Robust Color Image Watermarking Using Nonsubsampled Contourlet Transform

Document Sample
Robust Color Image Watermarking Using Nonsubsampled Contourlet Transform Powered By Docstoc
					                                                       (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                          Vol. 9, No. 3, March 2011 

            Robust Color Image Watermarking Using
             Nonsubsampled Contourlet Transform
          C.Venkata Narasimhulu                                               K.Satya Prasad
           Professor, Dept of ECE                                             Professor, Dept of ECE,
          HIET, Hyderabad, India                                              JNTU Kakinada, India

     Abstract-                                                       applications, embedded watermark should be invisible,
     In this paper, we propose a novel hybrid spread                 robust and have a high capacity. Invisibility refers to
spectrum watermarking scheme for authentication of                   degree of distortion introduced by the watermark and its
color images using nonsubsampled contourlet transform                affect on the viewers and listeners. Robustness is the
and singular value decomposition. The host color image               resistance of an embedded watermark against
and color watermark images are decomposed into                       intentional attack and normal signal processing
directional sub- bands using Nonsubsampled contourlet                operations such as noise, filtering, rotation, scaling,
transform and then applied Singular value decomposition              cropping and lossey compression etc. Capacity is the
to mid frequency sub-band coefficients. The singular                 amount of data can be represented by embedded
values of mid frequency sub-band coefficients of color
watermark image are embedded into singular values of
mid frequency sub-band coefficients of host color image in                    Watermarking techniques may be classified in
Red, Green and Blue color spaces simultaneously based on             different ways. The classification may be based on the
spread spectrum technique. The experimental results                  type of watermark being used, i.e., the watermark may
shows that the proposed hybrid watermarking scheme is
robust against common image processing operations such
                                                                     be a visually recognizable logo or sequence of random
as, JPEG, JPEG 2000 compression, cropping, Rotation,                 numbers. A second classification is based on whether
histogram equalization, low pass filtering ,median                   the watermark is applied in the spatial domain or the
filtering, sharpening, shearing ,salt & Pepper noise,                transform domain. In spatial domain, the simplest
Gaussian noise, grayscale conversion etc. It has also been           method is based on embedding the watermark in the
shown the variation of visual quality of watermarked                 least significant bits (LSB) of image pixels. However,
image for different scaling factors. The comparative                 spatial domain techniques are not resistant enough to
analysis reveals that the proposed watermarking scheme               image compression and other image processing
out performs the color image watermarking schemes                    operations.
reported recently.
    Keywords: Color image watermarking, Nonsubsampled
                                                                         Transform domain watermarking schemes such as
Contourlet Transform, Singular value decomposition, Peak             those based on the discrete cosine transform (DCT), the
signal to noise ratio, normalized Correlation coefficient.           discrete wavelet transform (DWT), contourlet
                                                                     transforms along with numerical transformations such
                      1. INTRODUCTION:                               as Singular value Decomposition (SVD) and Principle
    In recent years, multimedia products were rapidly                component analysis (PCA) typically provide higher
distributed over the fast communication systems such                 image fidelity and are much robust to image
as Internet, so there exist strong requirement to protect            manipulations.[2]Of the so far proposed algorithms,
the ownership and authentication of the multimedia                   wavelet domain algorithms perform better than other
data. Digital watermarking is a method of securing the               transform domain algorithms since DWT has a number
digital data by embedding additional information called              of advantages over other transforms including time
water mark into the digital multimedia content. This                 frequency localization, multi resolution representation,
embedding information can be later extracted from or                 superior HVS modeling, and linear complexity and
detected in the multimedia to make an assertion about                adaptively and it has been proved that wavelets are
the data authenticity. Digital watermarks remain intact              good at representing point wise discontinuities in one
under transmission/transformation, allowing us to                    dimensional signal. However, in higher dimensions,
protect our ownership rights in digital form. Absence of             e.g. image, there exists line or curve-shaped
watermark in a previously watermarked image would                    discontinuities. Since, 2D wavelets are produced by
lead to the conclusion that the data content has been                tensor products of 1D wavelets; they can only identify
modified. A watermarking algorithm consists of                       horizontal, vertical, diagonal discontinuities (edges) in
watermark structure, an embedding algorithm and                      images, ignoring smoothness along contours and
extraction or detection algorithm. In multimedia                     curves. Curvelet transform was defined to represent two

