International Journal of Computer Applications (0975 8887)

Volume 31 No.6, October 2011

40

Permutation Based Invisible Digital Watermarking
Technique using DCT Domain
Manish Choubisa
Research Scholar
Gyan Vihar University
Jaipur, India

Kamal Hiran
Research Scholar
Gyan Vihar University
Jaipur, India
S. K. Singh
Associate Professor
Gyan Vihar University
Jaipur, India


ABSTRACT
Digital watermarking is the modern idea in digital media for
copyright protection. Many watermarking algorithm has been
developed in recent years, but context of the purposes, as they
serve, they contrast from each other. Here we propose algorithm
of digital watermarking technique based on DCT (Discrete
Cosine Transformation) using permuting the image. Through
adjusting the block DCT coefficient of the image the
watermarks are invisible. The images are first permuted and then
converting into block allowing to 88 pixel and thus the
watermark images are embedded through adjusting their DCT
coefficient. The proposed paper also describes the experimental
results that the method has strong robust.
General Terms
Watermarking, Image processing, Watermark embedding and
Extracting, Permutation.
Keywords
Digital watermarking, DCT coefficient, PSNR and SM, MBEC
1. INTRODUCTION
In digital technology all types of multimedia digital products
such as text, image, video, audio, digital repositories and
libraries, web publishing image, video frequency and audio
frequency, are released by network mode. Watermarks used to
protect digital products with different algorithms and so research
of digital watermark technology is rapidly developed.
The watermark is a digital code embedded in the image and used
for the embedded transmission of binary information such that
the watermark signal is unobstructed and secure in the digital
mixture and never remove [1].
An efficient digital watermarking algorithm is the one which
finds a good balance between invisibility and robustness [2].
The DCT is common transform for image process. The
important feature of DCT that removing correlation makes it the
perfect solution to our problem.
We apply DCT to permuted image and then embed the
watermark. After this we can achieve a better balance between
robustness and invisibility. To ensure robustness, the watermark
image or information is usually redundantly distributed over
many samples (pixels) or sub-images of the host data, thus
providing "Holographic" robustness which means that the
watermark can usually be recovered from a small fraction of the
watermarked data. Usually robustness, imperceptibility and
watermark capacity have to be traded against each other.
In general, watermark systems use one or more
cryptographically secure keys to ensure security against
manipulation and erasure of the watermark. Figure 1 shows the
basic principal block diagram of watermarking.



Figure 1: block diagram of watermarking process

Watermarking is easy to manipulate multimedia products and
make its unauthorized duplication and distribution. This has
resulted in copyright protection of digital contents on the
internet and protecting rights of buyer and establishing his
ownership of legal copy.
2. CONCEPT AND PRINCIPLE OF DCT
BASED ALGORITHM
The DCT based on Mid Band Exchange Coefficient (MBEC)
algorithm. DCT-based methods divide image into 88 sized
blocks and then transforms image of size NN into the DCT
coefficients matrix with the same size [3]. MBEC use the one bit
of binary watermark image. The MBEC watermarking algorithm
encodes one-bit of the binary watermark image into one 8x8
DCT sub-block of the original image. If the difference of two
mid-band coefficients is positive in case of the encoded value is
1 means the first coefficient is small then second coefficient
then we encoded value is 1. Otherwise, these two mid-band
International Journal of Computer Applications (0975 8887)
Volume 31 No.6, October 2011
41

coefficients are exchanged. Classical middle-band based
algorithm is quite robust against JPEG compression and
common image manipulation operations. The basic idea of the
classical MBEC scheme was discussed in [4].
The middle-band frequencies (F
M
) of an 8*8 DCT block can be
shown below in figure 2. In 8x8 DCT block the middle-band
frequencies region is denoted by FM. The lowest frequency
component of the block is denoted by FL, while FH is used to
denote the higher frequency components. FM is chosen as the
embedding region as to provide additional resistance to lossy
compression techniques, while avoiding significant modification
of the cover image. Next, from the FM region, two locations P
i

(u
1
, v
1
) and P
i
(u
2
, v
2
) are chosen in the i
th
DCT block for
comparison.

Figure 2: Definition of DCT Region
The choice in selecting the two locations is dependent on the
content of the JPEG quantization table given below in table1.
The two locations which have identical quantization values are
chosen for embedding one watermark bit of information [5].
From the table the coefficients at (5, 2) and (4, 3) have value 22
or (2, 3) and (4, 1) have value 14, would make suitable
candidates for comparison, as their quantization values are
equal. The DCT block will encode a 0 if the value of the first
pixel position is greater than or equal to the second pixel
position other vise it will encode a 1. In other word if P
i
(u
1
,
v
1
) P
i
(u
2
, v
2
) then DCT block a value 0 and if P
i
(u
1
, v
1
) < P
i
(u
2
, v
2
) it will encode a 1. The coefficients are then swapped if
the relative size of each coefficient does not agree with the bit
that is to be encoded.

