You are on page 1of 4

2012 International Conference on Computer Science and Service System

A Blind Watermarking Thchnology Based on DCT Domain


Fang Ma,JianPing Zhang,Wen Zhang
(Institute of Information Engineering , Zhongzhou University, Zhengzhou, Henan, China 450044)

Abstract—As a new information hiding technology, digital DCT domain digital watermarking algorithm [4], meanwhile,
watermarking attracted more and more attention in copyright according to whether need the participation of original carrier
protection. This paper proposed a DCT domain blind images when it is extracted, can also divide for the blind
watermarking algorithm based on Arnold and quantification. watermark and watermarking algorithm, because of the
Firstly we conduct Arnold chaotic scrambling to the advantages of simple extraction method and strong robustness
watermarking before the embedding, to reduce the correlation etc, blind watermark algorithm has become a hot spot in
between pixels and improve the security of the watermarking; currently research [5]; According to the different methods of
then according to the bit values of watermarking which to be watermark, and can be divided into addition algorithm,
embedded, using different methods to quantize the frequency multiplicative algorithm, bit plane algorithm , statistical
coefficient of each sub-block; and then based on the coefficient algorithm , halftone , and quantization algorithm , substitution
characteristics of embedding watermarking to extract the algorithm , based on the relationship of algorithm , the adaptive
watermarking, in which the watermarking extraction process algorithm and the adaptive algorithm , etc. According to the
does not require the original image so that achieved the blind
characteristics of the technology mentioned by a large number
extraction. Numbers of experiments show that the algorithm has
good robustness and transparency in cope with common attacks. of documents and the advantage of Arnold chaotic scrambling,
this paper puts forward in the image DCT quantization
Keywords- Arnold;quantification;blind watermarking watermark algorithm.
II. DCT-BASED DOMAIN BLIND WATERMARKING
I. INTRODUCTION ALGORITHM
Along with the rapid development of computer technology The thought of DCT domain blind watermarking
and network technology, the copyright protection and algorithm is: in order to improve the security of watermark
information security of digital works become urgent needs to image watermark, such as scrambling processing before
solve the actual problem. As an important means to the digital
watermark image processing, in order to change the original
works to provide effective protection, digital watermarking
appearance of watermark image. Although there are many
technology has become one of the hotspot in academic study
[1]. As a kind of information hiding technology, Digital methods to Realize scrambling, the algorithm utilizes chaotic
watermarking [2] is a kind of new technology which can Arnold scrambling watermark image, this is because it is
realize the digital products copyright protection, tracking embed the carrier image watermark image after Arnold
piracy under the environment of opened network, it through scrambling processing, when the image is attacked, a portion of
special algorithms in Digital products such as digital images, the image is usually damaged or lost, so the embedded
video and audio products embedded some secret information, watermark image may also damaged or missing, while after the
and then through certain technology was used to extract iconic extracted watermark image reverse scrambling processing,
information in order to Anti-counterfeiting and validation such damage or loss will be scattered in the whole image, thus
copyright , and the product has robustness and not visible reducing the effect to human eye vision, accordingly improving
property requirements. the robustness of watermarking, Embedding regional chosen in
At present, digital watermarking research has attained great the intermediate frequency area of each word pieces of carrier
achievements and brought forward many excellent algorithm, image region after segmented 8*8, because carrier frequency
according to the difference of embedding area, it can be response the carrier's main profile, should not have larger
divided into spatial algorithm and transform domain algorithms distortion, the embedding will affect invisibility, And high
[3], and because spatial algorithm has the shortcomings such as frequency of the carrier is the not sensitive information
little information embedded, Lack of robustness, the watermark perceived by human perception system, but could easily be
research mostly concentrated in the transform domain in recent eliminated by compression technology, watermarking
years, the commonly used transform domain has discrete
embedding the area will influence the robustness, so embedded
Fourier transform (DFT),
discrete cosine transform (DCT) and discrete wavelet in intermediate frequency can also meet invisible and robust ;
transform (DWT) etc. In these transform of two-dimensional Embedding technology, this method using quantity of wavelet
DCT transform is the most common lossy JPEG compression coefficients of the variable is small, can satisfy not visible, The
system of the digital image, the core of its mature technology, algorithm combines the advantages of various technical, has
in DCT domain of watermark JPEG and MPEG compression the very good invisible and robust.
have stronger robustness, so now there are many studies in A. produce of watermark
Foundation item: This work was supported by the science and In order to enhance the security and robustness of
technology program item of ZhengZhou city(No. 064SGYG21128-1) watermark, The original binary image scrambling processing
Biographies: for enhancing its safety. This algorithm was used in Arnold
FangMa :( 1979- ), female, from Henan Province, Zhengzhou, scrambling [6]. Arnold transformation has periodic, it can find
interested in the research on mobile IP communication networks.