                                                                                                 ISSN 1947-5500
                                                     (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                        Vol. 9, No. 3, March 2011 
dimensional discontinuities more efficiently, with least                      2. NONSUBSAMPLED CONTOURLET
square error in a fixed term approximation. Curvelet                                   TRANSFORM
transform was proposed in continuous domain and its
discretisation was a challenge when critical sampling is           The Nonsubsampled contourlet transform is a new
desired. Contourlet transform was then proposed by DO              image decomposition scheme introduced by Arthur
and Vetterli as an improvement of Curvelet transform.              L.Cunha, Jianping Zhou and Minh N.Do [8]. NSCT is
The Contourlet transform is a directional multi                    more effective in representing smooth contours in
resolution expansion which can represents images                   different directions of in an image than contourlet
contains contours efficiently. The CT employs                      transform and discrete wavelet transform. The NSCT is
Laplacian pyramids to achieve multi resolution                     fully shift invariant, Multi scale and multi direction
decomposition and directional filter banks to achieve              expansion that has a fast implementation. The NSCT
directional decomposition [3] Due to down sampling                 exhibits a similar sub band decomposition as that of
and up sampling, the Contourlet transform is Shift                 contourlets, but without down samplers and up samplers
variant. However shift invariance is desirable in image            in it. Because of its redundancy the filter design problem
analysis applications such as edge detection, Contour              of nonsubsampled contourlet is much less constrained
characterization, image enhancement [4] and image                  than that of contourlet. The NSCT is constructed by
watermarking. Here, we present a NonSubsampled                     combining         nonsubsampled         pyramids       and
Contourlet transform (NSCT) [5] which is shift                     nonsubsampled directional filter bank as shown in
invariant version of the contourlet transform. The                 figure (1).The nonsubsampled pyramid structure results
NSCT is built upon iterated nonsubsampled filter banks             the multi scale property and nonsubsampled directional
to obtain a shift invariant image representation.
                                                                   filter bank results the directional property.
In all above transform domain watermarking techniques
including NSCT the watermarking bits would be
directly embedded in the locations of sub band
coefficients. Though here the visual of perception of
original image is preserved, the watermarked image
when subjected to some intentional attacks like
compression the watermark bits will get damaged.
Coming to the spatial domain watermarking using
numerical transformation like SVD (Gorodetski [6], liu
et al [7]) they provide good security against tampering
and common manipulations for protecting rightful
ownership. But these schemes are non adaptive, thus
unable to offer consistent perceptual transparency of                                                                                
watermarking of different images. To provide adaptive
transparency, robustness to the compressions and                                         (a)                    (b)
insensitivity to malicious manipulations, we propose a                  Figure 1 The nonsubsampled contourlet transform (a)
novel image hybrid watermarking scheme using NSCT                  nonsubsampled filter bank structure that implements the NSCT.
and SVD.                                                           (b) Idealized frequency partitioning obtained with NSCT

                                                                       2.1 Nonsubsampled Pyramids
In this paper, proposed method is compared with
another which is based on Contourlet Transform and
singular value decomposition (CT-SVD). The peak                       The nonsubsampled pyramid is a two channel
signal to noise ratio (PSNR) between the original image            nonsubsampled filter bank as shown in figure
and watermarked image and the normalized correlation               2(a).The H0(z) is the low pass filter and one then sets
coefficients (NCC) and bit error rate (BER) between                H1(z) =1-H0(z). the corresponding synthesis filters
the original watermark and extracted were calculated                   G0(z) =G1(z)=1.
with and without attacks. The results show high
improvement detection reliability using proposed                      the perfect reconstruction condition is given by
method. The rest of this paper is organized as follows.            Bezout identity
Section 2 describes the Nonsubsampled contourlet
transform, section 3 describes singular value
decomposition, section 4 illustrates the details of                    H0(z)G0(z)+H1(Z) G1 (Z) =1………………(1)
proposed method, in section 5 experimental results are
discussed without and with attacks, conclusion and
future scope are given in section 6.

