Gh2411361141

Mandeep Singh Saini, Venkata Kranthi B, Gursharanjeet Singh Kalra/ International Journal of
Engineering Research and Applications (IJERA)ISSN: 2248-9622
www.ijera.com Vol. 2, Issue 4, May-Jun 2012, pp.1136-1141
Comparative Analysis of Digital Image Watermarking
Techniques in Frequency Domain using MATLAB SIMULINK
Mandeep Singh Saini1, Venkata Kranthi B2, Gursharanjeet Singh Kalra3
Lovely Professional University
Phagwara, Punjab (INDIA)-144411

ABSTRACT
Digital multimedia data is easily created, involves an algorithm that directly operates on the
copied, processed, stored and distributed among pixel values of the host image. In transform domain
unauthorized users using freely available software. method, the pixel values are transformed into another
So, digital watermarking technique provides domain by applying appropriate transform technique
solution to the problem. Watermarking is the like discrete cosine transform (DCT) and discrete
process in which an informal data is incorporated wavelet transform (DWT). Generally, frequency
in original data to protect the owner’s copyright domain watermarking is more robust than the spatial
over that content. In this paper, comparative domain. In this paper, we will implement different
analysis of digital image watermarking techniques digital image watermarking techniques in MATLAB
in frequency domain is presented. Behavioral SIMULINK for comparative analysis.
MODEL of different image watermarking
techniques is introduced in MATLAB SIMULINK 1.1 Matlab Simulink
using Simulink library blocks. In this paper, Simulink is a graphical simulation tool for dynamic
comparative analysis is performed based on system modeling, simulation and comprehensive
robustness and computational complexity of analysis. It can call MATLAB powerful function
different watermarking algorithms after attacks. library, save a lot of duplicate codes writing work,
and the user can immediately see the results of
Keywords- Watermarking; discrete cosine transform simulation results [3]. Initially for behavioral model,
(DCT); Matlab Simulink; Robust; Image. several in-built function are used for watermarking
process and several blocks are created using Matlab
1. INTRODUCTION function block.
In the recent times, with the development of
network technology, protection of multimedia data
becomes increasingly prominent. Because of their
digital nature, multimedia data can be duplicated,
modified, transformed, and diffused very easily [1].
The growth of internet is so vast that any data can be
easily transferred to the other person in just a single
click. Due to its portability, the trend of piracy and
duplicity rapidly approaches to Everest now days.
The original producer of the file even doesn’t know
that the file created by him/her is available for free
through internet and even if knows, nothing can be
done[2]. It has been recognized for quite some time
that current copyright laws are inadequate for dealing
with digital data. Figure 1.1: Block creating using Matlab
So, recent development of digital watermarking function
technology can solve this problem. Watermarking is
the process to hide some data which is called In this block, we can write a function
watermark or label into the original data(image, according to our desired dimensions for cropping the
audio or video) such that watermark can be extracted input image. This solves the problem of non-
or detected later to make an assertion about the availability of in-built blocks. In the above Matlab
object. Watermarks of varying degree of visibility are function, image crop block is not available, so
added to presentation media as a guarantee of function for cropping image to desired size is written.
authenticity, quality, ownership and source.[1] Here u defines the input and y defines output of the
For insertion and extraction of watermark, there block.
are two primary methods that are spatial domain and
transform domain. The spatial domain method