Table 1: Pixel Matrix used in transforms Domain

DCT-based methods which is used in this thesis, divide image
into 8 8 sized blocks and then transforms image of size N N
into the DCT coefficients matrix with the same size [6][7]. The
equations for two-dimension DCT and Inverse DCT (IDCT) are:
( ) ( ) ( ) ( )
( ) ( )
1 1
0 0
2 1 2 1
, , cos cos
2 2
N N
x y
x u y v
C u v a u a v f x y
N N
t t

= =
+ + | | | |
=
| |
\ . \ .

( ) ( ) ( ) ( )
( ) ( )
1 1
0 0
2 1 2 1
, , cos cos
2 2
N N
u v
x u y v
f x y a u a v C u v
N N
t t

= =
+ + | | | |
=
| |
\ . \ .

Here C is the DC transformed block, f is the inverse DCT block
and u = 0,1,2, ...,K 1,v = 0,1,2, ...L1; K and L are length and
breadth of the image. a(u) and a(v) are defined in the following
equation.
( )
1
0
,
2
,
1 1
u
K
u K
K
a u
=
s s

( )
1
0
,
2
,
1 1
v
L
v L
L
a v
=
s s

16 11 10 16 24 40 51 61
12 12 14 19 26 48 16 55
14 13 16 24 40 57 69 56
14 17 22 29 51 87 80 62
18 22 37 56 68 109 108 77
24 35 55 64 81 104 113 92
49 64 78 87 103 121 120 101
72 92 95 98 112 100 103 99
International Journal of Computer Applications (0975 8887)
Volume 31 No.6, October 2011
42

3. WATERMARK EMBEDDING AND
EXTRACTING PROCESS
In Embedding process insert or embed the watermark
information within the original image by modifying all or
selected pixel values (spatial domain); or coefficients (frequency
domain), in such a way that the watermark is undetectable to
human eye and is achieved by minimizing the embedding
distortion to the host image [8].
The system block diagram for the embedding process is shown
in Figure 3.




Figure 3: Watermark Embidding Process
The watermark extraction follows a reverse embedding
algorithm, but with a similar input parameter set [9]. In this
paper we used the DCT domain for Watermark embedding and
extracting process on permuted image.
4. WATERMARK EMBEDDING
ALGORITHM
The algorithm which is used to embed a watermark on an image
is given below. We take input as Original Image and Watermark
data and produce output as Watermarked Image.

Step1: Start
Step2: Read input gray scale Image in 256 256 standard and
Watermark Data.
Step3: If the Watermark is small then it is padded with ones
(1s) so the small watermark image scale up to the max message
length for original image.
Step4: Permuted the original image using pseudo random
sequence.
Step5: Perform DCT on Each 8*8 block of image and Embed
the watermarking information into the (3,3) and (4,4) pixel of
the 8x8 DCT coefficient block by classical coefficient exchange
scheme. DCT of each block is calculated.
Step6: DCT co-efficient at the position say (3,3) and (4,4) are
compared for every block. The DCT block will encode a 0 if
pixel at position (3,3) is greater than or equal to the pixel at the
position (4,4) otherwise it will encode a 1. The coefficients are
then swapped if the relative size of each coefficient does not
agree with the bit that is to be encoded.
Step7: Re-permuted the image.
Step8: Stop.

The insertion of the watermark in the mid band of the coefficient
block of each averaged DCT block gives extra robustness to the
watermark.
5. WATERMARK EXTRACTING
ALGORITHM
For check the original image is watermarked or not, after
embedding the watermark into original image we apply the
watermark extracting algorithm. The algorithm which is used to
extract a watermark is given below.

Step1: Start
Step2: Permuted the Watermarked Image pseudo random
sequence.
Step3: Subdivide the Watermarked image into 8x8 sub-images
using DCT domain.
Step4: DCT co-efficient at the position say (3,3) and (4,4) are
compared for every block. If pixel at position (3,3) is greater
than or equal to the pixel at the position (4,4) then Watermark
bit hidden would be black or DCT block will encode a 0 else
white or DCT block will encode a 1.
Step5: Stop.
6. SIMULATION RESULTS AND THEIR
ANALYSIS
Two metrics for quality of watermarked images have been used
which are Peak Signal to Noise Ratio (PSNR) and Similarity
Factor(SM).
In order to test the performance of this watermarking scheme,
we have used 256256 gray scale images which are Lighthouse,
Girl and Pepper. The original watermark is shown in figure 4.
The watermarked images and the extracted watermark are
shown in figure 5-7.
For image with 255 gray levels, the PSNR is defined as:

( )
2
10
255
10log PSNR db
MSE
| |
= |
|
\ .

