Docstoc

Blind Robust Transparent DCT-Based Digital Image Watermarking for Copyright Protection

Document Sample
Blind Robust Transparent DCT-Based Digital Image Watermarking for Copyright Protection Powered By Docstoc
					                                                                (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                          Vol. 8, o. 7, 2010

  Blind Robust Transparent DCT-Based Digital Image
        Watermarking for Copyright Protection
                                                Hanan Elazhary and Sawsan Morkos
                                                  Computers and Systems Department
                                                    Electronics Research Institute
                                                            Cairo, Egypt
                                                     {hanan,sawsan}@eri.sci.eg


Abstract—This paper proposes a novel technique for grayscale                 degradation of host image quality. Unfortunately, the hiding
digital image watermarking using monochrome watermarks in                    capacity of secret data and the distortion of the host image are a
the DCT domain. The proposed technique combines the                          tradeoff since more hidden data always result in more distortion
contradictory goals of robustness against intentional and                    of the host image [2].
unintentional attacks with transparency in addition to blindness.
We propose the combination of three techniques: Torus                            Digital image watermarking is a method for ownership
Automorphism (TA) permutation, a pseudo-random number                        assertion and copyright protection. In such a technique, the
generator, and visual cryptography to achieve high degree of                 watermark is embedded into the host image such that the
robustness against intentional attacks. On the other hand,                   embedded watermark can be later extracted to make an
robustness against unintentional attacks is achieved not only by             assertion about the host image ownership. There are two
watermarking the low frequency DCT components, but also by                   essential requirements for this purpose. The first requirement is
intelligently utilizing the more robust relative values of these low         the invisibility or transparency of the embedded watermark. In
frequencies instead of their absolute values. Experimental results           other words, the embedded watermark should not be perceived
using two test images prove the robustness of the proposed                   by human eyes and should not degrade the quality of the
technique against several types of unintentional attacks: median             watermarked host image. The other requirement is the
filtering, blurring, sharpening, Gaussian noise addition, salt and           robustness of the embedded watermark. In other words, the
pepper noise addition, and JPEG compression. Blindness is                    embedded watermark should be able to resist both intentional
achieved using only few keys and two transparencies or shares
                                                                             and unintentional attacks. In intentional attacks, the attackers
for watermark extraction. In the literature, the size of the shares
is usually double the size of the host image and consequently the
                                                                             try to extract the embedded watermark for subsequent
process of generating the shares is usually time-consuming. In the           destruction. In other words, they alter some parameters of the
proposed technique, generating the shares is straightforward and             watermarked host image to prevent legal owners from
fast. Also, their size is exactly equal to the watermark size. To            extracting the watermark for ownership assertion. In
achieve 100% transparency, the proposed technique does not                   unintentional attacks, on the other hand, the watermarked host
change any pixel values of the host image.                                   image is treated using image processing techniques including
                                                                             compression and filtering. Digital image watermarking could
   Keywords- watermarking; image processing; security                        be done in the spatial domain [3-5] or in a transform domain
                                                                             including Discrete Cosine Transform (DCT) [6-8], Discrete
                                                                             Wavelet Transform (DWT) [9, 10] or a combination of such
                         I.   INTRODUCTION
                                                                             domains [11, 12]. Embedding a watermark into a transform
    The Internet has become the most popular channel for                     domain rather than the spatial domain is more robust, and has
transmitting various forms of multimedia digital data.                       higher resistance to various attacks [13]. Digital image
Multimedia data in digital format can be modified and illegally              watermarking techniques can be classified according to the
used with ease. Thus, the copyright protection of digital images             watermark extraction process as follows [4]:
transmitted over the Internet has become an important research
topic in recent years. In the past few years, two research topics               • Non-blind techniques that require both the secret key(s)
that are related to but different from the above topic have been                  for watermark embedding and the original host image.
proposed in the literature. One is image authentication and the                 •   Semi-blind techniques that require both the secret key(s)
other is image data hiding. The goal of image authentication is                     for watermark embedding and the embedded watermark
to verify the originality of an image by detecting malicious                        bit sequence.
manipulations. Most of the earlier techniques in the literature
treated some practical manipulations such as image                              •   Blind techniques that require only the secret key(s) for
compression and image enhancement as attacks. Lin and Chang                         watermark embedding. Neither the original host image
proposed an image authentication technique which can detect                         nor the embedded watermark bit sequence is needed.
malicious manipulations but allow JPEG lossy compression
                                                                                In this paper, we propose a novel digital-image
[1]. Hiding data in images involves embedding a large amount
                                                                             watermarking technique that incorporates the important
of secret data into the host image with minimal perceptible
                                                                             characteristics discussed above (robustness, transparency, and