1136 | P a g e


2. RELATED WORKS most of the visually significant information about the
There are many papers in which authors use image is concentrated in just a few coefficients of the
DCT and DWT transforms in their algorithm for DCT. For this reason, the DCT is often used in image
embedding watermark. Authors of [1] used DWT compression applications.
using fuzzy logic on FPGA for hardware for an In this algorithm, the original image is divided into
adaptive encrypted watermarking method aims to several 8x8 blocks. Then those several blocks are
achieve low power usage, real-time performance, discrete cosine transformed (DCT-2). The watermark
reliability. They conclude the watermarking scheme is embedded by 8*8 piecemeal and then again passed
is imperceptible and robust against geometric attacks. through Inverse DCT-2.
Importance of fuzzy logic to calculate the gain factor The definition of the two-dimensional DCT for an
with respect to texture sensitivity of the image was input image A and output image B is
𝑁−1 𝑁−1 𝜋 2𝑚 +1 𝑝 𝜋 2𝑚 +1 𝑞
also proved. Authors of [2] implement their algorithm Bpq = αp αq 𝑚 =0 𝑛 =0 𝐴 mncos cos
2𝑀 2𝑁
on DWT transform. In this, first they apply Arnold (1)
transform to the watermark after which, Chaos
transform is applied to the to the output of Arnold 0 ≤ p ≤ N-1
transform and then embedding is done to get the
watermarked image and conclude that the watermark 0 ≤ q ≤ N-1
becomes more secure with Arnold and Chaos pseudo 1
random sequence. Another encryption algorithm in , 𝑝=0
𝑁
𝛼 𝑝 = (2)
[3] based on color image watermarking is proposed in 2/𝑁, 1 ≤ 𝑝 ≤ 𝑁 − 1
SIMULINK. In this, they first separate color 1
components into RGB and then embedded watermark , 𝑞=0
𝑁
𝛼 𝑞 = (3)
on G components after DCT. Also a key is set for 2/𝑁, 1 ≤ 𝑞 ≤ 𝑁 − 1
encryption of watermark. They conclude that their p and q varies from 0 to N-1, and M-1.
embedding and detection algorithm is fast and
conveniently verify the correctness of the algorithm. Two dimensional inverse DCT can be calculated as
In [4] the authors used DCT and DWT both for below
watermarking scheme. In their algorithm, the 𝑁−1 𝑁−1 𝜋 2𝑚 +1 𝑝 𝜋 2𝑚 +1 𝑞
information of digital watermarking which has been Amn = αp αq 𝑝=0 𝑞 =0 𝐵 mncos cos
2𝑀 2𝑁
Discrete Cosine Transformed, is put into the high (4)
frequency band of image which has been wavelet
transformed. According to them, their simulation 0 ≤ m≤ N-1
results show that their algorithm is invisible and has
good robustness for some common image processing 0 ≤ n ≤ N-1
1
operations. Authors of [5] proposed hardware , 𝑝=0
𝑁
implementation using DCT/IDCT in Xilinx 𝛼 𝑝 = (5)
XC3S4000 FPGA which increases the speed close to 2/𝑁, 1 ≤ 𝑝 ≤ 𝑁 − 1
1
500% over the conventional methods. In [6], a dual , 𝑞=0
𝑁
image watermarking technique based on DWT is 𝛼 𝑞 = (6)
used. In this, watermarking technique incorporates 2/𝑁, 1 ≤ 𝑞 ≤ 𝑁 − 1
two watermarks in a host image, a watermark in from
of PN sequence is embedded in the wavelet domain Where M and N are the row and column size of
of a primary watermark before being embedded in the image A (Input data).
host image and concludes that algorithm proves its
robustness against several kinds of attacks. 3.1 Watermark Embedding
The watermarking scheme consist of DCT transform
of original image which will concentrate the main
3. WATERMARKING BASED ON DCT IN
image information into the least low frequency
MATLAB SIMULINK coefficients, and cause smallest image blocking
The DCT transform a signal from a spatial
effect, thus achieves good compromise between
representation into a frequency representation. Lower
centralized information and computational
frequencies are more obvious in an image than higher
complexity. Moreover, 2D-DCT is real
frequencies so if we transform an image into its
transformation. It has good energy compression
frequency components and throw away a lot of higher
ability, goes to the related ability envelope
frequency coefficients, we can reduce the amount of
compatibility with the international compression
data needed to describe the image without sacrificing
standard. It is a good choice about using the 2D-DCT
too much image quality. The discrete cosine
transform to implement the watermarking embedding
transform (DCT) represent an image as a sum of
and extraction. Figure 3.1 shows the embedding
sinusoids of varying magnitudes and frequencies. [4]
procedure of watermark.
The DCT has the property that, for a typical image,

1137 | P a g e

3.2 Watermark Extraction
Similarly the watermark can be extracted using the
same transform but here the difference is, that only
comparison is done according to the embedding
scheme.

Figure 3.1 Watermark embedding
algorithm

The watermark Embedding processes consists of Figure 3.3 Watermark extraction algorithm
following steps:
1. The original image is first resized into The watermark Extraction process consists of
256x256 pixels. following steps:
2. The watermark is resized into 32x32 pixels. 1. Firstly the watermarked image is resized into
3. Then original resized image is block- 256x256 pixels if needed.
processed into 8x8 blocks so as to 2. Then watermarked image is again block
completely embed the 32x32 watermark on processed into 8x8 pixels.
256x256 image. 3. Then Forward 2D-DCT is applied on the
4. After this, 2D- Discrete Cosine Transform is block processed image to have frequency
applied on these 8x8 sub-blocks individually separation.
to convert it from time domain to frequency 4. After this, Extraction process is done in the
domain. same way as embedding was done in
5. The watermark image is embedded on the embedding process.
8x8 blocks in the middle frequency band. 5. Depending upon the comparison done with
6. For embedding watermark, the value of particular pixels on which watermark data
frequency coefficient is changed in- was embedded, decision is make which is
accordance to watermark data. stored in again 32x32 empty array.
7. Reversed 2D-DCT is performed afterward 6. Then this matrix is the extracted image
on the watermark embedded image to which was original watermark image.
convert from frequency domain to time
domain.
8. The final image is the watermarked image.

