedding depth according to the images features to ensure
the invisibility of the watermark. By way of disporting and
stochastic embedding into color image watermark, it
increases the embedding robust city of watermark. C.
Podilchuk, W. Zeng [21] propose a watermarking technique
for digital images that is based on utilizing visual models,
which have been developed in the context of image
compression. The visual model gives a direct way to
determine the maximum amount of watermark signal that
each portion of an image can tolerate without affecting the
visual quality of the image. The watermark encoding
scheme consists of a frequency decomposition based on an
8×8 framework followed by just noticeable difference
(JND) calculation and watermark insertion. The watermark
scheme is robust to different attacks such as JPEG
compression, additive noise, scaling etc.
III. PROPOSED METHODOLOGY
Read Carrier Image: - Here as the image is the collection
of pixels where each pixel is representing a number
that are reflecting a number over there now for each number
depend on the format it has its range such that for the gray
scale format it is in the range of 0-255. So read an image
means making a matrix of the same dimension of the image
then fill the matrix correspond to the pixel value of the
image at the cell in the matrix.
DWT (Discrete Wavelet Transform):- Apply DWT on pre-
processed image. As the image are modify in the low
frequency region so the effect of compression attack is very
less. If direct change is done at this level then chance of
watermark recovery get decrease and quality of video also
degrade as at high frequency region human can detect it
easily. Here whole image is dividing into four part name as
LL, LH, HL, HH band where LL band is low frequency
band.
Chaotic Function:- In this step original image from the
database is jumble by utilizing the chaotic matrix where
each pixel position is multiply by the matrix, then new
position is obtain for the pixel value. In similar fashion all
pixels of the image is randomize. Chaotic Matrix =
CM (Chaotic Matrix),  is variable range from 1, 2…n.
Let P is matrix represent [row, column], then multiple CM
and p, will give N matrix which is a new pixel position of
the older pixel. N=CM*P
Make Image Block: Here image is dividing into same size
of three blocks. Here size is same as 8X8 in this case 8
means number of pixel in the row and column should be
8X8=64 pixels. Now watermark is also dividing into same
set of blocks of 8X8 so that Euclidean distance between
those blocks can be calculated. Finally block which has
minimum distance from the carrier image is replaced.
Eludician Distance: - In this step distance between the
blocks of the image is calculated. This formula is same as
the Pythagoras theorem. Let X and Y is the image feature
for distance calculation. ))(( 2
YXsumD  So as per the D
value of the watermark block embedded block in carrier
image is selected.
Embedding Block: - As work is done on image so
embedding is done on the matrix of the image, so whole
operation of embedding is done by finding the minimum
distance image from the watermark block. As most
matching block have minimum distance and replace that
block with watermark block. In similar fashion other blocks
of the carrier image is replace. In order to increase the
transparency block size can be reduced.
IDWT (Inverse Discrete Wavelet Transform):- Finally
IDWT is applying on the embedded image this will
retransform the matrix in original form. This is necessary as
matrix is dividing as per LL, HL, LH, HH quadrants. After
IDWT all the blocks are combining back to single image.
3.1 Proposed Encryption Algorithm
Input: O [Original Image], w [Watermark]
Output: EI [Embedded Image]
Loop 1: C // c: Cycle of chaotic function.
[LL LH HL HH]DWT (O)
While distance between new LL and original LL < T
// T is Threshold
LLChaotic_function (LL)
If c= Embedding_position
BLL[n] Blocks (LL) // n number of blocks
BW[m] Blocks (W) // n number of blocks
Loop 1: m
Loop 1: n
D[n] Euclidean (BLL[n], BW[m])
xMin (D) // MSB three bit
BLL[n] BW[m]
End Loop
End Loop

ISSN: 2395-3152
254
LLCombine_Block (BW)
End While
EIIDWT (LL LH HL HH)
Figure 3 Block diagram of proposed Embedding Work
3.2 Extraction
In this step whole process of extraction is explained which
is almost same as done in embedding. Here input is
embedded image while output is extracted image. Here
DWT is applying on the embedded image then LL band is
select for rest of operations. Now LL band undergoes a
chaotic function where at fix number of rotation it get stop.
This position is same as done at embedding time. Now
matrix at this position is use for extraction of watermark
data. Divide matrix into same size of blocks as done in
embedding, and then find the replaced and non-replaced
blocks of the input matrices. So this extraction of data is
repeat in each block till all watermark information is not
extract from the matrix. Once watermark is extract then
remaining chaotic rotation is continue to get image in
original form.
IV. Experiment and Result
This section presents the experimental evaluation of the
proposed Embedding and Extraction technique for privacy
of image. All algorithms and utility measures were
implemented using the MATLAB tool. The tests were
performed on 2.27 GHz Intel Core i3 machine, equipped
with 4 GB of RAM, and running under Windows 7
Professional. Experiment done on the standard images such
as mandrilla, Lena, pirate, etc. Result is compare at two
condition first is without attack and other is at noise and
filter attack.
4.1 Evaluation Parameter
Peak Signal to Noise Ratio (PSNR):- it is use to find the
amount of data present from the received signal as it may
corrupt by the presence of some noise. So it is term as the
peak signal to noise ratio. PSNR is the ratio between the
maximum possible received information and the noise that
affects the fidelity of its representation.
PSNR = 





