BLIND WATERMARKING SCHEME BASED ON RDWT-DCT FOR COLOR IMAGES

`International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume 2, Issue 4 (July-Aug 2014), PP. 117-121
117 | P a g e
BLIND WATERMARKING SCHEME BASED ON
RDWT-DCT FOR COLOR IMAGES
Ms. Shailaja Deshmukh, Prof. S.K. Bhatia, Dr. Rajneesh Talwar
Department of Electronics and Telecommunication Engineering, PG Student, JSPM’s Imperial College of Engg. &
Research, Wagholi, Pune, India
Department of Electronics and Telecommunication Engineering, Assistant Professor, JSPM’s Imperial College of
Engg. & Research, Wagholi, Pune, India
Principal, CGC group of Colleges, Landran, Punjab, India
Abstract— In Digital era sharing of images have become very
common and raises the risk of using it for unethical and
fraudulent purposes with the help of manipulation tools. Digital
image watermarking is one way to protect the digital information
(text, images, audio, and video) from fraudulent manipulations.
Digital Image Watermarking is a process of implanting data in
the original image for authentication. In this paper we are
providing one such watermarking scheme for color images. The
proposed method is designed to be robust for common attacks
with the aid of redundant discrete wavelet transform (RDWT)
and discrete cosine transform (DCT) properties. After applying
two levels RDWT decomposition to the blue channel of cover
image, we apply DCT to HH_LL subband i.e. 2nd
level
decomposed coefficient of HH band and to the watermark.
Divided the HH_LL sub band into 4x4 subblocks and DCT
coefficients of the last subblock of the cover image are replaced
with the DCT coefficients of watermark. Inverse DCT and
inverse RDWT is performed to get watermarked image. The
performance of the proposed technique is measured using the
parameters PSNR and NCC.
Keywords— RDWT, DCT, PSNR, NCC
I. INTRODUCTION
Copyright protection of digital media is the very first
application that comes to mind for digital watermarking. In
the past, duplicating artwork was quite complicated and
requires great efforts to create the work looks just like the
original. However, nowadays in digital world, it is very simple
for anyone to duplicate or manipulate digital data. The digital
image watermarking allows the watermark to be embedded
visibly or invisibly in the original image for identification of
the owner. This concept can also be used for other media,
such as digital video and audio. Visible watermarking inserts a
discernible watermark for the immediate identification of the
images. The insertion of a visible watermark should meet two
contradictory conditions, the watermark should be strong
enough to be visible, and also it should be light so that it is
unnoticeable and not disturbing the appearance of the original
image. On the other hand invisible watermarking inserts the
watermark such that they are indiscernible with the original
image.
Watermarking techniques are characterized with the
following features.
1) Robustness: It is the tenacity of the watermark to
detect accurately against various attacks on the
watermarked image.
2) Imperceptibility: The essence of the original image
should not be disturbed with the addition of
watermark to it.
3) Capacity: Amount of data that can be embedded in
the original image for authentication.
By embedding the watermark in the low frequencies
increase the robustness of the watermarked image but it
degrades or tampers the original image which in turn affects
the imperceptibility. Thus robustness and imperceptibility are
contradictory to each other. In this paper we proposed a
method where in we are using the HH_LL band for
embedding the watermark in the original image to balance
between the robustness and imperceptibility. In addition to
that DCT is used to separate out the low and high frequency
components. We are embedding the watermark in the high
frequency components that increases the imperceptibility.
The paper is organized as follows. Section II briefly
overviews the used concepts. Section III outlines the proposed
scheme for watermark embedding and extraction algorithm.
Section IV contains experimental results followed by
conclusion in section V.
II. BRIEF OVERVIEW OF CONCEPTS
A. Redundant Discrete Wavelet Transform
DWT family has remarkably become researchers first
choice for Image watermarking algorithms. RDWT is one
among that family which has same characteristics as DWT
with an added advantage of Shift invariance. DWT
decomposes an image using analysis filter and down samples
the frequencies. Due to the presence of down sampler it

