Dual-Layer Digital Image Watermarking for Intellectual Property Right Protection

Full text

ACEEE Int. J. on Signal & Image Processing, Vol. 4, No. 2, May 2013
© 2013 ACEEE
DOI: 01.IJSIP.4.2.
Short Paper
1
Dual-layer Digital Image Watermarking for Intellectual
Property Right Protection
1Mr. H. E. Suryavanshi, 2Prof. Amit Mishra and 3Prof. Amit Sinhal
Department of Information Technology, Technocrats Institute of Technology, Bhopal, India
1hitendra.suryavanshi@gmail.com, 2amitmishra.mtech@gmail.com and 3amit_sinhal@rediffmail.com
Abstract: Digital watermarking is a technique which is widely
used in various application areas such as copyright protection,
copy control, broadcast monitoring etc. In this paper a novel
digital image watermarking technique is presented. To protect
the intellectual property rights, two different watermarks are
inserted in host image. The first watermark is used to monitor
the any changes made in host image while second one is used
as proof of ownership. The performance of proposed system is
good and it withstands against various attacks such as Gaussian
noise, salt and pepper, tampering etc.
Keywords: Digital Watermarking, Discrete Wavelet Transform,
Fragile Watermark, LSB, Robustness.
I. INTRODUCTION
Internet allows the public to exchange the information
without any barriers. This information can be in the form of
text, image, audio and video. The unrestricted access to
information gives birth to some problems such as copyright
violation, piracies etc. watermarking is a techniques developed
to resolve these issues. Watermarking gains a lot of
importance since last decade. It can be defined as a practice
of undetectably modifying a work to embed a message about
that work. Where work can be image, audio, video clip. [1]
In general, digital image watermarking system consist of
two modules: embedder and extractor, as shown in Fig. 1.
The embedder takes two inputs one is message which acts
as watermark, and the other is host image in which we want
to embed the message. The output of embedder is the
watermarked image which can be stored for later use.
Extraction module takes watermarked image and extracts the
message. Most of the extraction modules only check whether
image carries any watermark or not.
A. Properties of Watermarking System
Watermarking system has number of properties and the
importance of each property depends upon the type of
applications and role that watermark plays. [2]
1) Robustness: The ability of the watermark to survive normal
processing of content.
2) Security: The ability of the watermark to resist hostile
attacks.
3) Fidelity:The perceptual quality of watermarked content.
4) Data payload: The amount of information that can be
carried in awatermark.
B. Classification of Watermarking
Watermarking can be classified into the number of types,
as given below [3][4][5].
Fig. 1 General concept of watermarking
1) Visible Watermarking: The idea of visible watermark is very
simple. The logos used in today’s world everywhere is an
example of visible watermark. They are especially used for
conveying the immediate claim of ownership.
2) Invisible Watermarking: In this type of watermarking, rather
than displaying logo, the information is concealed into the
content itself.
3) Fragile Watermarking: Fragile watermarks have limited
robustness. They are used to check whether any modification
had taken place into the watermarked data.
Public Watermarking: These types of watermark are not secure
4) because they can read or viewed by anyone using specific
algorithms.
II. DISCRETE WAVELET TRANSFORM
Discrete wavelet transform is mathematical tool for
hierarchically decomposing an image. Images are usually
non-stationary two-dimensional signals and wavelet transform
is effective in such case. When discrete wavelet transformation
(DWT) applied on image, it decompose image into four
frequency sub-bands (LL, HL, LH, HH) where LL refers to low
pass band and other three sub-bands corresponds to
horizontal (HL), vertical (LH) and diagonal (HH) high pass
bands [6].
Fig.2 Two-level DWT decomposition
Fig. 2 shows two-level DWT decomposition of image. In
38