                                                                                                  ISSN 1947-5500
                                                           (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                              Vol. 9, No. 3, March 2011 

                           (a)                                                     (b)
    Figure (2): Nonsubsampled pyramidal filter (a). Ideal frequency response of nonsubsampled pyramidal filter
              (b).The cascading analysis of three stages nonsubsampled pyramid by iteration of two channels
                     Nonsubsampled filter banks .

Multi scale decomposition is achieved from                                        invariant is constructed by eliminating the down and
nonsubsampled       pyramids     by     iterating the                             up samplers in the DFB.The ideal frequency response
nonsubsampled filter banks by up sampling all filters                             of nonsubsampled filter banks is shown in figure3 (a)
by 2 in both direction the next level decomposition is
achieved. The complexity of filtering is constant                                    To obtain multi directional decomposition, the
whether the filtering is with H(z) or an up sampled                               nonsubsampled DFBs are iterated. To obtain the
filter H(z m ) computed a Trous algorithm The                                     next level decomposition, all filters are up
cascading of three stage analysis part is shown in                                sampled by a quincunx matrix given by
figure 2( b)
                                                                                                             1   1 
    2.2 Nonsubsampled directional Filter Banks:                                                   Q=
The directional filter bank (DFB) is constructed from                                                        1  ‐1       ……………..(2)
the combination of critically-sampled two-channel
fan filter banks and resampling operations. The
outcome of this DFB is a tree-structured filter bank
splitting the 2-D frequency plane into wedges. The                                The analysis part of iterated nonsubsampled filter
nonsubsampled directional filter bank which is shift                              bank is shown in figure 3 (b)

                        (a)                                                                 (b)
    Figure (3) Nonsubsampled directional filter bank (a) idealized frequency response of nonsubsampled directional filter bank.(b) The
analysis part of an iterated nonsubsampled directional bank.

     3. SINGULAR VALUE DECOMPOSITION                                              data, for data compression and data denoising. If
                                                                                  A is any N x N matrix, it is possible to find a
   Singular value decomposition (SVD) is a                                        decomposition of the form
popular technique in linear algebra and it has
applications in matrix inversion, obtaining low
dimensional representation for high dimensional

                                                                                                           ISSN 1947-5500
                                                 (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                    Vol. 9, No. 3, March 2011 
                       A=USVT                                    including discrete cosine transform (DCT), discrete
                                                                 wavelet transform (DWT), Contourlet transform
                                                                 (CT) etc have been used to embed watermark into
                                                                 original image. here the proposed scheme uses
                                                                 nonsubsampled contourlet transform(NSCT) along
                                                                 with SVD for watermarking to obtain better
                                                                 performance compared to existing hybrid
                                                                              4. PROPOSED ALGORITHM
    Where U and V are orthogonal matrices of order
N x N and N x N such that UTU=I,VTV=I , and the                      In this paper, Nonsubsampled Contourlet
diagonal matrix S of order N x N has elements λi                 Transform and SVD based hybrid technique is
(i=1,2,3,..n) , I is an identity matrix of order N x N.          proposed for color image watermarking that uses
    The diagonal entries are called singular values of           true color images for both watermark and host
matrix A, the columns of U matrix are called the left            images. The robustness and visual quality of
singular values of A, and the columns of V are                   watermarked image is tested with three quantifiers
called as the right singular values of A.                        such as PSNR, NCC and Bit Error Rate. It is
    The general properties of SVD are [1], [2], [9]              investigated whether the NSCT-SVD advantages
    a) Transpose: A and its transpose AT have the                over CT-SVD for color image watermarking with
same non-zero singular values.                                   their extra features would provide any significance
    b) Flip: A, row-flipped Arf, and column-                     in terms of watermark robustness and invisibility.
flipped Acf have the same non-zero singular values.              4.1 , 4.2 explain the watermark embedding and
    c) Rotation: A and Ar (A rotated by an                       extraction algorithm [10],[11]
arbitrary degree) have the same non-zero singular                    4.1 Watermark Embedding Algorithm
    d) Scaling: B is a row-scaled version of A by                   The proposed watermark embedding algorithm
repeating every row for L1 times. For each non-zero              is shown in Figure 4. The steps of watermark
singular value λ of A, B has √L1λ. C is a column-                embedding algorithm are as follows.
scaled version of A by repeating every column for                   Step1: Separate the R G B color spaces of both
L2 times. For each nonzero singular value λ of A, C              host and watermark color images.
has √L2λ. If D is row-scaled by L1 times and
column-scaled by L2 times, for each non-zero                        Step2: Apply Nonsubsampled Contourlet
singular value λ of A, D has √L1L2λ.                             Transform to the R color space of both host image
    e) Translation: A is expanded by adding rows                 and watermark image to decompose them into sub
and columns of black pixels. The resulting matrix                bands.
Ae has the same Non-zero singular values as A.                      Step3: Apply SVD to mid frequency sub-band of
   The important properties of SVD from the view                 CT of R color space of both host and watermark
point of image processing applications are:                      image.
    1. The singular values of an image have very                     Step4: Modify the singular values of mid
    good stability i.e. When a small perturbation is             frequency sub-band coefficients of R color space of
    added to an image, their singular values do not              host image with the singular values of mid
    change significantly.                                        frequency sub-band coefficients of R color space of
                                                                 watermark image using spread spectrum technique.
   2. Singular value represents intrinsic algebraic
image properties.                                                     i.e.   λI’ = λI + α λW.,
    Due to these properties of SVD, in the last few                  Where α is scaling factor [9], λI is singular value
years several watermarking algorithms have been                  of R color space of host image, λW is singular value
proposed based on this technique. The main idea of               of R color space of watermark and λI’ becomes
this approach is to find the SVD of a original image             singular value of R color space watermarked image.
and then modify its singular values to embedded the                 Step5: Apply inverse SVD on modified singular
watermark. Some SVD based algorithms are purely                  values obtained in step4 to get the mid frequency
SVD based in a sense that only SVD domain is used                sub-band coefficients of watermarked image.
to embed watermark into original image. Recently
some hybrid SVD based algorithms have been                          Step6:      Apply inverse Nonsubsampled
proposed where different types of transform domain               Contourlet Transform to the mid frequency sub-

