Learning Center
Plans & pricing Sign in
Sign Out

DCT Based Robust Reversible Watermarking For Geometric Attack


International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) is an online Journal in English published bimonthly for scientists, Engineers and Research Scholars involved in computer science, Information Technology and its applications to publish high quality and refereed papers. Papers reporting original research and innovative applications from all parts of the world are welcome. Papers for publication in the IJETTCS are selected through rigid peer review to ensure originality, timeliness, relevance and readability. The aim of IJETTCS is to publish peer reviewed research and review articles in rapidly developing field of computer science engineering and technology. This journal is an online journal having full access to the research and review paper. The journal also seeks clearly written survey and review articles from experts in the field, to promote intuitive understanding of the state-of-the-art and application trends. The journal aims to cover the latest outstanding developments in the field of Computer Science and engineering Technology.

More Info
									   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

       DCT Based Robust Reversible Watermarking
                 For Geometric Attack
                           Mr.Navnath S. Narawade1 and Dr.Rajendra D.Kanphade2
                                                      Research Scholar,
                                       Dept of Electronics& Telecommunication Engg.
                                     Sant Gadgebaba Amravati University, Amravati, India
                                                Member IEEE and Principal
                                  Nutan Maharashtra Institute of Engineering and Technology,
                                                         Pune, India.

                                                                recovers 100 % similar to original map. Recently, with
                                                                the quick development of multimedia and network
Abstract: Reversible watermarking algorithms are used for       technologies, the transmission of digital products has
military security applications such as sending and receiving    become more widespread. Because digital products can
important map of a region. Here, we represent reversible and    quite easily be copied and modified, the problem of
roust watermarking technique based on DCT, which perfectly      protecting their copyrights will be seriously challenged.
embeds the watermark in image and also extracts the original    Digital watermarking, which is an auxiliary method for
watermark successfully, though image is affected with           traditional encryption measures, can effectively protect
geometric attacks such as scaling and translation. The          the copyrights of digital products.
proposed scheme embeds data by modifying those DCT
                                                                J.R.Hemandez,M.Amado have presented DCT domain
coefficients with high frequency characteristics. The
                                                                watermarking techniques for still images .They have
modifications are done in some of the high freq coefficients
value accordingly with the watermarks embedding bit. We
                                                                analyzed DCT domain with frequency coefficients[1]. Yi
present here embedding of bits using high frequency and         Du, Ting Zhang have analyzed A Reversible and Fragile
middle frequencies. In few images overflow mapping is           Watermarking Algorithm Based on DCT. They have
essential. In many images we need not to bother about           changed high frequency coefficients as per embedding bit
mapping. Original image is not essential for extraction of      value. This algorithm has less embedding capacity.But
original image. When bit 1 is embedded in the DCT block,        this algorithm is very robust to geometric attack[2]. B.
middle and high frequency coefficients are modified, while      Yang, M. Schmucker, X. Niu, C. Busch, S. Sun says that
bit 0 can be embedded without any modifications in the block    Integer DCT Coefficients techniques with Histogram
coefficients. This reversible method of watermarking is also    Modification is reversible[3]. Basic DCT algorithm is
robust to geometric attack. 100% similar watermark can be       useful for reversible watermarking[5-7].
recovered. In our method bits are embedded into high and
                                                                According to the differences in capacities of thwarting
middle frequency coefficients, which makes the watermark
more robust. 100% similar watermark and original image can
                                                                attacks, digital watermarking can be classified into robust
be recovered if our method is supported with average counter    watermarking, semi-fragile watermarking and fragile
geometric attack.                                               watermarking.
Keywords: Reversible watermarking, DCT, geometric               Robust watermarking is quite robust to modification and
attack, high and middle frequency coefficients, PSNR.           mainly used to protect intellectual property of digital
                                                                products. Semi-fragile watermarking is partly robust, but
                                                                it’s only fragile to intended tampering[8].
                                                                Fragile watermarking, which is fragile to each kind of
                                                                modification, is mainly used for authentication of
With increase in technology, usage of digital media has         integrity and reliability of digital products. In some
increased tremendously. With the increasing availability        fields, such as military, medical and legal fields, the
of digitally stored information and the development of          requirements for the quality of images are rigidly
new multimedia services, security questions are becoming        demanded, i.e., any drop in image quality isn’t permitted
even more urgent. The acceptance of new services                if the embedding of the watermarks leads to the drop. A
depends on whether suitable techniques for the protection       new type of watermarking technology called reversible
of the watermarks are available? The development of             watermarking is presented recently to meet such needs .
Internet and numerous hardware and software                     “Reversible” means that watermarked images can be
applications have created a need for security. In a military    restored to original ones, which also means that
applications, where the hidden data is of prime                 watermarks can be completely reversed. Reversible
importance. A map of a specific region is hidden inside a       watermarking has a widely used prospect, so it has caused
image. Though image is affected with geometric attack           many emphases from information scientists [2]. A
such as translation and scaling attack, our method
Volume 1, Issue 2 July-August 2012                                                                               Page 27
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

