Anda di halaman 1dari 6

e-ISSN (O): 2348-4470

Scientific Journal of Impact Factor (SJIF): 4.72


p-ISSN (P): 2348-6406

International Journal of Advance Engineering and Research


Development
Volume 4, Issue 2, February -2017

Digital Watermarking using SVD-DCT Method

Sureshkumar Natarajan 1, SnehaRaju Das2

1
Department Electronics and Telecommunication, Vishwaniketan Institute of Management Entreprenuership and
Engineering Technology
2
Department Electronics and Telecommunication, Vishwaniketan Institute of Management Entreprenuership and
Engineering Technology

AbstractAmong the various methods of watermarking, this paper proposes to use the combinational method of
Singular Value Decomposition (SVD) and Discrete Cosine Transform (DCT). Suppose a highly confidential information
has to be transferred from person A to person B, the secret information is hidden beneath a cover image, such that no lay
man or a hacker will know that this cover image is carrying some really confidential message within it. This is a
proposed technique of ensuring secure transfer from one place to another. Person B, on receiving the encoded image,
will now run an extraction code over the image to decode or retrieve the actual image or information. The image to be
hidden and the cover image are both transformed using the DCT. The discrete transformed values of the actual image
are reduced in value in comparison to the cover image to make the cover image dominant over the actual image. In this
way the image is hidden behind the cover image and transmitted to the destination address. This technique has been
implemented using the MATLAB software and we are able to retrieve almost the exact image from the watermarked
image with a high level of efficiency in terms of data loss.Among the many applications, watermarking is mainly used to
source track any data i.e., if an illegal copy of the original data is found, the watermark embedded into it can be
retrieved to know the source or the authenticated owner of the copyright. The applications can be extended to security of
official documents, detection of tampering and forgery and many more.

Keywords-digital watermarking; copyright protection; Singular Value Decomposition; Discrete cosine transform.;
imperceptibility.
I. INTRODUCTION

The rapid advancements in the technology paradigm, has enabled extensive availability of electronic devices such as
computer machines, printers, scanners and many more that have made the data transmission, exchange and acquisition
tasks very simple. This possibility of accessing each other's digital data has paved the way for malicious activities of
copying and distributing copyright content without the permission of the owner. This has given birth to a quest to invent
methods and techniques that could help bring such problems under control even if it could not be totally eradicated.
Digital watermarking is that technique which was invented with a view to control piracy. However, persistent work is
still being done to improvise the technique and make it more robust and efficient.

II. DIGITAL WATERMARKING

Digital watermarking is the proposed solution for the problems of copyright protection in the network environments.
Watermarking is a combination of those methods and technologies that are capable of embedding or hiding a detectable
and extractable message/code into any digital copyright content to reduce or control piracy. The content can be in any
form like a number, text, digital media such as videos and images. The watermarking is said to be inserted or embedded
into the piece of information which is to be secured. In other words, we are embedding nothing but two important
properties namely, inseparability and imperceptibility. Information can be watermarked by using any of the available
watermarking types such as, visible, invisible, public and fragile watermarking. Visible watermarking is visible to the
human eye in case of an image and detectable by ear in case of an audio file. The invisible watermarking is only detected
by some special software. For instance, a high frequency noise that a human ear cannot hear can be inserted into an audio
file. In fragile watermarking, the digital watermark is corrupted even if a small portion of the file is modified. And
finally, in robust watermarking, the digital watermark is protected and retrieved as it is even if a portion of the file or the
entire file is modified. Imperceptibility, robustness, security, complexity and verification are the various attributes of
watermarking. Among the many applications, watermarking is mainly used to source track any data i.e., if an illegal
@IJAERD-2017, All rights Reserved 288
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

copy of the original data is found, the watermark embedded into it can be retrieved to know the source or the
authenticated owner of the copyright. The applications can be extended to security of official documents, detection of
tampering and forgery and many more.