                                                                                             ISSN 1947-5500
                                                  (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                     Vol. 9, No. 3, March 2011 
band coefficients obtained in step 5 to get the R                     Step5: Apply inverse SVD to obtain mid
color space of watermarked image.                                 frequency coefficients of R color space of
                                                                  transformed watermark image using Step 3.
   Step7: Apply the same Steps from Step2 to
Step6 for the G and B color subspaces.                                Step6: Apply inverse NSCT using the
                                                                  coefficients of the mid frequency sub-band to obtain
   Step 8: Combine the R,G and B color spaces of
                                                                  the R color space of Watermark image.
watermarked image to obtain the color watermarked
image.                                                               Step7: Repeat the Steps 2 to 6 for G and B color
                                                                      Step8: Combine the R,G and B color spaces to
                                                                  get the color watermark. 

         Figure 4 Watermark Embeddign Algorithm
                                                                             Figure 5 Watermark Extracting Algorithm
    4.2 Watermark Extraction Algorithm
   The watermark extraction algorithm is shown in
Figure 5.      The Steps of watermark extraction                              5. EXPERIMENTAL RESULTS
algorithm are as follows.                                             In the experiments, we use the true color
                                                                  “tajmahal.jpg” of size 256X256 as host image as
   Step1: Separate the R,G,B color spaces of
                                                                  shown in the Figure 6 and true color “lena.jpg” of
watermarked image.                                                size 128 X 128 as watermark as shown in Figure 7.
   Step2: Apply Nonsubsampled Contourlet                          The experiment is performed by taking scaling
Transform to the R color space obtained in step1.                 factor alpha as 0.5.The results show that there are no
                                                                  perceptibly visual degradations on the watermarked
   Step3: Apply SVD to mid frequency sub-band of                  image shown in Figure 8 with a PSNR of
R color space of transformed watermarked image.                   45.2253dB. Extracted watermark without attack is
                                                                  shown in Figure 9 with NCC around unity and BER
    Step4: Extract the singular values from mid
                                                                  of 0.1339. MATLAB 7.6 version is used for testing
frequency sub-band of R color space of
                                                                  the robustness of the proposed method.  
watermarked and host image
                                                                      The proposed algorithm is tested for different host
    i, e λW   =   ( λI’ - λI )/ α                                 images such as “lotus.jpg”, ”Baboon.jpg”,
Where λI is singular value of watermarked image.                  ”Barbara.jpg”,     ”Way.jpg”      ,”Horse.jpg”     and
                                                                  “Wheel.jpg” as shown in Table 1 and it is observed
                                                                  that there are no visual degradations on the respected

