This action might not be possible to undo. Are you sure you want to continue?
Journal of Computer Science 3 (9): 740746, 2007
ISSN 15493636
© 2007 Science Publications
Corresponding Author: Ali AlHaj, Department of Computer Engineering, School of Electrical Engineering,
Princess Sumaya University for Technology, PO Box 1928, AlJubeiha,
11941 Amman, Jordan
740
Combined DWTDCT Digital Image Watermarking
Ali AlHaj
Department of Computer Engineering, School of Electrical Engineering,
Princess Sumaya University for Technology, PO Box 1928, AlJubeiha,
11941 Amman, Jordan
Abstract: The proliferation of digitized media due to the rapid growth of networked multimedia
systems, has created an urgent need for copyright enforcement technologies that can protect copyright
ownership of multimedia objects. Digital image watermarking is one such technology that has been
developed to protect digital images from illegal manipulations. In particular, digital image
watermarking algorithms which are based on the discrete wavelet transform have been widely
recognized to be more prevalent than others. This is due to the wavelets' excellent spatial localization,
frequency spread, and multiresolution characteristics, which are similar to the theoretical models of
the human visual system. In this paper, we describe an imperceptible and a robust combined
DWTDCT digital image watermarking algorithm. The algorithm watermarks a given digital image
using a combination of the Discrete Wavelet Transform (DWT) and the Discrete Cosine Transform
(DCT). Performance evaluation results show that combining the two transforms improved the
performance of the watermarking algorithms that are based solely on the DWT transform.
Key words: Digital image watermarking, image copyright protection, frequencydomain
watermarking, Discrete Wavelet Transform (DWT), Discrete Cosine Transform (DCT)
INTRODUCTION
The development of effective digital image
copyright protection methods have recently become an
urgent and necessary requirement in the multimedia
industry due to the everincreasing unauthorized
manipulation and reproduction of original digital
objects. The new technology of digital watermarking
has been advocated by many specialists as the best
method to such multimedia copyright protection
problem
[1,2]
. Its expected that digital watermarking
will have a widespan of practical applications such as
digital cameras, medical imaging, image databases, and
videoondemand systems, among many others
[3]
.
In order for a digital watermarking method to be
effective it should be imperceptible, and robust to
common image manipulations like compression,
filtering, rotation, scaling cropping, collusion attacks
among many other digital signal processing operations.
Current digital image watermarking techniques can be
grouped into two major classes: spatialdomain
and frequencydomain watermarking techniques
[4]
.
Compared to spatial domain techniques
[5]
,
frequencydomain watermarking techniques proved to
be more effective with respect to achieving the
imperceptibility and robustness requirements of digital
watermarking algorithms
[6]
.
Commonly used frequencydomain transforms
include the Discrete Wavelet Transform (DWT), the
Discrete Cosine Transform (DCT) and Discrete Fourier
Transform (DFT). However, DWT
[7]
has been used in
digital image watermarking more frequently due to its
excellent spatial localization and multiresolution
characteristics, which are similar to the theoretical
models of the human visual system
[8]
. Further
performance improvements in DWTbased digital
image watermarking algorithms could be obtained by
combining DWT with DCT
[9]
. The idea of applying
two transform is based on the fact that combined
transforms could compensate for the drawbacks of each
other, resulting in effective watermarking.
In this paper, we will describe a digital image
watermarking algorithm based on combining two
transforms; DWT and DCT. Watermarking is done by
altering the wavelets coefficients of carefully selected
DWT subbands, followed by the application of the
DCT transform on the selected subbands.
J. Computer Sci., 3 (9): 740746, 2007
741
THE DCT AND DWT TRANSFORMS
The DCT and DWT transforms have been
extensively used in many digital signal processing
applications. In this section, we introduce the two
transforms briefly, and outline their relevance to the
implementation of digital watermarking.
The DCT transform: The discrete cosine transforms is
a technique for converting a signal into elementary
frequency components
[9]
. It represents an image as a
sum of sinusoids of varying magnitudes and
frequencies. With an input image, x, the DCT
coefficients for the transformed output image, y, are
computed according to Eq. 1 shown below. In the
equation, x, is the input imagehaving N x M pixels,
x(m,n) is the intensity of the pixel in row m and column
n of the image, and y(u,v) is the DCT coefficient in row
u and column v of the DCT matrix.
(1)
where
u = 0
u = 1,2,…, N − 1
v = 0
v = 1,2,…, N − 1
The image is reconstructed by applying inverse
DCT operation according to Eq. 2:
(2)
The popular blockbased DCT transform segments
an image nonoverlapping blocks and applies DCT to
each block. This results in giving three frequency
subbands: low frequency subband, midfrequency
subban and high frequency subband. DCTbased
watermarking is based on two facts. The first fact is
that much of the signal energy lies at lowfrequencies
subband which contains the most important visual
parts of the image. The second fact is that high
frequency components of the image are usually
removed through compression and noise attacks. The
watermark is therefore embedded by modifying the
coefficients of the middle frequency subband so
that the visibility of the image will not be affected
and the watermark will not be removed by
compression
[10, 11,12, 13]
.