© 2013 ACEEE
DOI: 01.IJSIP.4.2.
ACEEE Int. J. on Signal & Image Processing, Vol. 4, No. 2, May 2013
Short Paper
1
general, the watermark can be inserted into low frequency
sub-bands (LL) because it increases the robustness of wa-
termark but at the same time it may degrade the image signifi-
cantly. High frequency bands (HH) contains edges and tex-
tures and changes that are caused due to watermark data
inserted in such band cannot be noticed by human eye [7].
III. PROPOSED WORK
In this section we present our proposed work. As
illustrated in Fig. 3, the watermarking system consists of two
embedders. The role of first embedder is to insert a watermark
into an image by using wavelet transform. This watermark
acts as robust watermark which cannot be removed from host
image. Second embeder inserts watermark which acts as fragile
watermark. This watermark is used to detect any alterations
in an image. Single change in an image can be easily identified
by using the second watermark. Spatial domain technique
such as LSB replacement is used to insert the second
watermark.
A. Watermark Insertion Method
1. Select the Original Image Ic [Mc, Nc] and Watermark Image
W1 [Mw, Nw]
2. Divide Ic and W1 into four sections [I1, I2, I3, I4] and [W1,
W2, W3, W4] by applying following equations
Part 1=Image (1 :( x/2), 1 :( y/2))
Part 2=Image (1 :( x/2), ((y/2) +1) :y)
Part 3=Image ((x/2) +1: x, 1 :( y/2))
Part 4=Image (((x/2) +1): x, ((y/2) +1): y)
Where [x, y] = Size of the Image
3. Select the pair (Ii, Wi) where i=1, 2, 3, 4 and repeat step no
4 to 13 for each pair
4. Apply DWT on Ii to get [CA, CH, CV, CD] coefficients
Fig.3 Proposed Waterma rking Scheme
5. Convert Wi into binary watermark (contains 0s and 1s)
6. Resize Wi as Wi [1, Mw×Nw]
7. Set PN sequence generator using Secrete Key-1
8. Repeat step 9 to 12 till (Mw×Nw)
9. Generate PN sequence
10. Calculate β factor as β=PN×K where, k is robustness
factor
11. Select the Coefficients [CH, CV]
12. IF Watermark (Wi) = 1
C=C+β
ELSE C=C
13. Apply inverse DWT to get Iwi
14. Collect all four section to get watermarked image (Iw)
15. Select the watermark image W2 [Mw, Nw]
16. Resize W2 to [Mc, Nc]
17. Set PN sequence generator using Key-2
18. Repeat the step 19 to 20 till [Mc × Nc]
19. Randomly generate a number R
20. IF R = Even Number
Set Iw (1st bit) to Watermark bit
ELSESet Iw (2nd bit) to Watermark bit
21. Final watermarked image (Iw) is obtained
22. Stop
B. Watermark Extraction Method
1. Select the Watermarked Image Iw [Mw, Nw]
2. Divide Iw into four sections [Iw1, Iw2, Iw3, Iw4] using same
equations specified above
3. Create four dummy matrix [M1, M2, M3, M4] containing all
zeros having size [Mw/2, Nw/2]
4. for each (Iwi, Mi) pair repeat step 5 to 9
where i=1, 2, 3, 4
39

ACEEE Int. J. on Signal & Image Processing, Vol. 4, No. 2, May 2013
© 2013 ACEEE
DOI: 01.IJSIP.4.2.
Short Paper
1
5. Apply DWT on Iwi to get [WA, WH, WV, WD]
coefficients
6. Set PN sequence generator using Key-1
7. Repeat step 8 till (Mw/2×Nw/2)
8. Select [WH, WV] coefficients and calculate correlation
9. IF Correlation >= Mean
Set Mi=1
ELSE Set Mi=0
10. Collect all Mi to get first watermark W1
11. Create a matrix W2 [Mw, Nw]
12. Set PN sequence generator using Key-2
13. Repeat the step 14 to 16
14. Randomly generate a number R
15. IF R = Even Number
Set W2 to Iw (1st bit)
ELSESet W2 to Iw (2nd bit)
16. Second watermark image (W2) is obtained
17. Stop
IV. RESULT ANALYSIS
This section presents the experimental results of the
proposed digital image watermarking scheme. For the entire
test in this paper MATLAB is used. The performance the
proposed method is tested on 8-bit grayscale image of
baboon, Lena and peppers of size 512×512. Fig 4. Shows the
cover image and two watermarks that can be inserted into
cover image.
(a) (b) (c)
(d) (e)
Fig.4 (a) Babbon (b) Lena (c) Peppers (d) First watermark (e)
Second watermark
The performance of the proposed watermarking technique
is evaluated in terms of the invisibility and robustness. The
PSNR (Peak-Signal-to-Noise Ratio) and MSE (Mean Square
Error) are used to measure the quality of the watermarked
image and attacked image. The PSNR is defined as follows
[6] [7]:
Where,
Where, I and I’ are cover image and watermarked image.
The normalized cross-correlation (NC) is used to check the
quality of original and extracted watermark. [8]
Where, W and W’ are original and extracted watermarks.
(a) (b) (c)
Fig.5 (a) Original Image (b) After inserting first watermark (c)
After inserting second watermark
Fig. 5 shows the original image of the baboon, the image
after the insertion of first watermark and second watermark.
The first watermark is inserted using blind watermarking
technique. It is a robust watermark. The second watermark is
a fragile watermark and embedded using LSB substitution
technique. Its purpose is to detect tampering.
(a) (b) (c)
Fig.6 (a) Watermarked image (b) Extracted first watermark (c)
Extra cted second watermark
The robustness of the proposed watermarking scheme is
tested against the various types of attacks such as, image
tampering, Salt and pepper, Gaussian and Poisson. It is shown
in the fig. 6 to 10.
(a) (b) (c)
Fig. 7 (a) Attacked ima ge (Tampering) (b) Extra cted first
watermark (c) Extracted second watermark
40