reversible and fragile watermarking algorithm is               DCT:-
proposed in this paper, which can simultaneously embed
the fragile and reversible watermark in an image.
Geometric attack can be easily handled by this algorithm.
Experiments have shown that the algorithm is effective.
Moreover the nature of digital media threatens its own
viability:                                                     IDCT:-
1) The replication of digital products is very easy.           f(x,y)=
2) The ease of transmission and multiple uses is very
worrying, too. Once a single pirate copy has been made,
it becomes available to anyone who wants it, without any
control of the original picture owner.                         where,
3) Eventually the plasticity of digital media is a great
menace. Any user (a pirate) can modify an image at will.         1.2 Implementation of DCT algorithm:
                                                               Block diagram of proposed method:-
Digital watermarking has been widely used to protect the
copyright of digital images. In order to strengthen the
intellectual property right of a digital image, a trademark                             Watermark
of the owner could be selected as a watermark and
embedded into the protected image. The image that
embedded the watermark is called a watermarked image.
Then the watermarked image could be published, and the            Original        DCT   Embedding                 Watermarked
owner can prove the ownership of a suspected image by             Image                 Process                   Image
retrieving the watermark from the watermarked image.
According to the retrieved results, we can determine the
ownership of the suspected image. Generally, a practical
                                                                                            Extraction           DCT
and useful watermarking scheme has to meet the                                              Process
following requirements after it is embedded.
1) persistence: watermarks must persist under common
image modifications (e.g., Rotationa, scaling and                     Watermark
translation )
2) robustness: watermarks must resist digital attacks to                                       Original
delete, modify or bury the watermark in another, illicit
one. Ideally, robustness also entails authorized, legitimate
recoverability of the watermark and the signature used to
create despite intervening editing.                                                Figure 1 Block Diagram
3) unobtrusiveness:both perceptible and imperceptible
watermarks should be sufficiently unobtrusive so that they     In this paper, watermarks can be embedded in the DCT
do not prevent the use and appreciation of the                 domain. The image with the size of M×N ( M and N both
watermarked object for the intended purposes.                  are the integral times of eight) will be divided into blocks
4) decodability: for some invisible watermarking               of 8×8 pixels, and then the total number of blocks is
applications, watermarks should be readily detectable by       defined as , which can be given as follows
the proper authorities, even if imperceptible to the
average observer. Such decodability without requiring the
original, un-watermarked image would be necessary for             represents the                     block. One way is
efficient recovery of property and subsequent prosecution.     to add zigzag scanned DCT coefficients, starting from
                                                                         as shown in a table I, which makes calculations
   1.1 DCT and IDCT equation:                                  complex, and other way to arrange the coefficients in a
Algorithms of digital watermarking can be classified into      simple linear order as shown in table II, starting from
three categories, which are algorithms of spatial domains,              .
transformation domains and compression domains. The
discrete cosine transform (DCT) and inverse discrete           Linear consideration of coefficients are nothing but
cosine transform (IDCT) are widely used among the              addition of middle and high frequency coefficients which
transformation algorithms of digital watermarking [6-8].       are bottom two rows. DCT coefficients whose orders are
The 2D-DCT and 2D-IDCT equations are respectively as           from                                                  in
follows:                                                       each      compose a set of high frequency coefficients
                                                               called      The number and the total value of elements