The DWT transform: Wavelets are special functions
which, in a form analogous to sines and cosines in
Fourier analysis, are used as basal functions for
representing signals
[7]
. For 2D images, applying DWT
corresponds to processing the image by 2D filters in
each dimension. The filters divide the input image into
four nonoverlapping multiresolution subbands LL1,
LH1, HL1 and HH1. The subband LL1 represents the
coarsescale DWT coefficients while the subbands
LH1, HL1 and HH1 represent the finescale of DWT
coefficients. To obtain the next coarser scale of
wavelet coefficients, the subband LL1 is further
processed until some final scale N is reached. When N
is reached we will have 3N+1 subbands consisting of
the multiresolution subbands LL
N
and LHx, HLx and
HHx where x ranges from 1 until N.
Due to its excellent spatiofrequency localization
properties, the DWT is very suitable to identify the
areas in the host image where a watermark can be
embedded effectively. In particular, this property
allows the exploitation of the masking effect of the
human visual system such that if a DWT coefficient is
modified, only the region corresponding to that
coefficient will be modified. In general most of the
image energy is concentrated at the lower frequency
subbands LLx and therefore embedding watermarks in
these subbands may degrade the image significantly.
Embedding in the low frequency subbands, however,
could increase robustness significantly. On the other
hand, the high frequency subbands HHx include the
edges and textures of the image and the human eye is
not generally sensitive to changes in such subbands.
This allows the watermark to be embedded without
being perceived by the human eye. The compromise
adopted by many DWTbased watermarking algorithm,
is to embed the watermark in the middle frequency
ubbands LHx and HLx where acceptable
performance of imperceptibility and robustness could
be achieved
[14,15,16,17,18,19,20]
.
( ) ( )
( ) ( )
M 1 N 1
u v
u 0 v 0
2 2
y u, v x m, n
M N
2m 1 u 2n 1 v
cos cos
2M 2N
− −
= =
= α α
+ ∏ + ∏
¿¿
u
v
1
2
1
1
2
1
=
¦
¦
¦
α
´
¦
¦
¹
¦
¦
¦
α =
´
¦
¦
¹
( ) ( )
( ) ( )
M 1 N 1
u v
u 0 v 0
2 2
x m, n y u, v
M N
2m 1 u 2n 1 v
cos cos
2M 2N
− −
= =
= α α
+ ∏ + ∏
¿¿
J. Computer Sci., 3 (9): 740746, 2007
742
Fig. 1: Combined DWTDCT watermark embedding procedure.
(a) (b)
Fig. 2: Multiresolution DWT subbands of the original
image
THE COMBINED DCTDWT ALGORTIHM
The watermark embedding procedure is depicted
in Fig. 1 followed by a detailed explanation.
Step 1: Apply DWT to decompose the cover host
image into four nonoverlapping multiresolution
subbands: LL
1
, HL
1
, LH
1
, and HH
1
.
Step 2: Apply DWT again to subband HL
1
to get four
smaller subbands and choose the HL
2
subband as
shown in Fig. 2 a. Or, apply DWT to subband HH
1
to
get four smaller subbands and choose the HH
2
subband as shown in Fig. 2 b
.
Step 3: Divide the subband HL
2
(or HH
2
) into 4 x 4
blocks.
Step 4: Apply DCT to each block in the chosen
subband (HL
2
or HH
2
) .
Step 5: Reformulate the greyscale watermark image
into a vector of zeros and ones.
Step 6: Generate two uncorrelated pseudorandom
sequences. One sequence is used to embed the
watermark bit 0 (PN_0) and the other sequence is sued
to embed the watermark bit 1 (PN_1). Number of
elements in each of the two pseudorandom sequences
must be equal to the number of midband elements of
the DCTtransformed DWT subbands.
Step 7: Embed the two pseudorandom sequences, PN_0
and PN_1, with a gain factor , in the DCT transformed
4x4 blocks of the selected DWT subbands of the host
image.
Embedding is not applied to all coefficients of the
DCT block, but only to the midband DCT
coefficients. If we donate X as the matrix of the mid
band coefficients of the DCT transformed block, then
embedding is done as follows:
If the watermark bit is 0 then
X
'
= X
+ * PN_0 (3)
otherwise,
if the watermark bit is 1 then,
X
'
= X
+ * PN_1 (4)
Step 8: Apply inverse DCT (IDCT) to each block after
its midband coefficients have been modified to embed
the watermark bits as described in the previous step.
Step 9: Apply the inverse DWT (IDWT) on the DWT
transformed image, including the modified subband, to
produce the watermarked host image.
The watermark extraction procedure is depicted in
Fig. 3, and described in details in the following steps.
The combined DWTDCT algorithm is a blind
J. Computer Sci., 3 (9): 740746, 2007
743
Fig. 3: Combined DWTDCT watermark extraction procedure
watermarking algorithm, and thus the original host
image is not required to extract the watermark.
Step 1: Apply DWT to decompose the watermarked
image into four nonoverlapping multiresolution sub
bands: LL
1
, HL
1
, LH
1
, and HH
1
.