                                                                       183                             http://sites.google.com/site/ijcsis/
                                                                                                       ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                       Vol. 8, o. 7, 2010
blindness) for embedding a monochrome watermark into a                    Automorphism (TA) permutation for scrambling the watermark
grayscale host image. Since, as mentioned in Section I, it has            [7, 14-16] before embedding into the host image. Using a
been shown that embedding a watermark into a transform                    pseudo-random number generator to select arbitrary host image
domain rather than the spatial domain is more robust, and has             blocks and/or watermark bits for embedding has also been
higher resistance to various attacks [13], the proposed                   proposed in the literature [8]. The proposed technique
technique embeds the watermark in the DCT domain. This                    combines Torus Automorphism (TA) permutation and a
technique is discussed in the rest of the paper.                          pseudo-random number generator to increase robustness
                                                                          against intentional attacks.
    The paper is organized as follows: Section II discusses
related research in the literature and discusses the shortcomings             To increase this robustness even further, digital image
that call for developing the technique proposed in the paper.             watermarking based on visual cryptography [13], [17] has been
Section III describes the proposed technique in terms of the              proposed. Visual cryptography uses a codebook to divide a
embedding and the extraction processes and emphasizes its                 watermarked host image into several different transparencies,
advantages. Section IV provides experimental results that                 shares, or sharing images. When the copyright problem needs
demonstrate the robustness of the proposed technique against              to be resolved, the ownership verification information can be
unintentional attacks. Finally, Section V provides the                    obtained by stacking several sharing images. One or more of
discussion and conclusions of the paper.                                  these shares could be registered to the Certified Authority (CA)
                                                                          for additional security and protection against intentional
                                                                          attacks. Because visual cryptography has good performance in
                      II.   RELATED WORK
                                                                          protecting copyright of digital media [18], it has gained much
     Many techniques in the literature have been proposed for             attention of researchers in recent years. The problem with these
digital image watermarking for copyright protection. It has               techniques is that generating the shares is always time
been shown that embedding a watermark into a transform                    consuming. Besides, the size of each share is usually too large.
domain rather than the spatial domain is more robust and has              For example, according to the Naor and Shamir scheme [17],
higher resistance to attacks [13]. Thus, most of these techniques         each pixel of a given host image is replaced by 2×2 pixels.
embedded the watermark in the transform domain(s) to achieve              Hence, a host image with M by N pixels can be divided into
robustness against unintentional attacks. The DCT domain is               two sharing images with 2M by 2N pixels. This scheme is the
one of these transform domains that has been frequently used in           predominant technique used in the literature for this purpose
the literature. In the DCT domain, embedding the watermark in             [13]. This calls for developing a technique that can quickly
the insignificant components (high frequencies) makes hiding              generate shares of much smaller sizes. The idea of our
the watermark easier, but the embedded watermark is less                  proposed technique is to quickly generate a different kind of
resilient to attacks. On the other hand, embedding the                    shares with much smaller size as long as they have the same
watermark in the most significant components (DC or low                   advantage: helping in obtaining ownership verification
frequencies) makes the embedded watermark robust against                  information. The proposed technique is described in the
unintentional attacks. But, this makes hiding the watermark               following section.
harder, which may degrade the host image quality. This is
because low frequency components contain more energy than
high frequency components and the human perception system                                            0    1
is more sensitive to low frequency components. Saryazdi et al.
                                                                                                     2    3
[6] embedded a given watermark bit in the five first low AC
coefficients. Similarly, Lin et al [7] embedded two watermark
bits in the seven low frequency locations. Patra et al. [8]
randomly selected a low frequency location for embedding a
given watermark bit. However, the disadvantage of these                           Figure 1. Possible embedding locations in a 4*4 DCT-transformed
                                                                                                               block.
techniques is that they physically embed the watermark bits
which result in changing some pixels of the host image and
degrading its quality. Also, another disadvantage of the above                             III. THE PROPOSED TECHNIQUE
techniques is that they utilize the absolute values of the DCT               The proposed technique can be described in terms of the
coefficients. Unfortunately, these values can easily change               watermark embedding process and the watermark extraction
under unintentional attacks. Thus, some techniques in the                 process.
literature called for 100% transparent watermarking without
physical embedding of the watermark [3, 4]. The result is a               A. The Watermark Embedding Process
watermarked image with no distortion.              The proposed
technique does not change any pixel value of the DCT-                        The steps of the proposed watermark embedding process
transformed host image and also attempts to utilize the more              can be described as follows:
robust relative values of the DCT coefficients instead of the                •   Step 1: Divide the host image into non-overlapping 4 *
absolute values.                                                                 4 blocks.
    Several research studies in the literature aimed at                      •   Step 2: Use Torus Automorphism (TA) permutation
developing watermarking techniques that are robust against                       [14] to disarrange the watermark bits using the
intentional attacks. This has been achieved by the use of Torus                  following equation:




                                                                    184                               http://sites.google.com/site/ijcsis/
                                                                                                      ISSN 1947-5500
                                                              (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                        Vol. 8, o. 7, 2010
                                                                              permuted watermark bit with value 0. It should be noted
     i*   1      1     i 
     =                 *   mod m                        (1)
                                                                              that the largest and the smallest values of the DCT
     j*   k      k + 1  j                                               coefficients are relative values that are unlikely to
                                                                        change under attacks. This technique, thus, attempts to
                                                                              utilize the more robust relative values of the DCT
    Equation (1) indicates that each bit of the watermark at                  coefficients instead of the absolute values that can easily
    location (i, j) will be moved to a new location (i*, j*).                 change under attacks. It should be also noted that we do
    Parameters k and m are secret keys needed for                             not need to physically embed the permuted watermark
    disarranging and rearranging the watermark bits. These                    bit. We only need to remember its embedding location.
    permuted watermark bits will be embedded bit by bit                       This implies that in the proposed technique the
    and line by line into the host image.                                     embedded watermark is 100% transparent. Since there
                                                                              are only 4 possible embedding locations, we need only
•   Step 3: Arrange the host image blocks sequentially one                    two bits to represent each embedding location. For
    by one and row by row. Use a pseudo-random number                         example, (0, 0), (0, 1), (1, 0), and (1, 1) can be used to
    generator to determine the sequence of host image                         represent embedding locations 0, 1, 2, and 3
    blocks used for embedding the permuted watermark                          respectively. Thus, the permuted watermark bit is
    bits. A pseudo-random number generator with a given                       replaced by two bits representing its embedding
    seed always generates the same sequence of random                         location. Steps 4 and 5 are repeated until all the
    numbers. Thus, the seed of the pseudo-random number                       permuted watermark bits have been embedded in the
    generator is a secret key needed to embed and extract                     DCT-transformed host image blocks. Since each
    the permuted watermark bits. Steps 2 and 3 offer                          permuted watermark bit is represented by its embedding
    cryptographic protection against intentional attacks                      location and since an embedding location is represented
    because the secret keys of the pseudo-random number                       by two bits, the final result is an array with double the
    generator and TA permutation are necessary for                            size of the watermark array.
    extracting and rearranging the embedded watermark for
    subsequent destruction.                                               •   Step 6: After embedding all the permuted watermark
                                                                              bits into the DCT-transformed host image blocks, the
•   Step 4: DCT-transform the next host image block from                      resultant array of the embedding locations of the
    the spatial domain to the frequency domain. Each block                    permuted watermark bits is decomposed into two arrays
    is DCT-transformed using the following equation [6]:                      with the same size. The first array stores the first bit
                                                                              representing each permuted watermark bit embedding
                                       −1   −1                                location, while the second array stores the second bit.
    DCT (i , j ) = C (i ) * C ( j ) * ∑    ∑ pixel ( x, y )                   The two arrays are of the same size as the watermark
                                     x =0 y = 0                               array and are considered to be the watermark shares or
                                                              (2)
           ( 2 x + 1)iπ   ( 2 y + 1) jπ                                   transparencies [17]. For example, suppose that the
    * cos
           2
                         cos  2
                                         