Figure 3.4 Watermark extraction model in
Simulink
Figure 3.2 Watermark embedding model in
Simulink

1138 | P a g e


The Watermark embedding process consists
4. WATERMARKING BASED ON DWT of the following steps:
In this algorithm, the original image is first 1. The original image is first resized into
resized to 512x512. Then this is decomposed through 512x512 pixels because after DWT
1- level Discrete Wavelet transform. decomposition, we will get only with
Discrete wavelets transform is frequency domain 256x256 pixel size image.
analysis method which can localize frequency domain 2. As, the size of our watermark is 32x32,
[4]. The basic idea of discrete wavelet transform 256x256 is the exact dimension to embedded
(DWT) in image processing is to multi-differentiated 32x32 pixels on 8x8 sub-blocks in complete
decompose the image into 4 frequency districts which image.
is one low frequency district(LL) and three high 3. Then this is decomposed through 1-level
frequency districts (LH,HL,HH). L and H represent Discrete Wavelet Transform which leads the
Low-pass and High-pass filter. . A two level DWT image into four parts [1] as explained above.
decomposed image is shown in fig. 4.1. 4. After this, watermark is embedded into the
Low-high frequency components.
LL2 HL2 5. Then again, 1-level IDWT is applied to this
image to synthesize it to complete image.
LH2 HH2 6. The final image is watermarked image.
HL1

1.2 Watermark Extraction Model in Matlab
LH1 HH1 Simulink
The watermarking process is same but in the reverse
Figure 4.1: 2-level DWT decomposition order.
The information of low frequency district
image is very close to the original image. The
frequency district of LH, HL, and HH respectively
represents the level detail, the upright detail and the
diagonal detail. The human eye is sensitive to the
change of smooth district of image, but not at the
edges and peaks. So, it is not safe to putting
watermark in coefficients of high frequency band of
DWT image because it can be easily noticed. The
brightness masking on human visual model shows
that the larger the background brightness, the more
the just noticeable difference of embeddable signal, Figure 4.2: watermark extraction model
which means low frequency approximate image can
be embedded by more watermarking capacity is Watermark extraction process consists of following
lower than JND, as human eyes cannot suspect the steps:
existence of signal. Some common attacking to low 1. The watermarked image is resized into
frequency coefficients, the host image is also 512x512 pixels if needed.
destroyed. So it is good to embed watermark in 2. The watermarked image is decomposed into
medium and low frequency. 1-level DWT transform.
3. Then according to the embedding process,
1.1 Watermark Embedding Model in Matlab the comparison is done for extraction.
Simulink 4. The generated image by comparison is the
extracted watermark image.

5. WATERMARKING BASED ON DCT
&DWT IN SIMULINK
In this algorithm, we use DCT and DWT
both the transforms for embedding and extraction of
watermark.
1.3 Watermark Embedding Model in Matlab
Simulink

Figure 4.1: watermark embedding model in
Simulink

1139 | P a g e

2. Then the low-high frequency sector is block-
processed into 8x8 blocks.
3. These 8x8 sub-blocks are passed through
Discrete Cosine Transform (DCT-2).
4. Then, according to the embedding scheme
on individual pixel, comparison decision as
done whose result is stored in an empty
array of 32x32 same as above.
5. Finally this empty array will contain the
complete decision information and this array
will be the extracted watermark image.

6. COMPARATIVE ANALYSIS OF
Figure5.1: watermark embedding model DIFFERENT ALGORITHM
The 32x32 is taken as watermark image
Embedding process consists of following steps: information in fig. 6.1 and 256x256 is taken as
1. The original image is first resized into the source image shown in fig. .6.2
512x512 pixels.
2. Then it is decomposed through 1-level
Discrete Wavelet transform (DWT) which
results the image into four parts.
3. Then the high-low frequency sector is block
processed into 8*8 blocks to let 32x32
watermark image embed completely into
host image correspond to a single pixel in
every 8x8 block.
4. After this, forward DCT-2 is applied on
each 8x8 sub-blocks for better frequency
band separation.
5. Then watermark image is embedded into the Figure 6.1 Figure 6.2
processed image in the middle frequency TABLE I. RESULTS AFTER ATTACKS (PSNR
band of sub-blocks. in dB)
6. Again inverse IDCT-2 is applied after
embedding the watermark. CATAGORY DCT DWT DCT&
7. After this, to get the original image, IDWT DWT
is applied to synthesize the complete image. WATERMARKED 34.45 36.38 41.40
8. The final image is the watermarked image. IMAGE
1.4 Watermark Extraction Model in Matlab SALT & 20.02 21.33 22.05
Simulink PEPPER(0.02)