Step 2: Apply DWT to HL
1
to get four smaller sub
bands, and choose the subband HL
2
, as shown in
Fig. 2 a. Or, apply DWT to the HH
1
subband to get
four smaller subbands, and choose the HH
2
subband
,
as shown in Fig. 2 b.
Step 3: Divide the subban HL
2
(or HH
2)
into 4×4
blocks.
Step 4: Apply DCT to each block in the chosen
subband (HL
2
or HH
2
), and extract the midband
coefficients of each DCT transformed block.
Step 5: Regenerate the two pseudorandom sequences
(PN_0 and PN_1) using the same seed used in the
watermark embedding procedure.
Step 6: For each block in the subband HL
2
(or HH
2
)
,
calculate the correlation between the midband
coefficients and the two generated pseudorandom
sequences (PN_0 and PN_1). If the correlation with
the PN_0 was higher than the correlation with PN_1,
then the extracted watermark bit is considered 0,
otherwise the extracted watermark is considered 1.
Fig. 4: 'Lina' host image
Fig. 5: Original watermark
Step 7: Reconstruct the watermark using the extracted
watermark bits, and compute the similarity between the
original and extracted watermarks.
PERFORMANCE EVALUATION
We evaluated the performance of the combined
DWTDCT image watermarking algorithms using a
512×512 'Lena' as the original cover host image, and a
256×256 greyscale image of the expression 'copyright'
as the watermark image. The two images are shown in
Fig. 4 and 5, respectively.
Performance Evaluation Metrics: Watermarking
algorithms are usually evaluated with respect to two
metrics: imperceptibility and robustness
[21]
. The two
metrics are described below.
J. Computer Sci., 3 (9): 740746, 2007
744
Table 1: The 'Lina' host image watermarked at different 2DWT subbands
Watermarked Image

DWTOnly DWTDCT DWTOnly DWTDCT
Original Image (HL2) (HL2) (HH2) (HH2)
PSNR = 80.190 PSNR = 97.072 PSNR = 77.098 PSNR = 97.083
Imperceptibility: Imperceptibility means that the
perceived quality of the host image should not be
distorted by the presence of the watermark. As a
measure of the quality of a watermarked image, the
peak signal to noise ratio (PSNR) is typically used.
PSNR in decibels (dB) is given below in Eq. 5.
(5)
Robustness: Robustness is a measure of the immunity
of the watermark against attempts to remove or degrade
it, internationally or unintentionally, by different types
of digital signal processing attacks
[22]
. In this chapter,
we will report on robustness results which we obtained
for three major digital signal processing operations
(attacks): Gaussian noise, image compression and
image cropping. The three attacks are a few, however,
they are good representatives of the more general
attacks. That is the Gaussian noise is a watermark
degrading attack, JPEG compression is a watermark
removal attack and cropping is a watermark dis
positioning geometrical attack. We measured the
similarity between the original watermark and the
watermark extracted from the attacked image using the
correlation factor ρ given below in Eq. 6.
(6)
where N is the number of pixels in watermark, w
and wˆ are the original and extracted watermarks
respectively. The correlation factor ρ may take values
between 0 (random relationship) to 1 (perfect linear
relationship). In general, a correlation coefficient of
about 0.75 or above is considered acceptable.
Performance results: We carried out the watermark
embedding algorithm described in the previous section.
Applying the 1level DWT of the 'Lina' image produced
four 256256 subbands; LL
1
, LH
1
, HL
1
and HH
1
. Since
embedding the watermark beyond the first DWT level
is more effective, we decided to embed the watermark
in HL
2
(or HH
2
). The selected 128×128 subband was
divided into 4×4 blocks giving a total of 1024 blocks.
The DCT transform was then applied to each block in
the chosen subband, after which the watermark was
embedded according to Eq. 3 and 4.
Imperceptibility: We evaluated imperceptibility of the
combined DWTDCT algorithm by measuring PSNR
for the HL
2
and HH
2
subbands. HL
2
gave a PSNR
value of 97.072 while HH
2
gave 97.083. The difference
is small between the two PSNR values indicating
comparable performance of the two subbands. In
contrast, and as shown in Table 1, the difference
between the PSNR values of the DWTOnly and the
combined DWTDCT algorithms is relatively large.
This indicates that improvement in imperceptibility can
be achieved by applying DCT on a DWT transformed
HL
2
(or HH
2
) subband.
Robustness: Table 2 shows the correlation values
between the original watermark and the watermarks
extracted from subbands HL
2
and HH
2
after being
subjected to different attacks, independently.
The correlation values given in Table 2 show
clearly that the combined DWTDCT watermarking
algorithm outperforms the conventional DWTOnly
approach with respect to robustness against the
Gaussian noise and cropping attacks. The results are
better regardless of whether the watermark was
embedded in HL
2
or HH
2
, however, HH
2
gave better
2
I
dB
10
I
10
MAX
PSNR 10 log
MSE
MAX
20 log
MSE
 
= ⋅

\ .
 
= ⋅

\ .