Volume 1, Issue 2 July-August 2012                                                                                              Page 28
      International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

in              each           respectively
                                   are          named        Coefficients in high-frequency domains always have
                               .                             relatively low absolute values. We have taken a high
 Table 1 DCT zigzag scanning order for high frequency        quality image called “Lena” (size: 256×256, gray-level: 0
                         .                                   - 255) was divided into blocks with the size of 8×8 pixels,
                                                             4x4 pixels, 16x16 pixels so the total number of blocks is
                                                                          If 4x4 blocks are used L=4096 and if it is
                                                             16x16 L=256. After DCT of each block, we obtain
                                                                                      of all the blocks in the image.
                                                             After plotting graph of              , most of the Si values
                                                             are between +15 and -15. If middle frequencies are
                                                             allowed then Si value increases to +20 to -20. In our
                                                             linear method also we got       very close to zero. So we
                                                             could conclude that all the       of blocks in images will
Table 2 DCT straight scanning order for high and middle
                                                             fluctuate in a minor interval named                   where
                                                                          are both quite small as shown in a graph. In
                                                             our method, num1 and num2 will always be set as
                                                                           in a linear scanning figure or last 16
                                                             coefficients in a zigzag scanning table.

                                                              1.3 Difference between Conventional and Reversible






                                                             Figure 2: Block diagram of Conventional watermarking
                0      200   400   600   800   1000   1200

          Graph1: Plot of Si Versus Di before embedding







                0      200   400   600
                                         800   1000   1200      Figure3: Block diagram of reversible watermarking
            Graph2: Plot of Si Versus Di after embedding

Volume 1, Issue 2 July-August 2012                                                                             Page 29
    International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

2. ALGORITHM                                                             7) Perform Inverse Discrete Cosine Transform to get the
                                                                         Watermarked Image.
1) Read the original image.
                                                                         8) Take average of watermark image and use it as a
2) Divide it into blocks of size 8*8.
                                                                         watermark key.
3) Perform DCT operation on each block of size 8*8.
                                                                         The exact reverse process should be followed for recovery
4) Take watermark to be embedded and find its binary
                                                                         of the Original Image and the Watermark.
5) If watermarking bit is 0,then no changes to be made in
8*8 blocks Si.                                              3. RESULT
6) If watermarking bit=1,then changes to be made in
DCTed blocks Si.
                                                 Table 3: Result table