errorSquareMean
valuepixelMax
__
__
log10 10
figure 4 Represent dataset images.
Structural Similarity index: - SSIM term is a method for
finding the similarity between two images. The SSIM
method use for evaluating the image quality based on an
initial uncompressed or distortion-free image as reference. It
is introduce to improve the traditional schemes like PSNR
and MSE, which have proven to be inconsistent with human
eye perception.
Extraction Rate: - This is the reverse of the BER where
value is obtained by the ratio of the correct bits received
after extraction to the total number of bits embeds at the
sandier. The extraction rate η is defined as follows:
Where nc is the number of correctly extracted bits, and na is
the total number of embedded bits.
From given table 2 and 3 it is obtain those three images are
chosen for testing, in presence of ideal condition. It is
observed that proposed work is better as compare to the
Embedding method in [9]. As in proposed work watermark
information is add in the image separately. From given table
4, 5, 6 and 7 it is obtain that three images are chosen for
testing, in presence of different attack. It is observed that
proposed work is approximately same as compare to the
Embedding method in [9]. As in proposed work watermark
information is add in the DWT LL band of image
separately. From above tables it is obtain that ETA values of
the information received has high percentage of recovery in
presence of noise attack. While ETA values of the
information received has low percentage of recovery in
presence of filter attack.

ISSN: 2395-3152
255
Table 1 Values at ideal conditions of previous embedding
method [9].
Table. 2 Values at Ideal condition Proposed Work method.
Table. 3 Noise attack on proposed Embedding method.
Table. 4 Noise attack on previous Embedding method [9]
method.
Table. 5 Filter attack on proposed Embedding method.
Table. 6 Filter attack on previous Embedding method [9].
V. CONCLUSION
In this paper a new approach of watermarking is studied in
detail. . So this work has presented a novel approach of the
embedding and extraction of the digital message in the
image which is quit unique as well as different from the
various existing methods. Result shows that the proposed
work is producing the results which maintain the image
quality as well as robustness against the data insertion. As
research is a continuous process of achieving new heights so
in future one can apply same work for color images.
REFERENCES
[1]. Hanieh Khalilian, Student Member, IEEE, And Ivan V.
Bajic Video “Watermarking With Empirical PCA-Based
Decoding” IEEE Transactions On Image Processing,
Vol. 22, No. 12, December 2013.
[2]. Walter Godoy Jr., Charles Way Hun Fung “A Novel
DWT-SVD Video Watermarking Scheme Using Side
View” 978-1-4577-1180-0/11/$26.00 ©2011 IEEE.
[3]. Tamanna Tabassum, S.M. Mohidul Islam “A Digital
Image Watermarking Technique Based On Identical
Frame Extraction In 3-Level DWT” Vol. 13, No. 7, PP
560 –576, July 2003.
[4]. Frank Hartung, Jonathan K. Su, And Bernd Girod
“Spread Spectrum Watermarking: Malicious Attacks
And Counterattacks”. Of Multimedia Contents”
International Journal Of Research In Engineering And
Technology ISSN: 2319-1163.
[5]. Chapter 2. Wavelet Transforms On Images” Sundocb.
Bibliothek. Uni-Halle. De/ Diss-Online /02 /03h033
/T4.Pdf
[6]. Priya Porwal, Tanvi Ghag, Nikita Poddar, Ankita Tawde
Digital Video Watermarking Using Modified LSB And
DCT Technique. International Journal Of Research In
Engineering And Technology EISSN: 2319-1163.
[7]. Kazuki Yamato, Madoka Hasegawa, Yuichi Tanaka‡
And Shigeo Kato. “Digital Image Watermarking Method
Using Between-Class Variance”. 978-1-4673-2533-
2/12/$26.00 ©2012 IEEE.
[8]. Angela Piper1, Reihaneh Safavi-Naini. “Scalable Fragile
Watermarking For Image Authentication”. Published In
Information Security, On 31st December 2012.
[9]. Mohammadreza Keyvanpour, Farnoosh Merrikh-Bayat.
”An Effective Chaos-Based Image Watermarking
Scheme Using Fractal Coding”. 1877-0509 2010.

A robust combination of dwt and chaotic function for image watermarking

More Related Content

What's hot (16)

Viewers also liked (19)

Similar to A robust combination of dwt and chaotic function for image watermarking (20)

More from ijctet (20)

Recently uploaded (20)

A robust combination of dwt and chaotic function for image watermarking