¿ ¿
¿
= =
=
=
N
i
i
N
i
i
N
i
i i
w w
w w
w w
1
2
1
2
1
ˆ
ˆ
) ˆ , ( ρ
J. Computer Sci., 3 (9): 740746, 2007
745
Table 2: Correlation values due to three different attacks
Correlation Factor

Fausain (Noise) Compression (Quality) Cropping (Block Size)
  
0 0.4 0.8 0 40 80 8 152 200
DWTOnly
(HL2) 0.736 0.660 0.445 0.478 0.733 0.741 0.746 0.623 0.550
DWTDCT
(HL2) 0.9738 0.8359 0.5029 0.012 0.604 0.958 0.968 0.778 0.549
DWTOnly
(HH2) 0.726 0.652 0.467 0.447 0.698 0.726 0.723 0.643 0.574
DWTDCT
(HH2) 0.989 0.868 0.546 0.023 0.065 0.314 0.989 0.884 0.696
Fig. 6: Extracted watermarks from the cropped HH2watermarked image using DWT only
Fig. 7: Extracted watermarks from the cropped HH2wateramrked image using DWT and DCT
Fig. 8: Extracted watermarks from the compressed HL2watermarked image using DWT only
Fig. 9: Extracted watermarks from the compressed HL2watermarked image using DWT and DCT
robustness against cropping compared with HL
2
.
Figure 6 shows the extracted watermarks from HH
2
when DWTOnly was used. In contrast, the better
combined DWTDCT results are shown in Figure 7
which shows the extracted watermarks from the HH
2
subband.
As given in Table 2 and illustrated in Fig. 8 and 9,
the robustness of the DWTOnly approach against the
JPEG compression attack is better than that of the
combined DWTDCT algorithm. This is due to the fact
that the DCT is the core component of the JPEG
compression, and therefore applying DCT at a DCT
transformed image degraded the quality of the
embedded watermark greatly.
RESULTS AND DISCUSSION
We described the performance of the combined
DWTDCT watermarking algorithm. For the sake of
comparison, we also evaluated the watermarking
performance when DWTOnly was used. The results
we obtained for the DWTOnly approach indicated a
better imperceptibility performance was obtained when
the watermark was embedded in the HL
2
or HH
2
sub
bands. The robustness performance, however, was not
acceptable. To improve performance, we combined
DWT with the another equally powerful transform; the
DCT. The combined DWTDCT watermarking
algorithm's imperceptibility performance was better
J. Computer Sci., 3 (9): 740746, 2007
746
than the performance of the DWTOnly approach.
Moreover, the improvement in robustness brought by
the combined DWTDCT algorithm was considerably
high.
CONCLUSIONS
The discrete wavelet transform (DWT) and the
discrete cosine transform (DCT) have been applied
successfully in many in digital image watermarking. In
this paper, we described a combined DWTDCT digital
image watermarking algorithm. Watermarking was
done by embedding the watermark in the first and
second level DWT subbands of the host image,
followed by the application of DCT on the selected
DWT subbands. The combination of the two
transforms improved the watermarking performance
considerably when compared to the DWTOnly
watermarking approach. In conclusion, in DWTbased
digital watermarking applications, combining
appropriate transforms with the DWT may have a
positive impact on performance of the watermarking
system.
REFERENCES
1. Cox, I., M. Miller and J. Bloom, 2002. Digital
Watermarking, Academic Press, USA.
2. Langelaar, G., I. Setyawan and R. Lagendijk, 2000.
"Watermarking Digital Image and Video Data: A
StateofArt Overview," IEEE Signal Processing
Magazine, 17(5):2046.
3. Arnold, M., M. Schumucker and S. Wolthusen,
2003. Techniques and Applications of Digital
Watermarking and Content Protection. Artech
House, USA.
4. Potdar, V., S. Han and E. Chang, 2005. A Survey
of Digital Image Watermarking Techniques, in
Proc. of the IEEE International Conference on
Industrial Informatics, pp: 709716, Perth,
Australia.
5. Chan, C. and L. Cheng, 2004. Hiding Data in
Images by Simple LSB Substitution, Pattern
Recognition, 37(3):469474.
6. Wang, R., C. Lin and J. Lin, " Copyright protection
of digital images by means of frequency domain
watermarking," Proc. of the SPIE Conference On
Mathematics of Data/Image Coding, Compression,
and Encryption, USA.
7. Vetterli, M. and J. Kovaevi, 1995. Wavelets and
Subband Coding. Prentice Hall, USA.
8. Wolfgang, R., C. Podilchuk and E. Delp, 1999.
"Perceptual Watermarks for Digital Images and
Video," Proc. of the IEEE, vol. 87, no. 7,
pp: 11081126.
9. Rao, K. and P. Yip. Discrete Cosine Transform:
algorithms, advantages, applications. Academic
Press, USA, 1990.
10. Chu, W, 2003. "DCTBased Image Watermarking
Using Subsampling," IEEE Trans. Multimedia,
5(1): 3438.