118 | P a g e
removes
the redundant
data causing
shift variance
problems.
RDWT
decomposes an image using only analysis filter resulting in
redundant data. With the aid of this redundant data even if
there is any minor shift in the image, the coefficients will not
vary causing shift invariant. The following figure depicts the
RDWT decomposition into sub bands.
Fig. 1 RDWT analysis filter bank
B. Discrete Cosine Transform
The discrete cosine transform describes an image as a
collection of sinusoids. In comparison with DFT, DCT is
simple because it represents the image in cosine functions and
is also faster. DCT transforms the original image into low
frequency, midband frequency and high frequency
coefficients. Most of the energy of the original image is
concentrated in the lower frequency coefficients and it keeps
decreasing as the frequency increases. Thus high frequency
coefficients contains the least energy and can be neglected
resulting in less number of coefficients for image. This
property of DCT is known as energy compaction. The
following figure shows how a DCT segregates the original
image information into different bands.
Fig. 2 DCT transformation of an image into DC and AC coefficients
The white blocks depicts the DC coefficients i.e.., they are
the low frequency components containing high energy, the
light gray are the midband frequency components while the
dark blocks are the ac coefficients representing high
frequencies and containing low energy.
Human eyes are insensitive to higher frequencies and any
changes in this region will not be perceptual. Thus we are
considering this advantage for our blind watermarking scheme.
III. PROPOSED METHOD
The proposed method consists of two steps. Watermark
Embedding step and Watermark Extraction step.
A. Watermark Embedding Process
The process of adding owners identifier or logo in an
invisible manner to the image is known as Watermark
embedding. It is done for the purpose of authentication or for
proving the ownership of the creator.
Following is the proposed procedure to embed a
watermark in an image.
Watermarked
image
Color
image
B panel RDWT DCT
EmbeddingDCT
Inverse
DCT
Logo
image
Inverse
RDWT
Input
Image
LPF
HPF
LPF
HPF
LPF
HPF
LL
LH
HH
HL
DC

119 | P a g e
Fig. 3 Watermark Embedding block diagram
1) Select input color image of size NxN and is parted into
red, green and blue images.
2) Using RDWT, the blue panel of the cover image
disintegrated into four sub bands LL, LH, HL, HH each of
same size.
3) Select the each band (lets say HH) and apply RDWT to
decompose it into sub bands HH_LL, HH_LH, HH_HL,
HH_HH of same size.
4) DCT is applied to the subband HH_HH and are then
partitioned into 8x8 subblocks of size N/8x N/8 i.e., 8 rows
and 8 columns.
5) The DCT coefficient matrix D is formed from the sub
block of HH_HH sub band.
6) The logo of size N/8xN/8 is used for embedding and
DCT is applied to form the coefficient matrix DW.
7) The two dimensional matrices D and DW each of size
N/8xN/8 are converted to one dimensional matrices for
embedding process.
8) The one dimensional matrix D is substituted with DW.
9) Remodel the one dimensional coefficient matrix D back
to two dimensional and combine it with the other sub blocks
to form the subband HH_ HH*.
10) Inverse DCT and Inverse RDWT are applied to the
transformed subband HH_ HH* and other subbands LH, HL,
HH to form the HH* subband.
11) Second level Inverse RDWT is applied to the LL, LH,
HL, HH* subbands to generate the blue panel of the
watermarked image.
12) The red panel, green panel and transformed blue panels
are integrated to form the Watermarked Color Image.
B. Watermark Extraction Process
The process of extricating the owners identifier or logo
from the watermarked image is known as Watermark
extraction. The extracted logo should resemble with the one
which was added at the embedding process.
Following is the proposed procedure to extract a
watermark from a watermarked image.
Fig. 4 Watermark Extraction block diagram
1) Select the blue panel from the watermarked color image
for extraction process.
2) The blue panel watermarked image is disintegrated using
RDWT, into 4 subbands: LL, LH, HL and HH.
3) Select the HH band to decompose it into HH_LL,
HH_LH, HH_HL, HH_HH sub bands using RDWT.
4) Repeat the same process as done in embedding of
applying DCT to the HH_HH subband and partition the
subband into 8x8 subblocks.
5) Select the extreme bottom right sub block and apply
Inverse DCT to extract the logo.
6) Thus the watermark is extracted from the input color
image.
IV.EXPERIMENTAL RESULTS
The proposed algorithm is tested with Lena image of size
512x 512 as original image and a watermark image with size
64x64. Following figure 6 and 7 shows the input color image
and watermark image respectively. Figure 8 and figure 9 are
the watermarked image and extracted watermark image
respectively.
Fig. 5 Original Color image
Watermarked
image
B
panel
Extraction
RDWT DCT
Inverse
DCTWatermark