                                           
                                                                              embedding location of the permuted watermark bit (a, b)
                                                                              is location 1 as shown in Figure 1. This embedding
                                                                              location is represented by (0, 1). Thus, the
                                                                              corresponding value of the bit (a, b) in share or
                             1   for i , j = 0                                transparency 1 is 0 and in share or transparency 2 is 1.
    where C(i ), C ( j ) =                                                    One of the generated transparencies or shares is public,
                             2 otherwise                                      while the other is secret and is registered to the Certified
                                                                              Authority (CA) for additional security and protection
                                                                              against intentional attacks (that attempt to extract the
    In equation (2), DCT (i, j) represents the value at                       embedded watermark for subsequent destruction) as
    location (i, j) in the DCT-transformed block, while pixel                 mentioned before.
    (x, y) represents the value at location (x, y) in the                 •   Step 7: Finally, inverse DCT-transform each
    original block. N is the number of locations (pixels) in                  watermarked block from the frequency domain to the
    each block.                                                               spatial domain forming the watermarked host image
•   Step 5: Embed the next permuted watermark bit into the                    using the following equation [6]:
    low frequencies of the DCT-transformed block. To
    increase the robustness of the embedded watermark                                             −1 −1
    against unintentional attacks, as explained in Section II,                pixel ( x, y ) = ∑     ∑ C (i ) * C ( j ) * DCT (i, j )
                                                                                               i =0 j =0
    the locations for embedding the permuted watermark bit                                                                               (3)
    are restricted to the low frequency locations 0, 1, 2, and                       ( 2 x + 1)iπ   ( 2 y + 1) jπ 
                                                                              * cos
    3 as shown in Figure 1. The embedding location with                              2
                                                                                                   cos  2
                                                                                                                   
                                                                                                                     
    the largest DCT coefficient is selected as the embedding
    location for a permuted watermark bit with value 1,
    while the embedding location with the smallest DCT
    coefficient is selected as the embedding location for a