11. Lin, S. and C. Chin, 2000. "A Robust DCTbased
Watermarking for Copyright Protection," IEEE
Trans. Consumer Electronics, 46(3): 415421.
12. Deng, F. and B. Wang, 2003. "A novel technique
for robust image watermarking in the DCT
domain," in Proc. of the IEEE 2003 Int. Conf. on
Neural Networks and Signal Processing, vol. 2,
pp: 15251528.
13. Wu, C. and W. Hsieh, 2000. "Digital watermarking
using zerotree of DCT," IEEE Trans. Consumer
Electronics , vol. 46, no. 1, pp: 8794.
14. Hsieh, M., D. Tseng, and Y. Huang, 2001. "Hiding
Digital Watermarks Using Multiresolution Wavelet
Transform," IEEE Trans. on Industrial Electronics,
48(5): 875882.
15. Reddy, A. and B. Chatterji, 2005. "A New Wavelet
Based Logowatermarking Scheme," Pattern
Recognition Letters, 26(7): 10191027.
16. Wang, S. and Y. Lin, 2004. "Wavelet Tree
Quantization for Copyright Protection
Watermarking," IEEE Trans. Image Processing,
vol. 13, no. 2, pp: 154164.
17. Hsu, C. and J. Wu, 1998. "Multiresolution
Watermarking for Digital Images," IEEE Trans. on
Circuits and Systems II, 45(8): 10971101.
18. Tay, P. and J. Havlicek, 2002. "Image
Watermarking Using Wavelets," in Proc. of the
IEEE Midwest Symposium on Circuits and
Systems, pp: 258261, Oklahoma, USA.
19. Nikolaidis, A. and I. Pitas, 2003. "Asymptotically
optimal detection for additive watermarking in the
DCT and DWT domains," IEEE Trans. Image
Processing , 2(10): 563571.
20. Tsai, M. and H. Hung, 2005. "DCT and DWT
based Image Watermarking Using Subsampling,"
in Proc. of the 2005 IEEE Fourth Int. Conf. on
Machine Learning and Cybernetics, pp: 53085313,
China.
21. Ejima, M. and A. Myazaki, 2001." On the
evaluation of performance of digital watermarking
in the frequency domain," in Proc. of the IEEE
Int. Conf. on Image Processing, 2: 546549.
22. Voloshynovskiy, S., S. Pereira and T. Pun, 2001.
"Attacks on Digital Watermarks: Classification,
EstimationBased Attacks, and Benchmarks,"
Comm. Magazine, 39(8): 118126.
is to embed the watermark in the middle frequency ubbands LHx and HLx where acceptable performance of imperceptibility and robustness could be achieved[14. Computer Sci. In particular. LH1.19.2. y ( u. v ) N u =0 v = 0 2N ( 2m + 1) u ∏ cos ( 2n + 1) v ∏ cos (2) The popular blockbased DCT transform segments an image nonoverlapping blocks and applies DCT to each block. DCTbased . HL1 and HH1 represent the finescale of DWT coefficients. 1 shown below.18. The DCT transform: The discrete cosine transforms is a technique for converting a signal into elementary frequency components [9]. v ) = 2 M M −1 N −1 2 αuαv x ( m. x. On the other hand. and outline their relevance to the implementation of digital watermarking. could increase robustness significantly.…. To obtain the next coarser scale of wavelet coefficients. x. is the input imagehaving N x M pixels. are computed according to Eq. n ) = 2M 2 M 2 M −1 N −1 α u α v y ( u. only the region corresponding to that coefficient will be modified. Due to its excellent spatiofrequency localization properties.…. In the equation.12. N − 1 v=0 v = 1.J. The subband LL1 represents the coarsescale DWT coefficients while the subbands LH1. The DWT transform: Wavelets are special functions which.16. This results in giving three frequency subbands: low frequency subband. this property allows the exploitation of the masking effect of the human visual system such that if a DWT coefficient is modified. 2: x ( m.20]. For 2D images. The filters divide the input image into four nonoverlapping multiresolution subbands LL1. n ) N u = 0 v =0 watermarking is based on two facts. midfrequencysubban and high frequency subband. are used as basal functions for representing signals[7].v) is the DCT coefficient in row u and column v of the DCT matrix. The compromise adopted by many DWTbased watermarking algorithm. It represents an image as a sum of sinusoids of varying magnitudes and frequencies. the subband LL1 is further processed until some final scale N is reached. This allows the watermark to be embedded without being perceived by the human eye. In general most of the image energy is concentrated at the lower frequency subbands LLx and therefore embedding watermarks in these subbands may degrade the image significantly. we introduce the two transforms briefly. x(m. the DWT is very suitable to identify the areas in the host image where a watermark can be embedded effectively. y.2. With an input image. 3 (9): 740746.17. HLx and HHx where x ranges from 1 until N. 741 ( 2m + 1) u ∏ cos ( 2n + 1) v ∏ cos 2M 2N (1) where αu = 1 2 1 1 2 1 u=0 u = 1. 2007 THE DCT AND DWT TRANSFORMS The DCT and DWT transforms have been extensively used in many digital signal processing applications. The second fact is that high frequency components of the image are usually removed through compression and noise attacks. the DCT coefficients for the transformed output image. and y(u. in a form analogous to sines and cosines in Fourier analysis. When N is reached we will have 3N+1 subbands consisting of the multiresolution subbands LLN and LHx. The first fact is that much of the signal energy lies at lowfrequencies subband which contains the most important visual parts of the image. however.15. 13]. applying DWT corresponds to processing the image by 2D filters in each dimension. Embedding in the low frequency subbands. The watermark is therefore embedded by modifying the coefficients of the middle frequency subband so that the visibility of the image will not be affected and the watermark will not be removed by compression[10. In this section.n) is the intensity of the pixel in row m and column n of the image. HL1 and HH1. the high frequency subbands HHx include the edges and textures of the image and the human eye is not generally sensitive to changes in such subbands. N − 1 αv = The image is reconstructed by applying inverse DCT operation according to Eq.. 11.