Where MSE is the mean square error of two images of N x N
pixels is defined as

( )
2
'
2
1 1
1
N N
ij ij
i j
MSE p p
N
= =
=


Where P
ij
is the original pixel value and
'
ij
p is the reconstructed
pixel value.
The similarity factor has value [0,1] calculated using following
equation . If SM = 1 then the embedded watermark and the
extracted watermark are same. Generally value of SM>.75 is
accepted as reasonable watermark extraction.

( ) ( )
( ) ( )
*
1 1
2 2
*
1 1 1 1
, ,
, ,
M N
M M
i j
M N M N
M M
i j i j
W i j W i j
SM
W i j W i j
= =
= = = =
=

Where W
M
is Original Watermark and
*
M
W is detected
watermark.
International Journal of Computer Applications (0975 8887)
Volume 31 No.6, October 2011
43


Figure 4: Original Watermark and extract watermark


Figure 5: Permuted Lighthouse image and Watermarked
Lighthouse image

Figure 6: Permuted Girl image and Watermarked Girl
image


Figure 7: Permuted Pepper image and Watermarked Pepper
image
The following Table 2 shows the PSNR of the different
watermarked images and the Similarity factor (SM) of their
extract watermarks.

Table 2: PSNR and SM values for different images

Image Degree of
Permutation
PSNR(db) SM(db)
Lighthouse 100
200
400
40.36
39.93
36.13
0.89
0.90
0.97
Girl 100
200
400
43.94
43.22
40.10
0.99
0.99
1.00
Pepper 100
200
400
44.87
40.79
34.62
0.99
0.99
1.00

When PSNR is higher than 30, Watermarked image has a very
good quality and the eye could hardly tell the difference between
the original and the Watermarked image. While when SM is
higher than 0.75, the extracted Watermarked is considered as
valid one. From the above Table we can safely say that the
watermarking schema discuss in this paper has a good
invisibility and can extract the marks correct. Figure 8 show the
graph between SM and different degree of permutation.

Figure 8: SM values of different watermarked images are
increases using permutation


7. CONCLUSION
Digital Watermarking is the process of inserting watermark data
into original images in a way that the degradation of quality is
minimized and remain in an invisible level. Many digital
watermarking algorithms have been proposed in special and
transform domains. The techniques in the spatial domain still
have relative low-bit capacity. On the other hand, frequency
domain-based techniques can embed more bits for watermark
International Journal of Computer Applications (0975 8887)
Volume 31 No.6, October 2011
44

and are more robust to attack. In this paper, we use DCT
transform watermark algorithms based on robustness.
We use the permutation method for more secure watermarking
algorithm. The robustness of the watermarking methods has
been measured by computing the Peak Signal to Noise Ratio
(PSNR) of the Watermarked Image and Similarity between
original Watermark and extract Watermark using Similarity
Factor (SM).
8. ACKNOWLEDGEMENTS
We would like to express our gratitude to experts Professor
Naveen Hemrajani, Dean (Engineering), Associate Prof.
Vibhakar Pathak (HOD, IT Department) and Inder Pratap Singh
(Asst.Professor, IT) for their guidance and contributions. We
would also like to thank for the valuable informations they
provided us. We would like to thank our family members for the
love and care.
9. REFERENCES
[1] J. R. Hemandez, M. Amado, DCT domain watermarking
techniques for still images as detector performance analysis
and a new structure, in IEEE Transactions on Image
Processing, 2000, vol. 9, pp. 55-68.
[2] Q. Du, Color image digital watermarking algorithm based
on DCT and quantifying, in Journal of Soochow
university, 2006, vol. 26, pp. 47-51.
[3] Z. M. Zhang, L. Wang, Semiblind image watermarking
algorithm in DCT domain with chaotic encryption, in
Computer Engineering, 2003, vol. 29, pp. 10.
[4] L. S. Liu, R. H. Li, Q. Gao, Method of embedding digital
watermark into the green component of color image, in
Journal of XianJiaotong university, 2004, vol. 38, pp. 1256-
1259.
[5] N. Bourbakits and C. Alexopoulos, Picture data
encryption using scan patterns, Pattern Recognition, Vol.
25, No. 6, 1992, pp. 567581.
[6] L. Wei, H. T. Lu, F. L. Chung, Robust digital image
watermarking based on subsampling, in Applied
Mathematics and Computation, 2006, vol. 181, pp. 886-
893.
[7] S. Z. Yu, A color image-adptive watermark based on
wavelet transform, in Computer Simulation, 2006, vol. 23,
pp. 132-134.
[8] C. C. Chang and H. M. Tsai, A generalized secret sharing
scheme, Journal of Systems and Software, Vol. 36, No. 3,
1997, pp. 267272.
[9] C. C. Chang, J. Y. Hsiao, and C. S. Chan, Finding optimal
LSB substitution in image hiding by using dynamic
programming strategy, Pattern Recognition, Vol. 36, No.7,
2003, pp.15831595.