                                                                    185                           http://sites.google.com/site/ijcsis/
                                                                                                  ISSN 1947-5500
                                                              (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                        Vol. 8, o. 7, 2010
                                                                           C. The Secret Keys
                                1   for i, j = 0                               Several kinds of secret keys are utilized in the proposed
       where C(i ), C ( j ) =                                              technique. These secret keys are necessary in the proposed
                                2 otherwise                                watermark embedding and extracting processes. They should
                                                                           be preserved well for watermark extraction and ownership
                                                                           verification and to increase the robustness of the embedded
       It is worth noting that in the proposed technique, this             watermark against intentional attacks (that attempt to extract
       step does not need to be explicitly performed. This is              the embedded watermark for subsequent destruction). These
       because the watermark bits are not physically embedded              secret keys are as follows:
       into the host image blocks as explained above.                         •   The two shares or transparencies needed for
                                                                                  reassembling the embedding locations of the permuted
B. The Watermark Extraction Process                                               watermark bits. The size of the transparencies is the
    The steps of the proposed watermark extraction process are                    same as that of the watermark.
the reverse of the steps of the watermark embedding process.
They can be described as follows:                                             •   The seed of the pseudo-random number generator
                                                                                  needed for determining the sequence of host image
   •   Step 1: Divide the watermarked host image into non-                        blocks used for embedding (and thus extracting) the
       overlapping 4 * 4 blocks.                                                  permuted watermark bits.
   •   Step 2: Reassemble the two transparencies or shares to                 •   The k and m parameters of Torus Automorphism (TA)
       generate the permuted watermark bits embedding                             permutation. These are needed to disarrange the
       locations. Remember that the two values in each two                        watermark bits before embedding and thus to rearrange
       corresponding locations in the two transparencies are                      the extracted permuted watermark bits.
       the two values representing the embedding location of
                                                                               It is clear that the proposed technique is blind. This is
       the corresponding bit of the permuted watermark. For
                                                                           because the watermark extraction process requires only the
       example, suppose that the value of bit (a, b) in share 1 is
                                                                           secret key(s) for watermark embedding in addition to the two
       0 and that of bit (a, b) in share 2 is 1. This implies that
                                                                           transparencies representing the embedding locations of the
       the embedding location of bit (a, b) of the permuted
                                                                           permuted watermark bits. Neither the original host image nor
       watermark is that represented by (0, 1), which is
                                                                           the embedded watermark bit sequence is needed.
       location 1 as shown in Figure 1.
   •   Step 3: Arrange the watermarked host image blocks                                    IV. EXPERIMENTAL RESULTS
       sequentially one by one and row by row. Use a pseudo-
       random number generator (with the same seed as the                      This section provides experimental results that demonstrate
       embedding pseudo-random number generator) to                        the robustness of the proposed technique against unintentional
       determine the sequence of host image blocks used for                attacks. Two host images have been used in the experiments.
       embedding (and thus extracting) the permuted                        These are the boat image and the baboon image shown in
       watermark bits.                                                     Figures 2 and 3 respectively. Each of these images is of size
                                                                           256*256. The rose watermark shown in Figure 4 is of size
   •   Step 4: DCT-transform the next watermarked host                     128*128. It has been used for watermarking these two host
       image block from the spatial domain to the frequency                images.
       domain.
                                                                              Figures 5 and 6 show the extracted watermarks from the
   •   Step 5: Extract the next permuted watermark bit from                boat image and the baboon image respectively under different
       the corresponding embedding location in the low                     types of unintentional attacks: median filtering, blurring,
       frequencies of the DCT-transformed block. The                       sharpening, Gaussian noise addition, salt and pepper noise
       corresponding bits in the reassembled transparencies are            addition, and JPEG compression.
       used for specifying the embedding location in the DCT-
       transformed block as explained above. If the embedding                  It is clear that the watermarks, embedded using the
       location has the largest value among the four low                   proposed technique, are robust to these different types of
       frequency locations shown in Figure 1, the embedded                 attacks. This is not only due to embedding the watermark bits
       permuted watermark bit is a 1. Otherwise, it is a 0. Steps          in the low frequencies of the DCT-transformed blocks of the
       4 and 5 are repeated until all the embedded permuted                host image, but also due to intelligently utilizing the more
       watermark bits have been extracted.                                 robust relative values of the low frequency components instead
                                                                           of the absolute values as explained in Section III.
   •   Step 6: After extracting all the permuted watermark
       bits, use Torus Automorphism (TA) permutation to
       rearrange the watermark bits (using the same keys used
       for disarranging the watermark bits before embedding).