2 b. Computer Sci. Step 3: Divide the subband HL2 (or HH2 ) into 4 x 4 blocks. 3 (9): 740746. in the DCT transformed 4x4 blocks of the selected DWT subbands of the host image. Step 9: Apply the inverse DWT (IDWT) on the DWT transformed image. 2: Multiresolution DWT subbands of the original image THE COMBINED DCTDWT ALGORTIHM The watermark embedding procedure is depicted in Fig. then embedding is done as follows: If the watermark bit is 0 then X '= X + * PN_0 (3) (a) (b) Fig. HL1. 1 followed by a detailed explanation.. 3. including the modified subband. PN_0 and PN_1. 1: Combined DWTDCT watermark embedding procedure. 2 a. if the watermark bit is 1 then. with a gain factor . 2007 Fig. The watermark extraction procedure is depicted in Fig. Step 4: Apply DCT to each block in the chosen subband (HL2 or HH2) . to produce the watermarked host image. LH1. apply DWT to subband HH1 to get four smaller subbands and choose the HH2 subband as shown in Fig. Step 6: Generate two uncorrelated pseudorandom sequences. The combined DWTDCT algorithm is a blind . One sequence is used to embed the watermark bit 0 (PN_0) and the other sequence is sued to embed the watermark bit 1 (PN_1). If we donate X as the matrix of the midband coefficients of the DCT transformed block. and described in details in the following steps. Embedding is not applied to all coefficients of the DCT block. Step 5: Reformulate the greyscale watermark image into a vector of zeros and ones. Step 2: Apply DWT again to subband HL1 to get four smaller subbands and choose the HL2 subband as shown in Fig. and HH1. Step 1: Apply DWT to decompose the cover host image into four nonoverlapping multiresolution subbands: LL1. Step 7: Embed the two pseudorandom sequences. X '= X + * PN_1 (4) Step 8: Apply inverse DCT (IDCT) to each block after its midband coefficients have been modified to embed the watermark bits as described in the previous step.J. but only to the midband DCT coefficients. 742 otherwise. Number of elements in each of the two pseudorandom sequences must be equal to the number of midband elements of the DCTtransformed DWT subbands. Or.
5: Original watermark Step 7: Reconstruct the watermark using the extracted watermark bits. 743 Fig. 4: 'Lina' host image Fig. and thus the original host image is not required to extract the watermark. Or. LH1. and choose the subband HL2. calculate the correlation between the midband coefficients and the two generated pseudorandom sequences (PN_0 and PN_1). Performance Evaluation Metrics: Watermarking algorithms are usually evaluated with respect to two metrics: imperceptibility and robustness [21]. apply DWT to the HH1 subband to get four smaller subbands. . Step 6: For each block in the subband HL2 (or HH2). 2007 Fig. Computer Sci.J. and a 256×256 greyscale image of the expression 'copyright' as the watermark image. Step 3: Divide the subban HL2 (or HH2) into 4×4 blocks. 2 b. as shown in Fig.. Step 5: Regenerate the two pseudorandom sequences (PN_0 and PN_1) using the same seed used in the watermark embedding procedure. PERFORMANCE EVALUATION We evaluated the performance of the combined DWTDCT image watermarking algorithms using a 512×512 'Lena' as the original cover host image. 3: Combined DWTDCT watermark extraction procedure watermarking algorithm. Step 1: Apply DWT to decompose the watermarked image into four nonoverlapping multiresolution subbands: LL1. Step 2: Apply DWT to HL1 to get four smaller subbands. as shown in Fig. 3 (9): 740746. otherwise the extracted watermark is considered 1. then the extracted watermark bit is considered 0. The two images are shown in Fig. and HH1. respectively. Step 4: Apply DCT to each block in the chosen subband (HL2 or HH2). and compute the similarity between the original and extracted watermarks. and extract the midband coefficients of each DCT transformed block. HL1. If the correlation with the PN_0 was higher than the correlation with PN_1. and choose the HH2 subband. 4 and 5. 2 a. The two metrics are described below.