978-0-7695-4719-0/12 $26.00 © 2012 IEEE 398


DOI 10.1109/CSSS.2012.106
a relatively and the best scrambling degree, and regard the best if P(i)=0 (4)
scrambling degree as the watermark extraction keys to
increase the security of watermark. Set the original digital DCT c
ij
k u step
image for size for M*M, transformation equations of Arnold ª DCTijc º
matrix is showed as formula 1:
In formula, k round « »
¬« step ¼»
Step said quantization step length, round said round,
§ x i 1 ·
§11 ·§ x i · DCT 、ij said medium frequency coefficients after the
¨¨ ¨¨ ¸¸¨¨ ¸¸ mod M
¸¸ (1)
©12 ¹© y i ¹
© y i 1 ¹ watermark..
4) Take inverse discrete cosine transform (IDCT) for each
Among them, x i and y i mean the abscissa denotes and
sub-block, and get image G of embedded watermark.
y-coordinate in pixels of primitive watermark image the concrete embedded process, as shown in figure 1
And x i , y i mean the abscissa denotes and y-coordinate in shows:
pixels of image after scrambling processing
Arnold transform has periodic [5], that is to say ,there are Carrier 8*8 block DCT Select frequency Wat
images transform coefficients erm
a positive integer T meet formula 2 arki
ng

AT ri mod M ri
em
(2) bed
Select frequency N times A gleam of din
ª11 º coefficients Arnold conversion g
In formula, A «12 » Called 2-d Arnold matrix,
¬ ¼
Contain watermark Each sub-block
§x · image IDCT transform
ri = ¨¨ i ¸¸ ,T is for the original image is transformed by
© yi ¹
Arnold transform the cycle of original image through times, Figure 1 watermarking embedding procedure
namely Arnold transformation will get a chaos image after
n ^n 0 d n d T ` times transformation, and then will revert C. watermark extraction process
to the original image after T - n times transformation. Watermark extraction process and embedding procedure
is exactly the opposite, and the method is very simple, need
not the participation for the original image, specific steps are
B. embedding as follows:
Embedding stage is the key to the whole algorithm, 1) Segment the carriers image after embed watermark
which determines the robustness and not visible of the into 8 u 8 sub-blocks, for these sub-block ,with Zcd covered
algorithms, Procedure is as follows: NX NX
1) After n times Arnold scrambling follow the method each (0c 8 , 0d 8 ) said. And then take discrete
showed in formula 1 before embedding, then having one- cosine transform (DCT) to each sub-block Zcd, then extracted
dimensional transformation to the scrambling image, and we intermediate frequency coefficients of watermark in each sub-
will get the digital watermarking one-dimensional sequence P. block , extracted watermark bits for them, specific judgment
2) First segment size Nx u Nx for the original carriers into 8*8 methods such as formula 5, formula 6 and 7 shown as follow:
sub-blocks, for these sub-block ,with Bcd covered each
NX NX
(0c 8 , 0d 8 ) said. Then, uniform selected for ª DCTijc º (5)
s (i) round « »
watermark blocks (the algorithm selected is even block) take ¬ step / 2 ¼
discrete cosine transform (DCT) to each sub-block Bcd, choose

intermediate frequency coefficients DCT ij , then sequence In formula, Step said quantization step length, round said

embedded one-dimensional sequence P = {P (I), 0i - 1} after round, DCT ij said medium frequency coefficients after the
scrambling watermark.
watermark..
3) When the watermark embedding, being embedded
intermediate frequency coefficients according to the formula 3 When watermarking extraction, Watermark extraction of
and 4 [7] : bits follow the formula 6 and 7 shown:
If P (I) = 1
­ 1
step
t If mod (s (i), 2 = = 0) then T (i) = 0 (6)
° k u step  2 u step 若 mod( DC T c ,step) 2 Else, T (i) = 1 (7)
®
ij
DCT c 1
ij
°̄ k u step  2 u step 其他 (3) In formula, T (i) said the watermark extraction place.

399
2)Extracted watermark one-dimensional sequence T using
the above method, then put the sequence convert M u M 2d
matrix, get a scrambling watermark image of the image T - n
times Arnold transform, for the ultimate extracted watermark
image w.
Specific extraction process as figure 2 shows:

Containin 8*8 block Select frequency Extracted


g DCT coefficients watermar
watermar transform k bits a3scrambling watermark after