                                                                     186                            http://sites.google.com/site/ijcsis/
                                                                                                    ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                       Vol. 8, o. 7, 2010
                                                                          remembering the embedding locations within the low
                                                                          frequency coefficients and not on changing the pixels values of
                                                                          these coefficients.
                                                                             The robustness of the proposed technique against
                                                                          unintentional attacks is demonstrated using two test images
                                                                          under several attacks: median filtering, blurring, sharpening,
                                                                          Gaussian noise addition, salt and pepper noise addition, and
                                                                          JPEG compression with different quality factors 30, 50, and 70.
                        Figure 2. The boat image.
                                                                                In summary, the paper has the following contributions:
                                                                                •     Combining the contradictory goals of robustness against
                                                                                      intentional and unintentional attacks with transparency
                                                                                      in addition to blindness.
                                                                                •     Achieving 100% transparency of the embedded
                                                                                      watermark.
                                                                                •     Achieving high degree of robustness against several
                                                                                      types of unintentional attacks.
                      Figure 3. The baboon image.
                                                                                •     The technique is blind and requires only 3 keys and two
                                                                                      transparencies.
                                                                                •     Using three techniques to achieve high robustness
                                                                                      against intentional attacks.
                                                                                •     Unlike other techniques in the literature, generating the
                                                                                      transparencies is straightforward and fast.
                      Figure 4. The rose watermark.
                                                                                •     Unlike other techniques in the literature where the size
                                                                                      of each share is usually equal double the size of the host
               V. DISCUSSION AND CONCLUSIONS                                          image, the size of each share in the proposed technique
    This paper is concerned with digital image watermarking                           is exactly equal to the watermark size
for ownership assertion and copyright protection. A novel
technique to achieve high degree of robustness with blindness                                                   REFERENCES
in addition to transparency is proposed in this paper. The                [1]       C.Y. Lin and S.F. Chang, "A Robust Image Authentication Method
proposed technique incorporates three techniques to achieve                         Distinguishing JPEG Compression from Malicious Manipulation,"
robustness against intentional attacks: Torus Automorphism                          IEEE Transactions on Circuits and Systems for Video Technology, vol.
(TA) permutation to disarrange the watermark bits before                            11, no. 2, pp. 153-168, 2001.
embedding into the host image), a pseudo-random number                    [2]       C. Yanga, W. Hua, and C. Lin, "Reversible Data Hiding by Coefficient-
generator to select a random sequence of host image blocks for                      Bias Algorithm," Journal of Information Hiding and Multimedia Signal
                                                                                    Processing, vol. 1, no. 2, pp. 91-100, 2010.
embedding the scrambled watermark bits, and visual
                                                                          [3]       C. Chang and J. Chuang, "An Image Intellectual Property Protection
cryptography to produce two shares and register one to the                          Scheme for Gray-Level Images Using Visual Secret Sharing Strategy,"
Certified Authority (CA). On the other hand, to achieve                             Pattern Recognition Letters, vol. 23, pp. 931–941, 2002.
robustness against unintentional attacks, the proposed                    [4]       M. Hua, D. Loub, and M. Chang, "Dual-Wrapped Digital
technique not only watermarks the low frequency DCT                                 Watermarking Scheme for Image Copyright Protection," Computers &
components of the host image blocks, but also intelligently                         Security, vol. 26, pp. 319-330, 2007.
utilizes the more robust relative values of these low frequencies         [5]       J. Hussein, "Spatial Domain Watermarking Scheme for Colored Images
instead of their absolute values.                                                   Based on Log-Average Luminance," Journal of Computing, vol. 2, no.
                                                                                    1, 2010.
    To achieve blindness, the technique is designed such that it          [6]       S. Saryazdi and M. Demehri, "A Blind DCT Domain Digital
only requires few keys and two transparencies for watermark                         Watermarking," in proceedings of the 3rd International Conference on
extraction. In the literature, the size of the shares is usually                    Sciences of Electronic, Technologies of Information and
                                                                                    Telecommunications, 2005.
double the size of the host image and consequently the process
                                                                          [7]       S. D. Lin, S. Shie, and J.Y. Guo, "Improving the Robustness of DCT-
of generating the shares is usually time-consuming. In the                          Based Image Watermarking against JPEG Compression," Computer
proposed technique, generating the shares is straightforward                        Standards & Interfaces, vol. 32, pp. 54–60, 2010.
and fast and the size of each share is exactly equal to the               [8]       J. Patra, J. E. Phua, and C. Bornand, "A Novel DCT Domain CRT-
watermark size.                                                                     Based Watermarking Scheme for Image Authentication Surviving
                                                                                    JPEG Compression," Digital Signal Processing, 2010.
    In addition to the above advantages, the proposed technique
                                                                          [9]       C. Temi, S. Choomchuay, and A. Lasakul, "A Robust Image
is designed to be 100% transparent so as not to degrade the                         Watermarking Using Multiresolution Analysis of Wavelet," In
quality of the host image. Embedding the watermark relies on                        proceedings of ISCIT2005, 2005.




