Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more
Download
Standard view
Full view
of .
Look up keyword
Like this
3Activity
0 of .
Results for:
No results containing your search query
P. 1
Ubiquitous Computing and Communication Journal_13

Ubiquitous Computing and Communication Journal_13

Ratings: (0)|Views: 33|Likes:
Published by Usman Tariq

More info:

Published by: Usman Tariq on Oct 31, 2008
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

05/09/2014

pdf

text

original

 
Ubiquitous Computing and Communication Journal
 
1
 
AN EFFICIENT BLOCK-BY-BLOCK SVD-BASED IMAGEWATERMARKING SCHEME
R. A. Ghazy
#
, N. A. El-Fishawy
#
, M. M. Hadhoud
$
, M. I. Dessouky
#
and F. E. Abd El-Samie
#
 # Dept. of Electronics and Elect. Communications., Fac. of Electronic Eng., Menoufia Univ., 32952, Menouf ,EGYPT.
$
Dept. of Inform. Tech., Faculty of Computers and Information , Menoufia Univ., 32511, Shebin Elkom , EGYPT.E-mails:eng_rasg@yahoo.com, nelfishawy@hotmail.com, mmhadhoud@yahoo.com and fathi_sayed@yahoo.com
ABSTRACT
This paper presents a block based digital image watermarking scheme that isdependent on the mathematical technique of singular value decomposition (SVD).Traditional SVD watermarking already exists for watermark embedding on the image as awhole. In the proposed approach, the original image is divided into blocks, and then thewatermark is embedded in the singular values (SVs) of each block separately. Thissegmentation and watermarking process makes the watermark much more robust to theattacks such as noise, compression, cropping. Watermark detection is implemented byextracting the watermark from the SVs of the watermarked blocks. Experiments showthat extracting the watermark from one block at least is enough to ensure the existence of the watermark.
Keywords
: Image Processing, Watermarking, Singular Value Decomposition.
1 INTRODUCTION
The spreading of digital multimedianowadays has made copyright protection anecessity. Authentication and information hidinghave also become important issues. To achievethese issues, watermarking technology is used.Several researchers have worked in the field of watermarking for its importance [1-11]. The work in this field has led to several watermarkingtechniques such as correlation-based techniques,frequency domain techniques, DFT basedtechniques and DWT based techniques [2].Watermarking means embedding a pieceof information into multimedia content, such asvideo, audio or images in such a way that it isimperceptible to a human observer, but easilydetected by a computer or detector [1]. Before theemergence of digital image watermarking, it wasdifficult to achieve copyright protection,authentication and data hiding but now it is easyto achieve these goals using watermarkingtechniques. Every watermarking algorithmconsists of an embedding algorithm and adetection algorithm.Watermarking has several properties such asrobustness, fidelity, and tamper-resistance [1].The robustness means that the watermark must berobust to transformations that include commonsignal distortions such as digital-to-analog,analog-to-digital conversion, and lossycompression. Fidelity means that the watermark should be neither  noticeable to the viewer nor degrading for the quality of the content. Tamper-resistance means that the watermark is oftenrequired to be resistant to signal processingalgorithms. The existence of these propertiesdepends on the application. The watermark can be embedded in the spatial domain or in thetransform domain [2].The SVD mathematical technique provides an elegant way for extracting algebraicfeatures from an image. The main properties of the SVD matrix of an image can be exploited inimage watermarking. The SVD matrix of animage has good stability. When a small perturbation is added to an image, large variationof its SVs does not occur [3], [4]. Using this property of the SVD matrix of an image, thewatermark can be embedded to this matrixwithout large variation in the obtained image.Liu et al. have proposed an SVD basedwatermarking scheme in which the watermark isadded to the SVs of the whole image or a apart of it [3]. A single watermark is used in this schemewhich may be lost due to attacks. To avoid thisdisadvantage, we propose an approach in which ,the original image is segmented into blocks andthe watermark is added to the SVs of each block in a modified manner. The SVs of the modified
 