Watermar T-N times Two- Two-


k image Arnold dimensional dimensional
transform transformation watermarking

Figure 2 watermark extraction process

III. SIMULATION EXPERIMENT AND ANALYSIS


In order to check the effectiveness of the method, using
a4 contain the carrier image a5 extracted watermark image
simulated experiments, MATLAB7.0 ,carrier image choose Figure 3 carrier images and watermark image
512 u 512 standard test image .watermark image choose
32 u 32cadets binary images, in experiments, for watermark B. Robustness experimental analysis
scrambling for Arnold, scrambling for 11 times, the effect
fig.3 shows, quantization step length elected 72, in order to When it comes examining algorithm robustness, there are
more clearly reflect the efficiency of the algorithm, all the salt-pepper noises attack, gauss noise attack, shear common
experiments pictures are not zoom processing. geometric attacks test, combining eye observation and
common peak value signal-to-noise ratio (PSNR) and the
correlation coefficient (NC) results on a test.
A. transparency experimental analysis containing watermark image and extracted watermark
Original carriers image watermark of PSNR value is image after noise attacks are shown as figure 4 and figure 5,
35.59. with good watermark transparency, with the naked eye the quality of carrier image after attacked is good, the
from in fig. 3 a1 and b1, both can be found no significant extracted watermark distortion are small, it can be directly
difference, When it is not attacked for the including observed with the naked eye; Watermarking extraction of NC
watermarks carrier images, extracted watermark of NC value value was shown as table 1 ,which shows that this algorithm
is 1, namely it is same for the watermark extraction and the anti-noise attack effect is better.
original image, realized completely can extract, so whether
from objective or subjective speaking, modify the algorithm
has very good invisibility. Specific fig.03 shows.

b1 Noise is the carrier image. 0.001 b2 Extracted watermark

a1 primitive carrier images a2 original image

b3 Noise is the carrier of 0.005 images b4 Extracted watermark


Figure 4 salt-pepper noises attack containing watermark image and extracted
watermark image

400
simple, do not need original image of participation, realize a
blind extraction, the results of this experiment show that the
algorithm has good robustness and transparency.
Problems: to compare strong noise attack robustness
weakly. Future work will focus on resisting the attack of
strong noise, and begin to have research on DCT domain
fragile watermark..

REFERENCES
c1 Noise is the carrier of 0.001 images c2 Extracted watermark
[1] r. b. Digital Watermarking arnett J J point and Challenges: is a
Figure 5 gauss noise attack containing watermark image and extracted marketing method [J] Electronics&Communication Engineering: 27
watermark image (4) : 1029-183
[2] G.V oyatzis and I.P itas. The Use of Protection of The
TABLE 1 NOISE ATTACK WATERMARKING EXTRACTION OF NC VALUES Watermarks in Digital Products in [J] Proceedings of The IEEE,
1999,87 (7) : 1197-1207
0.001 salt- 0.005 salt- [3]. JinCong .digital watermarking theory and technology [M].
0.001
Noise pepper pepper
gaussian Beijing: tsinghua university press, 2008
attack noises noises
noise attack [4], JinCong, zheng WeiWenFen etc, Digital image watermark
attack attack
algorithm in DCT domain with chaotic encryption [J].computer
NC 0.98268 0.92082 0.96386
application, 11 (33) : 2365-2366
[5] I J, COX MILL ER M L, BLOOM J a. Digital Water2marking
2) For shear attack, and the test has the shear attack to [M]. Wang, HuangZhiBei, translate. Beijing: publishing house of
contain the watermark image separately from the carrier of electronics industry. 2003.
upper and lower and upper right, left-down, and middle. [6] Liu Fang,JiaCheng,YuanZheng. A binary image
because scrambling watermark image technology is adopted in watermarking algorithm based on Arnold transform [J]. Computer
Applications,2008,(6) :1404 - 1406.
algorithm, weakened watermark image correlation between
[7] V. Saravanan, P. K. Bora and Dghosh.Oblivious Image-Adaptive
pixels, enhanced his ability against shearing attack, specific Watermarking Using QuantizationIndex Modulation[A].The Eighth
results shows as figure 6 and table 2, extracted watermark is National Conference On Communications[C].Mumbai:Indian
very clear, high quality, which shows that this algorithm has Institute Of Technology Bombay,2002.26-27.
good effect against clipping attack .

d1 Intermediate 1/8 shear attack d2 Extracted watermark


Figure 6 shear attack containing watermark image and extracted watermark
image

TABLE 2 SHEAR ATTACK WATERMARKING EXTRACTION OF NC VALUES


Shear Upper Lower Upper Lower
middle
attack left left right right

NC 0.95673 0.96821 0.94447 0.95866 0.96757

IV. CONCLUSIONS
Combined with Arnold and quantify technology,
adopting watermarking algorithm for image watermarking
technology Arnold scrambling, not only to reduce the
correlation between pixels, but also can strengthen the security
of watermark, After using the quantification dither modulation
technology, the changes of original image become smaller, at
the same time, the adjustment for robustness and transparency
is more flexible, The algorithm of watermark extraction is

401

You might also like