                                                                                              ISSN 1947-5500
                                                 (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                    Vol. 9, No. 3, March 2011 
watermarked images. For all the different Host test                   color space of host image using eq.3 [12]. The
images, the watermark is effectively extracted with               final PSNR of watermarked image is taken as mean
around unity NCC. Various intentional and non-                    of PSNR obtained with three color spaces. The
intentional attacks are tested for robustness of the              similarity of extracted watermark with original
proposed       watermark       algorithm      includes            watermark embedded is measured using NCC. The
JPEG,JPEG2000compressions, low pass filtering,                    NCC is calculated using eq. (4) [13]for the three
Rotation, Histogram Equalization ,Median Filtering,               color spaces and their mean is taken as the resultant
Salt &Pepper Noise, Weiner Filtering, Gamma                       Normalized Correlation coefficient.  The proposed
Correction, Gaussian Noise, Rescaling, Sharpening                 method is also tested for binary and grayscale
Blurring ,Contrast Adjustment ,Automatic cropping,                watermark image of size 128x128 and watermarked
Dilation, Row Colum Copying, Row Colum                            and extracted watermark are shown in table 3.
removing, color to Gray scale conversion ,shearing
and sharpening. The term robustness describes the
watermark resistance to these attacks and can be
measured by the bit-error rate which, is defined as the
ratio of wrong extracted bits to the total number of                                 ……….….(3)
embedded bits.                                                        Normalized Correlation Coefficient:
    In table 2, extracted watermark and attacked
watermarked image with NCC and BER are shown.
The quality and imperceptibility of watermarked
image is measured by using PSNR. The PSNR is                                                                ………..(4)
calculated separately for R, G, B color space of
watermarked image with respect to the respective

    Fig 6:Original image-      Fig 7:Watermark image-           Fig 8:Watermarked Lena            Fig 9:Extracted
       "Tajmahal.jpg”                "Lena.jpg”                     PSNR= 45.2253                   Watermark


           Original image           Watermark image             Watermarked image with           Extracted image
            “lotus.jpg”              “LENA.jpg”                     PSNR=46.2785              NCC= 0.9983,Ber=0.1610

                                                                                             ISSN 1947-5500
                                (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                   Vol. 9, No. 3, March 2011 

                                               Watermarked image with
    Original image   Watermark image                                            Extracted image
    “baboon.jpg”      “LENA.jpg”                                             NCC=0.9992, Ber=0.1342

    Original image   Watermark image           Watermarked image with           Extracted image
    “barbara.jpg”     “LENA.jpg”                   PSNR=44.4930             NCC=0.9994,Ber=0.1299

    Original image   Watermark image           Watermarked image with           Extracted image
      “way.jpg”       “LENA.jpg”                   PSNR=44.7550             NCC= 0.9994, Ber=0.1140

    Original image   Watermark image           Watermarked image with          Extracted image
     “horse.jpg”      “LENA.jpg”                  PSNR= 44.7308             NCC= 0.9994, Ber=0.1201

    Original image   Watermark image           Watermarked image with          Extracted image
     “wheeljpg”       “LENA.jpg”                  PSNR= 45.5204             NCC= 0.9985, Ber=0.1614

                                                                            ISSN 1947-5500
                                                    (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                       Vol. 9, No. 3, March 2011 

                                       WATERMARKED IMAGE

       Jpeg compression Ncc= 0.9985,Ber=0.3306                               Jpeg2000Ncc= 0.9995,Ber=0.1056

      Salt & pepper noise Ncc= 0.6948, Ber=0.4503                       Low Pass filtering Ncc= 0.9729 Ber=0.2995

      utomatic cropping Ncc= 0.9538 Ber=0.3449                       Histogram Equalization Ncc= 0.9808 Ber=0.3128

          Rotation Ncc= 0. 0.9951 Ber=0.2958                             Median filtering Ncc= 0.9484 Ber=0.3178

    Contrast adjustment Ncc= 0.9985    Ber= 0.1613                         Weiner filter Ncc= 0.9982 Ber=0.2051