                                                               DCT                      DCT with counter distortion
                                         Embedding in High       Embedding in High &    Embedding in High      Embedding in High &
                                         freq components         Middle freq            freq components        Middle freq components
Sr.No.   Image     Attack                PSNR     Similarity     PSNR        Similarity  PSNR       Similarity  PSNR         Similarity
                                                  measure of                 measure of             measure of               measure of
                                                  extracted                  extracted              extracted                extracted
                                                  watermark                  watermark              watermark                watermark
                                                  with                       with                   with                     with
                                                  original                   original               original                 original
                                                  watermark                  watermark              watermark                watermark
1        Lena      Scaling attack(1.1)   18.07    100%           24.1        100%         18.07      100%       24.1         100%
                   Translation(10)       18.07    100%           24.1        100%        18.07       100%             24.1    100%
                   Rotation              18.07    60.93%         24.1        60.93%      18.07       100%             24.1    100%
2        Boat      Scaling attack(1.1)   22.46    100%           20.03       100%        22.46       100%             20.03   100%
                   Translation(10)       22.46    100%           20.03       100%        22.46       100%             20.03   100%
                   Rotation(90)          22.46    60.54%         20.03       60.93%      22.46       100%             20.03   100%
3        Barbara   Scaling attack(1.1)   9.75     100%           12.60       100%        9.75        100%             12.60   100%
                   Translation(10)       9.75     100%           12.60       100%        9.75        100%             12.60   100%
                   Rotation(90)          9.75     60.54%         12.60       60.93%      9.75        100%             12.60   100%
4        Brain     Scaling attack(1.1)   13.23    100%           19.14       100%        13.23       100%             19.14   100%
                   Translation(10)       13.23    100%           19.14       100%        13.23       100%             19.14   100%
                   Rotation(90)          13.23    60.54%         19.14       60.93%      13.23       100%             19.14   100%
5        Cartoon   Scaling attack(1.1)   19.50    100%           16.64       100%        19.50       100%             16.64   100%
                   Translation(10)       19.50    100%           16.64       100%        19.50       100%             16.64   100%
                   Rotation(90)          19.50    60.54%         16.64       60.93%      19.50       100%             16.64   100%
6        Baboon    Scaling attack(1.1)   24.09    100%           25.95       100%        24.09       100%             25.95   100%
                   Translation(10)       24.09    100%           25.95       100%        24.09       100%             25.95   100%
                   Rotation(90)          24.09    60.54%         25.95       60.93%      24.09       100%             25.95   100%
7        C’man     Scaling attack(1.1)   22.89    100%           15.54       100%        22.89       100%             15.54   100%
                   Translation(10)       22.89    100%           15.54       100%        22.89       100%             15.54   100%
                   Rotation(90)          22.89    60.54%         15.54       60.93%      22.89       100%             15.54   100%
8        veg       Scaling attack(1.1)   23.57    100%           19.54       100%        23.57       100%             19.54   100%
                   Translation(10)       23.57    100%           19.54       100%        23.57       100%             19.54   100%
                   Rotation(90)          23.5     60.54%         19.54       60.93%      23.5        100%             19.54   100%

                                  *For exponential attack all images are giving 100% similarity

Volume 1, Issue 2 July-August 2012                                                                                              Page 30
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

          Watermarked Image           recovered original image            shifting method. This method’s embedding capacity can
                                                                          be increased by 4x4 DCT and further increased by 2x2
                                                                          DCT. But it reduces quality of watermarked image.
                                                                          DCT method is quite suitable for geometric attacks such
                                                                          as scaling and translation attack. In case of scaling attack,
                                                                          as every coefficient is changed but it do not affect overall
              watermark       Diff between original and recovered image
                                                                          Si out of [T1,T2]. As the sum Si of higher and middle
                                                                          frequency coefficients are not affected, a 100% similar
                                                                          watermark recovery is possible. But if image is attacked
                                                                          with rotational attack, all pixel values in image changes,
                                                                          which results into hardly 61% similarity between original
                                                                          and recovered watermark. Result table shows that this
                                                                          method is not supporting to rotational attack.

         Figure 3: Figure shows(a)watermarked
         image(b)recovered watermarked image
                                                                          4. CONCLUSION AND FUTURE SCOPE
(c)watermark(d)difference between original and recovered                  When watermarked image is affected with scaling and
                         image                                            translation attack, individual high frequency coefficients
                                                                          and middle frequency coefficients of DCT domain image
                                                                          changes, but overall addition sum( Si) do not change.
                                                                          Hence we got 100% similar watermark and recovered
                                                                          This method gives less PSNR than any other method as
                                                                          well as it gives less embedding capacity. But no any other
                                                                          method is suitable for geometric attack.
                                                                          Rotational attack and other attacks are difficult to manage
                                                                          in DCT method also. So our aim is to develop a algorithm
                                                                          for rotational attack.