III. SINGULAR VALUE DECOMPOSITION

Any real matrix of order m x n whether a square matrix or a rectangular matrix can be easily decomposed into a product
of three matrices, where the first matrix is a square matrix of order , denoted by , the second matrix is a diagonal
matrix of order , denoted by , whose diagonal values are the Eigen values of the original matrix and the last one
is another square matrix of order , denoted by which is the transpose matrix of the matrix obtained after
normalizing the Eigen vectors of the product of transpose of matrix and itself. The matrices on the extreme ends i.e.,
and are orthogonal matrices as their inverse are equal to their corresponding transposes respectively. In general, SVD
can be mathematically expressed as below:
= 1
where,
1 0 . 0 0
0 2 . 0 0
= 0 0 . 0 0 2
0 0 . 1 0
0 0 . 0
and and are orthogonal matrices i.e.,
= = 3
The central matrix i.e., the diagonal matrix plays a vital role in watermarking the image. The diagonal elements of
the matrix are the singular values of the original matrix A and each of them are responsible for the perceived brightness,
luminance and contrast of the image. While the central matrix has such an important role to play, the matrices at the
extremities also contribute a good deal. The matrix contains columns which are nothing but the left singular vectors of
and the matrix contains columns which are the right singular vectors of .Both of these orthogonal vectors are
responsible for managing the images' geometry. The product of all these three matrices retraces back to the original
matrix .This is the means by which the image obtained after performing DCT can be retraced to its original form so that
the image can be converted to visual mode. Because the image is not enabled for visualization as the image is
transformed into integers after the image undergoes DCT.

IV. DISCRETE COSINE TRANSFORM

The discrete cosine transform converts the image from spatial domain into frequency domain in the form of integers so
that they can be easily worked upon. It is considered to be the most efficient transform for image compression of JPEG
images. The DCT is similar to FFT (Fast Fourier Transform) but is more efficient than any other transform in terms of
energy compaction. The energy in DCT domain is concentrated into very few coefficients.

The DCT for a 1D image is given by:


1
2 + 1
= cos , 0 1 4
2
=0

1
= 0

=
2
0

The inverse DCT for a 1D image is given by:

1
2 + 1
= cos , 01 5
2
=0
@IJAERD-2017, All rights Reserved 289
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

The forward 2D discrete cosine transform of a signal is given by:

1 1
2 + 1 2 + 1
, = , cos cos , 0 , 1 6
2 2
=0 =0

1
= 0

=
2
0

Similarly,

1
= 0

=
2
0

The inverse2D discrete cosine transform of a signal is given by:

1 1
2 + 1 2 + 1
, = , cos cos , 0 , 1 7
2 2
=0 =0

4.1. The DCT operation

Figure 1. DCT Frequency Band Region

@IJAERD-2017, All rights Reserved 290


International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

An input image of order is considered. (, ) is the pixel intensity stored in rows and columns. , is the
matrix containing the image frequency integers in rows and columns after performing DCT over the image. These are
nothing but the DCT coefficients. Most of the information or energy of the signal is stored in the lower frequencies
represented by the top-left corner of the image as shown below. The down-right corner consists of high frequency
distribution while the mid portion contains the middle frequency distribution. Since maximum information is stored by
the lower frequencies, the higher frequencies almost carry negligible value and can be neglected. The negligence of these
high frequencies cause extremely minute distortion which is not visible. Factually, neglecting 50% of the high
frequencies results in only 5% loss of information. This supports the reason for DCT being the best among any other
transform. The image pixels are broken down into groups of 64 bytes initially i.e., each group containing 8x8 pixels
making a total of 64 bytes. After DCT, this number reduces to 2 to 20 bytes per group. Thus we can say that compression
has been successful. While uncompressing the image the inverse DCT algorithm is applied over 2 to 20 bytes to make an
approximation of the 8x8 pixel groups. These groups when put together form the original image. TheFigure 2.shows the
accuracy of the recovered image after inverse DCT. It is 98.869% identical to the original image.

Figure 2.