120 | P a g e
Fig. 6 Watermark image
Fig. 7 Watermarked image
Fig. 8 Extracted Watermark image
The watermarked image was tested for Gaussian noise, salt
& pepper noise, speckle noise, Poisson noise, blurring, and
sharpening and rotation attacks. The PSNR and NCC values
for each subband are tabulated below.
S.N
o
Attacks LL_HH Watermarked image
1 Without Attack 0.9033 51.2016
2 salt and pepper 0.0824 40.7829
3 Gaussian 0.0375 38.3065
4 Speckle noise 0.2174 41.1621
5 Poisson noise 0.1418 41.9451
6 Blurring attack -0.505 47.9769
7 Sharpening 0.9007 41.67
8 Rotation by 90 -0.0657 32.0772
TABLE. 1 NCC AND PSNR VALUES FOR LL SUBBAND
S.No Attacks LH_HH Watermarked image
3 Gaussian 0.1406 38.3186
7 Sharpening 0.9665 41.6708
8 Rotation by 90 0.0813 32.0775
TABLE. 2 NCC AND PSNR VALUES FOR LH SUBBAND
S.No Attacks HL_HH Watermarked image
3 Gaussian 0.2032 38.3185
7 Sharpening 0.9583 41.6714
8 Rotation by 90 -0.0796 32.0774
TABLE. 3 NCC AND PSNR VALUES FOR HL SUBBAND
S.No Attacks HH_HH Watermarked image

121 | P a g e
3 Gaussian 0.552 38.31
7 Sharpening 0.9951 41.0402
8 Rotation by 90 0.0695 32.0772
TABLE. 4 NCC AND PSNR VALUES FOR HH SUBBAND
The PSNR obtained with the proposed algorithm is 45.53 to
ensure that the watermarked image closely resembles the input
color image. The acceptable PSNR values are above 40.
NCC is the parameter to assess the correlation between the
extracted watermark and the embedded watermark. We have
achieved a NCC value of 0.999 which is approximated to 1
implies that extracted watermark and the input watermark are
same.
V. CONCLUSIONS
Thus the blind color image watermarking based on RDWT-
DCT scheme is proposed. From the results it is observed that
by increasing the level of decomposition and using RDWT
with DCT the robustness and imperceptibility is improved as
compared to the existing techniques. It is observed that by
implanting watermark in HH_LL subband the proposed
algorithm became more robust to common attacks. It is
affirmed from the results that watermark extracted from
HH_LL subband is robust to salt and pepper noise, Gaussian
noise, speckle noise, Poisson noise and sharpening attack. In
the future we will extend our work by embedding the copy of
watermark in all the four sub bands to increase the robustness
to other attacks.
REFERENCES
[1] Samira Lagzian, Mohsen Soryani, Mahmood Fathy “Robust
watermarking scheme based on RDWT-SVD: Embedding data in all
subbands” AISP International 2011.
[2] V.Santhi Member, IACSIT, Prof. Arunkumar Thangavelu “DC
Coefficients Based Watermarking Technique for color Images Using
Singular Value Decomposition” IACSIT Int. J. of Computer and
Electrical Engineering, Vol.3, No.1, February, 2011
[3] Nidhi Divecha Dr. N. N. Jani “Implementation and performance
analysis of DCT-DWT-SVD based watermarking algorithms for color
images” International Conference on ISSP 2013.
[4] Baisa L. Gunjal and Suresh N.Mali, “SECURED COLOR IMAGE
WATERMARKING TECHNIQUE IN DWT-DCT DOMAIN,”
International Journal of Computer Science, Engineering and
Information Technology (IJCSEIT), Vol.1, No.3, August 2011.
[5] D. Kundur, D. Hatzinakos, Digital watermarking using multiresolution
wavelet decomposition, Proceedings of IEEE,
[6] Baisa L Gunjal and Dr. Suresh N Mali, “Handling Various Attacks in
Image Watermarking” CSI communications 2013.
[7] The MathWorks, Inc., Getting Started with MATLAB. http://
www.mathworks.com/access/helpdesk/help/techdoc/learn_matlab/learn
_matlab.shtml
[8] Y. Shantikumar Singh, B. Pushpa Devi, and Kh. Manglem Singh, “A
Review of Different Techniques on Digital Image Watermarking
Scheme”, IJER Volume No.2, Issue No.3, pp : 193-199.
[9] Priyanka Singh, Suneeta Agarwal, Akanksha Pandey, “A Hybrid
DWT-SVD Based Robust Watermarking Scheme for Color Images and
its Comparative Performance in YIQ and YUV Color Spaces”, IEEE
IACC 2013.
[10] R. A. Ghazy , N. A. El-Fishawy,M. M. Hadhoud, M. I. Dessouky and
F. E. Abd El-Samie, “ An Efficient Block-By-Blocksvd-Based Image
Watermarkingscheme” UBIC.

BLIND WATERMARKING SCHEME BASED ON RDWT-DCT FOR COLOR IMAGES

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to BLIND WATERMARKING SCHEME BASED ON RDWT-DCT FOR COLOR IMAGES (20)

More from International Journal of Technical Research & Application (20)

Recently uploaded (20)

BLIND WATERMARKING SCHEME BASED ON RDWT-DCT FOR COLOR IMAGES