ROTATION (2 14.91 15.57 17.61
degree)

ENHANCEMENT 30.41 32.65 37.41
(contrast limit-0.6-
0.9)
GAUSSIAN 16.23 17.11 21.34
NOISE(mean-
0.01,var-0.01)
SPECKLE (0.06) 15.85 16.21 18.83

Figure5.2: watermark extraction model
TABLE II. RESULTS AFTER ATTACKS (NC
The extraction process consists of the following in dB)
steps: CATAGORY DCT DWT DCT&
1. The watermark image is decomposed DWT
through 1-level DWT transform to again WATERMARKED 1 1 1
have four different frequency sectors.[1] IMAGE

1140 | P a g e

SALT & 0.991 0.996 0.998 [5] Rajesh Kannan Megalingam, Venkat
PEPPER(0.02) Krishnan B., Vineeth Sarma V., Mithun M.,
ROTATION (2 0.952 0.963 0.982 Rahul Srikumar, "Hardware Implementation
degree) of Low Power, High Speed DCT/IDCT
Based Digital Image Watermarking," icctd,
ENHANCEMENT 0.983 1 1 vol. 1, pp.535-539, 2009 International
(contrast limit-0.6- Conference on Computer Technology and
0.9) Development, 2009
GAUSSIAN 0.964 0.972 0.985 [6] Maha Sharkas, Dahlia ElShafie, and Nadder
NOISE(mean- Hamdy, “A Dual Digital-Image
0.01,var-0.01) Watermarking Technique” World Academy
of Science, Engineering and Technology 5
SPECKLE (0.06) 0.965 0.972 0.988
2005
[7] Wang Huai-bin; Yang Hong-liang; Wang
Chun-dong; Wang Shao-ming; , "A New
7. CONCLUSION Watermarking Algorithm Based on DCT
In this paper, we implemented different and DWT Fusion," Electrical and Control
watermarking algorithms using MATLAB Engineering (ICECE), 2010 International
SIMULINK based on DCT, DWT and combination Conference on , vol., no., pp.2614-2617, 25-
of DCT&DWT transform. Simulation results show 27 June 2010
that DWT is somewhat better than DCT but, [8] Dongyang Teng; Renghui Shi; Xiaoqun
combination of these two gives much better results Zhao; , "DCT image watermarking
than individual one. The capability of watermarking technique based on the mix of time-
algorithm is robust to salt & pepper, enhancement domain," Information Theory and
and speckle, but somewhat weaker to rotate and Information Security (ICITIS), 2010 IEEE
Gaussian noise. Moreover, we can implement robust International Conference on , vol., no.,
watermarking using combined DCT&DWT pp.826-830, 17-19 Dec. 2010
transforms with advanced modification in the [9] Chandra, M.; Pandey, S.;, "A DWT domain
algorithm for much better results in Simulink and visible watermarking techniques for
hardware implementation can be done in the next digital images," Electronics and Information
step. Engineering (ICEIE), 2010 International
Conference On , vol.2, no., pp.V2-421-V2-
REFRENCES 427, 1-3 Aug. 2010
[1] Pankaj U.Lande, Sanjay N. Talbar, G.N.
Shinde “Robust Image Adaptive
Watermarking Using Fuzzy Logic an FPGA
Approach”, International Journal of Signal
Processing, Image Processing and Pattern
Recognition, Vol. 3, No. 4, December, 2010
[2] Kalra, G. S., Talwar, R., & Sadawarti, H.
(2011). Robust Blind Digital Image
Watermarking Using DWT and Dual
Encryption Technique. 2011 Third
International Conference on Computational
Intelligence Communication Systems and
Networks, 225-230. Ieee.
[3] Shanshan Zhang, Xiaohong Wang, Shizheng
Zhou, "The Research of Image
Watermarking Encryption Algorithm," cso,
pp.821-824, 2011 Fourth International Joint
Conference on Computational Sciences and
Optimization, 2011
[4] Mei Jiansheng, Li Sukang and Tan Xiaomei
“A Digital Watermarking Algorithm Based
On DCT and DWT”, Proceedings of the
2009 International Symposium on Web
Information Systems and Applications,
Nanchang, P. R. China, May 22-24, 2009,
pp. 104-107

1141 | P a g e

Gh2411361141

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Gh2411361141 (20)

Gh2411361141