© 2013 ACEEE
DOI: 01.IJSIP.4.2.
ACEEE Int. J. on Signal & Image Processing, Vol. 4, No. 2, May 2013
Short Paper
1
(a) (b) (c)
Fig. 8 (a) Attacked image (Salt & Pepper) (b) Extracted first
watermark (c) Extracted second watermark
(a) (b)
Fig. 9 (a) Attacked image (Gaussian Noise) (b) Extracted first
watermark
(a) (b)
Fig. 10 (a) Attacked image (Poisson) (b) Extra cted first watermark
The performance of the system is shown in table 1 to 3 for
different images such as Baboon, Lena and Peppers. For
different values of K, robustness factor, different results are
obtained as given in tables below.
TABLE I. PERFORMANCE OF WATERMARKING SYSTEM
TABLE II. PERFORMANCE OF WATERMARKING SYSTEM
TABLE III. PERFORMANCE OF WATERMARKING SYSTEM
CONCLUSIONS
In this paper a novel digital image watermarking technique
is presented. This method is based on wavelet. The watermark
is inserted using wavelet coefficient blocks. Watermark
extraction process is independent on the original image.
Watermarks can be extracted in any order. This scheme is
tested against various attacks such as tampering, Gaussian
noise. In the future, we will try to enhance our algorithm to
obtain watermarked images with less distortion and to recover
the watermark with good accuracy.
ACKNOWLEDGMENT
The authors wish to thank Prof. Amit Mishra and Prof.
Amit Sinhal for their valuable guidance.
REFERENCES
[1] Ingemar J. Cox, Matthew L. Miller, Jeffrey A. Bloom, Jessica
Fridrich , Ton Kalker, “Digital watermarking and
steganography”, Second Edition, Morgan Kaufmann
Publishers, 2007.
[2] Husrev T. Sencar, Mahalingam Ramkumar, Ali N.
Akansu,”Data hiding fundamentals and applications”, Elsevier
Academic Press, 2004.
[3] Keshav S Rawat, Digital Watermarking Scheme for
Authorization against Copying or Piracy of Color Images,
Indian Journal of Computer Science and Engineering, Vol. 1,
No. 4, 2010, pp. 295-300
[4] Mohamed Abdulla Suhail, Digital watermarking for protection
of intellectual property, (University of Bradford, UK, 2005)
[5] Stefan Katzenbeisser, Fabien A. P. Petitcolas, Information
hiding techniques for steganography and digital watermarking,
Artech House Inc., 2000
[6] Peining Tao and Ahmet M. Eskicioglue, A robust multiple
watermarking scheme in the Discrete Wavelet Transformation
Domain, Proc. SPIE 5601, Internet Multimedia Management
Systems, Philadelphia, PA, 2004
[7] Ali Al-Haj,” Combined DWT-D CT digital image
watermarking”, in Journal of Computer Science 3(9): 740-
746, 2007, ISSN 1549-3636, © 2007 Science Publications
[8] Hanaa A. Abdallah et. Al. “Blind wavelet-based image
watermarking”, in International Journal of Signal Processing,
Image Processing and Pattern Recognition, Vol. 4, No. 1, March
2011.
K Image-1: Babbon.bmp
PSNR MSE NC
0.1 38.6712 7.9527 0.5429
0.2 32.8323 31.5237 0.5968
0.3 29.4246 70.7842 0.6508
0.4 27.0669 125.7632 0.6952
0.5 25.1983 196.4499 0.7333
0.6 23.6141 282.9246 0.7651
0.7 22.2814 384.5371 0.7841
0.8 21.1269 501.6401 0.8190
0.9 20.1092 634.1065 0.8413
1.0 19.1963 782.4404 0.8603
K Image-2: Lena.bmp
PSNR MSE NC
0.1 38.8532 7.9446 0.5841
0.2 33.0062 31.5301 0.6952
0.3 29.6291 70.8231 0.7651
0.4 27.1323 125.8479 0.8254
0.5 25.1939 196.6462 0.8730
0.6 23.6085 283.2924 0.8889
0.7 22.2732 385.2623 0.9270
0.8 21.1145 503.0735 0.9524
0.9 20.0935 636.4056 0.9651
1.0 19.1777 785.8025 0.9810
K Image-3: Peppers.bmp
PSNR MSE NC
0.1 38.1598 7.9413 0.5175
0.2 32.3653 31.4904 0.6159
0.3 29.0096 70.5561 0.6952
0.4 26.7119 124.8649 0.7460
0.5 24.9696 194.2827 0.8032
0.6 23.5794 278.5260 0.8413
0.7 22.3656 377.1564 0.8857
0.8
21.2234
490.6154
0.9111
0.9 20.2171 618.5500 0.9429
1.0 19.3119 761.8840 0.9460
41