                                                                    187                                     http://sites.google.com/site/ijcsis/
                                                                                                            ISSN 1947-5500
                                                                        (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                  Vol. 8, o. 7, 2010
[10] C. Yongqiang, Z. Yanqing, and P. Lihua, "A DWT Domain Image                    [17] N. Naor and A. Shamir, "Visual Cryptography," in Proceedings of
     Watermarking Scheme Using Genetic Algorithm and Synergetic Neural                   Eurocrypt'94, pp. 1–12, 1994.
     Network," in proceedings of the 2009 International Symposium on                [18] C. Wang, S. Tai, and C. Yu, “Repeating Image Watermarking
     Information Processing (ISIP’09), pp. 298-301, 2009.                                Technique by the Visual Cryptography,” IEICE Transactions on
[11] E. Ganic and A. M. Eskicioglu, "Robust DWT-SVD Domain Image                         Fundamentals, pp. 1589–1598, 2000.
     Watermarking: Embedding Data in All Frequencies," in proceedings of
     the 2004 workshop on Multimedia and security, 2004.
                                                                                                                 AUTHORS PROFILE
[12] S. Bedi, A. Kumar, and P. Kapoor, "Robust Secure SVD Based DCT-
     DWT Oriented Watermarking Technique for Image Authentication," in                   Hanan Elazhary received her B.Sc. degree in Electronics and
     proceedings of International Conference on IT to Celebrate S.                  Communications Engineering and her M.Sc. degree in Computer Engineering
     Charmonman's 72nd Birthday, 2009.                                              from the faculty of Engineering, Cairo University, Egypt. She received her
                                                                                    Ph.D. degree in Computer Science and Engineering from the University of
[13] D. Lou, H. Tso, and J. Liu, "A Copyright Protection Scheme for Digital         Connecticut, USA. Currently, she is working as a researcher at the Electronics
     Images Using Visual Cryptography Technique," Computer Standards &              Research Institute, Cairo, Egypt. She is also working as an assistant professor
     Interfaces, vol. 29, pp. 125–131, 2007.                                        at Akhbar Elyom Academy, Cairo, Egypt. Her research interests include high
[14] G. Voyatzis and I. Pitas, "Applications of Toral Automorphisms in              performance computing (HPC), software engineering, artificial intelligence,
     Image Watermarking," In proceedings of the Image Processing                    and computer networks. Elazhary has supervised several graduation projects
     International Conference, 1996.                                                and is currently supervising one Ph.D. student in the field of high-performance
[15] C. Chang, J.Y. Hsiao, and C.L. Chiang, "An Image Copyright                     computing.
     Protection Scheme Based on Torus Automorphism," in proceedings of                   Sawsan Morkos received her B.Sc. degree in Electronics and
     the 1st International Symposium on Cyber Worlds, pp. 217–224, 2002.            Communications Engineering and her M.Sc. degree and Ph.D. degree in
[16] M. Engedy, V.N.K. Munaga, and A. Saxena, "A Robust Wavelet Based               Computer Engineering from the faculty of Engineering, Cairo University,
     Digital Watermarking Scheme Using Chaotic Mixing," in proceedings              Egypt. Currently, she is working as a researcher at the Electronics Research
     of the 1st International Conference on Digital Information                     Institute, Cairo, Egypt. Her research interests include image processing and
     Management, pp. 36–40, 2006.                                                   artificial intelligence.




 Median filtering 9*9          Median filtering 3*3                 Blurring 9*9                      Blurring 3*3                     Sharpening




    Gaussian noise             Salt and pepper noise          JPEG Compression 70               JPEG Compression 50             JPEG Compression 30
      addition                        addition
                                        Figure 5. The extracted watermarks from the boat image under different attacks.




 Median filtering 9*9          Median filtering 3*3                 Blurring 9*9                      Blurring 3*3                     Sharpening




    Gaussian noise             Salt and pepper noise          JPEG Compression 70               JPEG Compression 50             JPEG Compression 30
      addition                        addition
                                      Figure 6. The extracted watermarks from the baboon image under different attacks.




                                                                              188                                  http://sites.google.com/site/ijcsis/
                                                                                                                   ISSN 1947-5500

				
DOCUMENT INFO
Description: Vol. 8 No. 7 October 2010 International Journal of Computer Science and Information Security