                                                                          [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]Yi Du, ”, Ting Zhang,”A Reversible and Fragile
                                                                          Watermarking Algorithm Based on
                                                                          DCT”,2009,ICAICI,978-0-7695-3816-07/09-IEEE DOI
     Figure 4: Different images on which we have
                                                                          10.1109/AICI 2009.30
   implemented our algorithm and lastly a watermark
                                                                          [3] B. Yang, M. Schmucker, X. Niu, C. Busch, S. Sun,
                                                                          "Reversible Image Watermarking by Histogram
  3.1 Experiment and Discussion
                                                                          Modification for Integer DCT Coefficients", in IEEE .6th
We have performed experimentation on different images
                                                                          Workshop on Multimedia Signal Processing, pp. 143-
such as high quality images medical images and carton
                                                                          146,Siena, Italy, Sept. 2004.
images. Above result table shows that quality of
                                                                          [4]A. K. Jain, “Fundamentals of Digital Image
watermarked image reduces. We get less PSNR as
                                                                          Processing,” New Jersey: Prentice Hall Inc.,1989.
compared to histogram shifting method. But it is
                                                                          [5]A. C. Hung and TH-Y Meng, “A Comparison of fast
comparable with IWT method. As each and every
                                                                          DCT algorithms,” Multimedia Systems, No. 5 Vol. 2, Dec
coefficients are affected with this method, the quality of
watermarked image reduces. But this DCT method is
                                                                          [6] G. Aggarwal and D. D. Gajski, “Exploring DCT
robust to geometric attacks such as rotation, translation
                                                                          Implementations,” UC Irvine, Technical Report ICS-TR-
and scaling attack. Result table clearly shows that this
                                                                          98-10, March 1998.
method is 100% reversible to geometric attack. The
                                                                          [7] J. F. Blinn, “What's the Deal with the DCT,” IEEE
embedding capacity of this method is also very less as
                                                                          Computer Graphics and Applications,July 1993, pp.78-
compared to other methods such as chaotic neural
network method, difference expansion method and RCM
method. But it is quite comparable with histogram

Volume 1, Issue 2 July-August 2012                                                                                           Page 31
    International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

[8]Mohammad Awrangjeb, Manzur Murshed, and
Guojun Lu,”Global Geometric Distortion Correction in


                           Mr. Navnath S. Narawade is a
                         research scholar at Electronics and
                         Telecommunication Engg department
                         at    Sant   Gadgebaba      Amravati
                         University , Amravati. He received
                         the        M.E.(Electronics       and
                         Telecommunication Engg) in 2005
                         from Govt. College of Engg, Pune
                         under      University    of     Pune,
B.E.(Electronics ) from Walchand College of Engg, Sangli
under Shivaji University ,Kolhapur. His research interests are
focused on image and signal processing, particularly in robust
reversible watermarking.

                      Dr. Rajendra D. Kanphade is
                      presently        Principal    of Nutan
                      Maharashtra Institute of Engineering and
                      Technology, Pune. He has joined SSGM
                      College of Engineering, Shegaon in
                      1987. He has been Head of the
                      Electronics Department for the period
                      Oct 2003 to Jan 2006 and also Incharge
                      of “VLSI & Embedded System Design
                             Center” of SSGMCE, Shegaon.
He has completed B.E. (Electronics) degree from SGB
Amravati University, Amravati in the year 1987 & M.E
(Electronics) from Dr. Babasaheb Ambedkar Marathwada
University, Aurangabad in the year 1993. He received
Ph.D.(Electronics Engg.) degree from SGB Amravati
University, Amravati. His areas of research are VLSI and
Embedded system Design, Analog and Mixed Signal Design.
He has completed his Ph. D degree from SGB Amravati
University, Amravati. He is a member of IEEE, IETE and ISTE.
He has published different papers in international journals and

Volume 1, Issue 2 July-August 2012                                                    Page 32

To top