V. WATERMARKING PROCESS

5.1. The SVD-DCT Watermarking Process


The image to be hidden and the cover image are both transformed using the DCT. The discrete transformed values of the
actual image are reduced in value in comparison to the cover image to make the cover image dominant over the actual
image. This is done by multiplying the image values by a constant whose value is negligibly small. The resultant image
values are stored in a variable and called into the SVD function for decomposition. After decomposition, the resultant
values are stored in terms of a vector [U S V']. With this the image watermarking is declared complete. But at this stage
the image cannot be viewed as it is required to be transformed back from the frequency domain to the spatial domain.
The product of the image components U, S and V does the domain conversion and displays the watermarked image.

5.2. Extraction Process


The extraction process deals with the recovery of the actual image by the removal of the cover image and displaying the
actual image to the viewer. For this, SVD is again performed over the watermarked image, whose diagonal matrix is
considered. The cover image is extracted from this actual image and the frequency integers are again divided by the
constant factor to retain the original image after performing the inverse DCT.

VI. ALGORITHM
6.1. Watermarking Process:
Step1: Read the cover image
Step2: Apply DCT to cover image
Step3: Apply SVD to the resultant image of DCT
Step4: Read the actual image
Step5: Apply DCT to actual image

@IJAERD-2017, All rights Reserved 291


International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

Step6: Apply each values of diagonal matrix of singular value decomposition of the resultant image of DCT of cover
image is added with each values of image which is to be watermarked, and the image which is to be watermarked
is first multiplied with a constant value.
Step7: Apply SVD to watermarked image
Step8: To observe watermarked image, multiply the values of unitary matrices and of cover image with the diagonal
matrix values of singular value decomposition of the watermarked image. The Watermarking process is shown in
Figure 3.

6.2. Extraction Process:


Step1: Perform SVD over the watermarked image and store the values in a different set of variables of an array to obtain
the singular values of the image.
Step2: Multiply the obtained singular values with the orthogonal vectors and store the values in another variable.
Step3: Along the rows and columns of the watermarked image subtract the row and column values of the cover image
and divide the values by the constant factor that was initially multiplied, to make the original image visible.
Step4: Perform the inverse Discrete Cosine Transform over this image for the final extraction of the image.The
Extraction process is shown in Figure 4.

Figure 3.Watermarking Process

Figure 4.Extraction Process


@IJAERD-2017, All rights Reserved 292
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

VII. EXPERIMENTAL RESULTS

Following are the experimental results of the SVD-DCT based algorithm. Figure 5(a)indicates the cover image;Figure
5(b)is the actual image which is to be hidden. Figure 5(c)is the watermarked image. This image is carrying the actual
image hidden within it. And finally, Figure 5(d) is the image obtained after extraction i.e., the actual image of the girl is
separated from the cover image.
Cover Image Key Image

(a) (b)

The Watermarked Image The Extracted Image

(c) (d)

Figure 5.Experimental Results

VIII. CONCLUSION

The proposed paper highlights the success of the SVD-DCT based Watermarking technique in efficient image recovery,
in terms of absolute recovery of the actual image from the watermarked image with negligible loss of information after
executing the process of extraction. Its application is mostly extended to securing highly confidential official data by
means of encryption and decryption using the proposed algorithm.

REFERENCES

[1] Md. MaklachurRahman, A DWT, DCT and SVD Based Watermarking Technique to Protect the Image Piracy,
International Journal of Managing Public Sector Information and Communication Technologies (IJMPICT), Vol.4.
No.2, pp.21-32June2013

[2] Alexander Sverdlov, Scott Dexter, Ahmet M. Eskicioglu, "Secure DCT-SVD Domain Image Watermarking:
Embedding Data in All Frequencies.

[3] KeJi, Jianbiao Lin, HuiLi,Ao Wang, Tianjing Tang, "A DCT And SVD Based Watermarking Technique ToIdentify
Tags, arXiv: 1502.02969v1 [cs.MM], 10 Feb 2015

@IJAERD-2017, All rights Reserved 293

Anda mungkin juga menyukai