190 PSNR = 97. LH1. by different types of digital signal processing attacks[22]. The DCT transform was then applied to each block in the chosen subband. That is the Gaussian noise is a watermark degrading attack. Computer Sci. internationally or unintentionally. N ρ ( w. w and wˆ are the original and extracted watermarks respectively. The correlation factor ρ may take values . In general. Imperceptibility: We evaluated imperceptibility of the combined DWTDCT algorithm by measuring PSNR for the HL2 and HH2 subbands. 3 (9): 740746.072 while HH2 gave 97. however. 2007 Table 1: The 'Lina' host image watermarked at different 2DWT subbands Watermarked Image DWTOnly DWTDCT DWTOnly DWTDCT Original Image (HL2) (HL2) (HH2) (HH2) PSNR = 80. we decided to embed the watermark in HL2 (or HH2). JPEG compression is a watermark removal attack and cropping is a watermark dispositioning geometrical attack. 5. LL1. the difference between the PSNR values of the DWTOnly and the combined DWTDCT algorithms is relatively large. The difference is small between the two PSNR values indicating comparable performance of the two subbands. they are good representatives of the more general attacks. HL2 gave a PSNR value of 97. 6. HL1 and HH1. a correlation coefficient of about 0. image compression and image cropping.083. The correlation values given in Table 2 show clearly that the combined DWTDCT watermarking algorithm outperforms the conventional DWTOnly approach with respect to robustness against the Gaussian noise and cropping attacks. independently. In contrast. This indicates that improvement in imperceptibility can be achieved by applying DCT on a DWT transformed HL2 (or HH2) subband. Robustness: Table 2 shows the correlation values between the original watermark and the watermarks extracted from subbands HL2 and HH2 after being subjected to different attacks. PSNRdB = 10 ⋅ log10 = 20 ⋅ log10 MAX I MSE MAX 2 I MSE between 0 (random relationship) to 1 (perfect linear relationship). w) = ˆ i =1 N i =1 ˆ wi wi 2 N i =1 (6) wi ˆ wi 2 where N is the number of pixels in watermark.. The three attacks are a few. In this chapter. The selected 128×128 subband was divided into 4×4 blocks giving a total of 1024 blocks. As a measure of the quality of a watermarked image.J. Since embedding the watermark beyond the first DWT level is more effective. the peak signal to noise ratio (PSNR) is typically used. HH2 gave better 744 (5) Robustness: Robustness is a measure of the immunity of the watermark against attempts to remove or degrade it. The results are better regardless of whether the watermark was embedded in HL2 or HH2.083 Imperceptibility: Imperceptibility means that the perceived quality of the host image should not be distorted by the presence of the watermark.072 PSNR = 77. 3 and 4.098 PSNR = 97. after which the watermark was embedded according to Eq. and as shown in Table 1. Performance results: We carried out the watermark embedding algorithm described in the previous section. PSNR in decibels (dB) is given below in Eq. We measured the similarity between the original watermark and the watermark extracted from the attacked image using the correlation factor ρ given below in Eq. we will report on robustness results which we obtained for three major digital signal processing operations (attacks): Gaussian noise.75 or above is considered acceptable. Applying the 1level DWT of the 'Lina' image produced four 256256 subbands. however.
the robustness of the DWTOnly approach against the JPEG compression attack is better than that of the combined DWTDCT algorithm.9738 0. 6: Extracted watermarks from the cropped HH2watermarked image using DWT only Fig. was not acceptable. 745 RESULTS AND DISCUSSION We described the performance of the combined DWTDCT watermarking algorithm.989 0.5029 0.958 0.550 DWTDCT (HL2) 0. The results we obtained for the DWTOnly approach indicated a better imperceptibility performance was obtained when the watermark was embedded in the HL2 or HH2 subbands.660 0.884 0.868 0. 9: Extracted watermarks from the compressed HL2watermarked image using DWT and DCT robustness against cropping compared with HL2.8 0 40 80 8 152 200 DWTOnly (HL2) 0. Computer Sci.478 0.623 0. the better combined DWTDCT results are shown in Figure 7 which shows the extracted watermarks from the HH2 subband.023 0. and therefore applying DCT at a DCTtransformed image degraded the quality of the embedded watermark greatly.065 0.741 0.696 Fig.J. 3 (9): 740746. Figure 6 shows the extracted watermarks from HH2 when DWTOnly was used. however.989 0. The robustness performance. As given in Table 2 and illustrated in Fig. we also evaluated the watermarking performance when DWTOnly was used.447 0.549 DWTOnly (HH2) 0. we combined DWT with the another equally powerful transform.746 0. 2007 Table 2: Correlation values due to three different attacks Correlation Factor Fausain (Noise) Compression (Quality) Cropping (Block Size) 0 0.652 0.643 0. 8: Extracted watermarks from the compressed HL2watermarked image using DWT only Fig. The combined DWTDCT watermarking algorithm's imperceptibility performance was better . To improve performance.604 0.445 0. the DCT.778 0. For the sake of comparison.4 0.8359 0.467 0. 7: Extracted watermarks from the cropped HH2wateramrked image using DWT and DCT Fig. In contrast.723 0.546 0. 8 and 9. This is due to the fact that the DCT is the core component of the JPEG compression.733 0.968 0..726 0.726 0.012 0.574 DWTDCT (HH2) 0.314 0.736 0.698 0.