                                                                                                ISSN 1947-5500
                                            (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                               Vol. 9, No. 3, March 2011 

    Gamma correction Ncc= 0.9989 Ber=0.1387                      Gaussian Noise Ncc= 0.8399 Ber=0.3120

       Sharpening Ncc= 0.8379 Ber=0.3967                         Gaussian Blurring Ncc= 0.9719 Ber=0.3003

        Shearing Ncc= 0.9744 Ber=0.2889                                Dilatations= 0.9443 Ber=0.3332

    Color to grayscale Ncc= 0.8163 Ber=0.3490                  Row & column removal Ncc=0.9977 Ber=0.1930

                                                                                        ISSN 1947-5500
                                                (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                   Vol. 9, No. 3, March 2011 
       Row column copying Ncc= 0.9902 Ber=0.9734                              Scaling (150%)     Ncc = 0.9187



Original image                 Binary Watermark image                Watermarked image                Extracted image
“tajmahal.jpg                        “ksp.bmp”.                   PSNR= 47.6710                 Ncc= 0.9995, Ber=0.0157

      Original image            Binary Watermark image            Watermarked image                    Extracted image
       “tajmahal.jpg                  “lena.bmp”.                        PSNR= Inf                     Ncc= 1,Ber= 0

Original image “tajmahal.jpg   Gray scale Watermark image         Watermarked image                   Extracted image
                                       “Lena.jpg”.                 PSNR=45.2629                 Ncc= 0.9992,Ber= 0.1345

                                                                    salt pepper noise, Rotation, Gaussian Noise,
   In table 4, the proposed method is compared
                                                                    Sharpening, Row and Colum removal and Row
with contourlet and SVD based algorithm [11].It
                                                                    and column copying.
demonstrates that proposed method is superior to