2003. "Watermarking Digital Image and Video Data: A StateofArt Overview. Artech House. C. pp: 154164. 2004. 2004. Consumer Electronics. 5(1): 3438. 2003. of the IEEE. in Proc. Chu. of the IEEE 2003 Int. "Multiresolution Watermarking for Digital Images. In conclusion.." in Proc. Arnold. China. M. Tay. Wavelets and Subband Coding. S. Voloshynovskiy. 12. C. USA. Academic Press. "A Robust DCTbased Watermarking for Copyright Protection. Podilchuk and E. in DWTbased digital watermarking applications. A." IEEE Trans. A Survey of Digital Image Watermarking Techniques. pp: 15251528.. 46. . and A. Hsu. 4. "Digital watermarking using zerotree of DCT. EstimationBased Attacks. and Benchmarks. 1995. D. M. Multimedia. and H. 3 (9): 740746. Image Processing . Bloom. "Perceptual Watermarks for Digital Images and Video. 2003. followed by the application of DCT on the selected DWT subbands. the improvement in robustness brought by the combined DWTDCT algorithm was considerably high." in Proc." Pattern Recognition Letters. 2002. Discrete Cosine Transform: algorithms. Huang. S. Tseng. 21." IEEE Trans. 1998. pp: 11081126. 20." IEEE Trans. 2000. Setyawan and R. Lin.J. on Industrial Electronics. 6. and J. Havlicek. Chang. S. Hsieh. Potdar. V. on Circuits and Systems. Nikolaidis. Chan. "Wavelet Tree Quantization for Copyright Protection Watermarking. M. Compression. pp: 8794. and Y. 1999. Langelaar.. A.. C. 15. Lin. vol. and W. "DCT and DWTbased Image Watermarking Using Subsampling. on Image Processing. Schumucker and S." IEEE Trans. Wang. Pitas. M. 2001. 2. R. 2: 546549. pp: 709716. 39(8): 118126." Proc. 2002. "A New Wavelet Based Logowatermarking Scheme. and P. Wang. vol. Digital Watermarking. 17.. Hung. Pattern Recognition. "Attacks on Digital Watermarks: Classification. Myazaki. pp: 53085313. K. USA.. and J. "A novel technique for robust image watermarking in the DCT domain. Academic Press. I. 37(3):469474. 9. vol. 2000. 1990. "Image Watermarking Using Wavelets. Cox. vol. and Encryption. 2005. no. we described a combined DWTDCT digital image watermarking algorithm." On the evaluation of performance of digital watermarking in the frequency domain. Techniques and Applications of Digital Watermarking and Content Protection. 18. Image Processing. pp: 258261. and B. and J. USA. 2001. Ejima. Chatterji. 17(5):2046. 87. S. P. advantages. 19. Chin. Consumer Electronics . Perth. Lin. 2003.. Miller and J. Magazine. 45(8): 10971101. 2005. and L. of the 2005 IEEE Fourth Int. Reddy. 2. REFERENCES 1. Hiding Data in Images by Simple LSB Substitution. CONCLUSIONS The discrete wavelet transform (DWT) and the discrete cosine transform (DCT) have been applied successfully in many in digital image watermarking. and I. 5. no. of the IEEE International Conference on Industrial Informatics. Rao. Pun.II. USA. Moreover. combining appropriate transforms with the DWT may have a positive impact on performance of the watermarking system. 2007 than the performance of the DWTOnly approach. of the IEEE Midwest Symposium on Circuits and Systems. 48(5): 875882. USA. S. 10. Wang. 2001. 22. Wu. Vetterli. "Asymptotically optimal detection for additive watermarking in the DCT and DWT domains. 46(3): 415421. M. Wu." IEEE Trans. Delp." IEEE Trans. "Hiding Digital Watermarks Using Multiresolution Wavelet Transform. Prentice Hall. Yip. R. 26(7): 10191027. 11. M. M. of the SPIE Conference On Mathematics of Data/Image Coding. Conf. 3." IEEE Trans. Watermarking was done by embedding the watermark in the first and second level DWT subbands of the host image. and Y. Australia." in Proc. USA. 7. on Machine Learning and Cybernetics. and B. F. 7. Wolthusen. " Copyright protection of digital images by means of frequency domain watermarking. G. In this paper. I." in Proc. W. Pereira and T. Kova evi . on Neural Networks and Signal Processing. 2. 2(10): 563571. Han and E. Wolfgang. Tsai. C. 2005. 13. C. Computer Sci. Lagendijk. The combination of the two transforms improved the watermarking performance considerably when compared to the DWTOnly watermarking approach. applications." Comm. Hsieh." IEEE Signal Processing Magazine. Oklahoma. Cheng. 8. 746 16. no." Proc. and C. Conf... Deng. "DCTBased Image Watermarking Using Subsampling. Conf. of the IEEE Int. 2000. 14. 1. 13. Lin and J.