                                                                                              ISSN 1947-5500
                                                  (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                     Vol. 9, No. 3, March 2011 
           TABLE 4: COMPARISON OF CT+SVD AND                                            7. REFERENCES:
                       NSCT + SVD
                                                                  [1].      C.Venkata Narasimhulu &K.Satya Prasad:”A novel robust
                                                                         watermarking technique based on nonsubsampled contourlet
    S.No    ATTACK              Normalized Correlation                   transform and SVD”, International Journal of multimedia and
                               NSCT+SVD      CT+SVD                      Applications.vol.3, no.1, Feb2011.
    1       Jpeg                 0.9985        0.9996
            compression                                           [2].    C.Venkata Narasimhulu &K.Satya Prasad:”A hybrid
    2       Jpeg2000             0.9995        0.9996                    watermarking scheme using contourlet transform and
    3       Salt & pepper        0.6948        0.6823                    singular value Decomposition”, IJCSNS: International
                                                                         Journal of Computer Science and Network Security.vol.10,
            noise                                                        no.9, Sep2010.
    4       Low         pass     0.9729        0.9839
            filtering                                             [3]     Minh N. Do, and Martin Vetterli,        “The Contourlet
    5       Automatic            0.9538        0.9658                    Transform: An Efficient Directional Multiresolution Image
            cropping                                                     Representation” IEEE Transaction on image processing, vol
    6       Histogram            0.9808        0.9733                    14, issue no 12, pp 2091-2106, Dec 2005
                                                                  [4]     Jianping Zhou; Cunha, A.L, M.N.Do, “Nonsubsampled
    7       Rotation             0.9958        0.9750                    contourlet transform construction and application in
    8       Median               0.9484        0.9680                    enhancement” IEEE Trans. Image Proc Sept. 2005.
    9       Contrast             0.9985        0.9991             [5]      Arthur L. Cunha, J. Zhou, and M. N. Do, “Nonsubsampled
            adjustment                                                   contourlet transform: filter design and applications in
    10      Weiner filter        0.9982        0.9989                    denoising” IEEE International conference on image
                                                                         processing, September 2005.
    11      Gamma                0.9989        0.9995
            correction                                            [6]     V.I.Gorodetski L.J.Popyack, and V.Samoilov, “SVD-based
    12      Gaussian Noise       0.8399        0.7538                     approach to transparent embedding data into digital
    13      Sharpening           0.8379        0.8212                    images,” in proc. int. Workshop, MMM-ACNS,
    14      Gaussian             0.9719        0.9841                     St Peterburg, Russia, May 2001, pp.263-274.10.
                                                                  [7]    R.Liu and T.Tan, “An SVD-Based Watermarking scheme
    14      Shearing             0.9744        0.9857                     for Protecting rightful ownership,” IEEE Trans. Multimedia,
    16      dilatations          0.9443        0.9678                    vol.4, no.1, pp.121-128, Mar.2002.
    17      Color         to     0.8163        0.8693
            grayscale                                             [8]     A. L. Cunha, J. Zhou, and M. N. Do, “The Nonsubsampled
    18      Row & Colum          0.9977        0.9972                     contourlet transform: theory, design and applications,”
            removal                                                       IEEE Trans. Image Proc., vol.15, no.10, October 2006.
    19      Row       Colum      0.9902        0.9820
                                                                  [9]       Emir Ganic and ahmet M. Eskicioglu “ Robust embedding
            copying                                                         of visual watermarks using discrete wavelet transform and
    20      Scaling (150%)       0.9187        0.9417                     singular value decomposition Journal. Of Electron.
                                                                           Imaging, Vol. 14, 043004 (2005); doi:10.1117/1.2137650
                                                                          Published 12 December 2005
                   6. CONCLUSION:
                                                                  [10]    Dongyan liu,wenbo Liu,Gong Zhang,”An adaptive
In this paper, a novel robust hybrid watermarking                          watermarking scheme based on nonsubsampled contourlet
scheme is proposed for authentication of color                            transform for color image authentication”.Proceedings of
images using nonsubsampled contourlet transform                           the 2008 the 9th international conference for Young
                                                                          computer Scientist,ISBN:978-0-7695-3398-8.
and singular value decomposition. Watermark is
embedded in all color spaces of host image by                     [11]    C.Venkata Narasimhulu &K.Satya Prasad:”A new SVD
modifying singular values of mid frequency sub band                       based hybrid color image watermarking for copy right
coefficients with respect to watermark mid frequency              \         protection using Contourlet transform”, Communicated
                                                                           to    International   Journal     of     computer   and
sub band coefficient with suitable scaling factor. The                       Applications(IJCA) in March 2011.
robustness of watermark is improved for common
image procession operations by combining both the                 [12]      Ashraf. K. Helmy and GH.S.El-Taweel “Authentication
concepts of nonsubsampled contourlet transform and                         Scheme Based on Principal Component Analysis for
                                                                           Satellite Images” International Journal of Signal
singular value decomposition. The proposed                                  Processing, Image Processing and Pattern Recognition
algorithm is tested for different host images and                           Vol. 2, No.3, September 2009.
respective watermark images are obtained without
any visual degradation. The proposed algorithm                    [13]     Matlab 7.6 version, Image Processing Tool Box.
preserves high perceptual quality of the watermarked
image and shows an excellent robustness to attacks
like Salt and Pepper Noise, Gaussian Noise, Row                    
Column Copying, and Row Column Removal.

                                                                                                 ISSN 1947-5500
                                                 (IJCSIS) International Journal of Computer Science and Information Security,  
                                                                                                    Vol. 9, No. 3, March 2011 

                                                                 K.Satya Prasad
C.V Narasimhulu
                                                                 Received his Ph.D degree from IIT Madras, India. He
He received his Bachelor degree in Electronics and               is presently working as professor in ECE department,
Communication Engineering from S.V. University,                  JNTU college of Engineering Kakinada and Rector of
Tirupati, India in 1995 and Master of Technology in              JNT University, Kakinada, India. He has more than
Instruments and Control Systems from Regional                    30 years of teaching and research experience. He
Engineering College Calicut, India in 2000.He is                 published 30 research papers in international and 20
currently pursuing the Ph.D degree in the department             research papers in National journals. He guided 8
of Electronics and Communication Engineering from                Ph.D thesises and 20 Ph.D thesises are under his
Jawaharlal     Nehru     Technological      University           guidance. His area of interests is digital signal and
Kakinada, India. He has more than 15 years                       image processing, communications, adhoc networks
experience of teaching under graduate and post                   etc.., 
graduate level. He is interested in the areas of signal
processing and multimedia security



                                                                                             ISSN 1947-5500