Ubiquitous Computing and Communication Journal
 
2watermarked blocks are used to extract thewatermark after the attacks. As a result of usingseveral watermarked blocks, several watermarkscan be recovered. So if any attack affects thewatermarked image, some of the watermarks willsurvive. This block-by-block method givesrobustness against JPEG compression, cropping, blurring, Gaussian noise, resizing and rotation asthe results will reveal. The watermark can either  be a pseudo-random number, or an image. In this paper the watermark used is an image.This paper is organized as follows:Section 2 briefly explains the SVD-Basedwatermarking scheme. Section 3 introduces the proposed scheme. Section 4 introduces theexperimental results and section 5 gives theconcluding remarks.
2 TRADITIONAL SVD-BASEDIMAGE WATERMARKING
The SVD of an image is computed to obtaintwo orthogonal matrices U and V and a diagonalmatrix S [7]. In the approach proposed by Liu etal., the watermark W is added into the matrix Sthen a new SVD process is performed on the newmatrix S+kW to get U
w
, S
w
and V
w
[3]. k is thescale factor that controls the strength of thewatermark embedded to the original image. Thenthe watermarked image F
w
is obtained bymultiplying the matrices U, S
w
, and V
T
. The stepsof watermark embedding are summarized asfollows:1.
 
The SVD is performed on the originalimage (F matrix).F=USV
T
(1)2.
 
The watermark (W matrix) is added to theSVs of the original matrix.D=S+kW (2)3.
 
The SVD is performed on the new modifiedmatrix (D matrix).D=U
w
S
w
V
wT
(3)4.
 
The watermarked image (F
w
matrix) isobtained by using the modified matrix (S
w
 matrix).5.
 
F
w
=US
w
V
T
(4)To extract the possibly corrupted watermark from the possibly distorted watermarked image,given U
w
, S, V
w
matrices and the possiblydistorted image F
w, ,
the above steps are reversedas follows:1.
 
The SVD is performed on the possiblydistorted watermarked image (F*
w
matrix).F*
w
=U*S*
w
V*
T
(5)2.
 
The matrix that includes the watermark iscomputed.D*=U
w
S*
w
V
wT
(6)3.
 
The possibly corrupted watermark isobtained.W*=(D*-S)/k (7)The * refers to the corruption due to attacks.
3 THE PROPOSED WATERMARKINGAPPROACH3.1 Watermark Embedding:
In this approach the original matrix (Fmatrix) is divided into blocks and the watermark is embedded to the diagonal matrix (S matrix) of each block giving new matrices. An SVD is performed on each of these new matrices to getthe SV matrices of the watermarked image blocks. Then, these SV matrices are used to buildthe watermarked image blocks. By combiningthese blocks again into one matrix of the originalimage dimensions, the watermarked image F
w
is built in the spatial domain. The steps of embedding the watermark can be summarized asfollows:1. Divide the original image (F matrix) into non-overlapping blocks.2. Perform SVD on each block (B
i
matrix) toobtain the SVs (S
i
matrix) of each block.Where i=1,2,3,…..,N, and N is number of blocks.B
i
=U
i
S
i
V
iT
(8)4.
 
Add the watermark image (W matrix) to theS matrix of each block.D
i
=S
i
+kW (9)5.
 
Perform SVD on each D
i
matrix to obtainthe SVs of each (S
wi
matrix).D
i
=U
wi
S
wi
V
wiT
(10)6.
 
Use the (S
wi
matrix) of each block to buildthe watermarked blocks in the spatialdomain.
 
Ubiquitous Computing and Communication Journal
 
3B
wi
=U
i
S
wi
V
iT
(11)7. Rearrange the watermarked blocks back intoone matrix to build the watermarked image in thespatial domain (F
w
matrix).
3.2 Watermark Detection:
 Having U
wi
, V
wi
, S
i
, matrices and possiblydistorted image F*
w
, we can follow the stepsmentioned below to get the possibly corruptedwatermarks.1.
 
Divide the watermarked image (F*
w
matrix) into blocks having the same size used in theembedding process.2.
 
Performs SVD on each watermarked block (B*
wi
 matrix) to obtain the SVs of each one (S*
wi
 matrix).B*
wi
=U
i
*S*
wi
V
i
*
T
(12)
 
3.
 
Obtains the matrices that contain the watermark using U
wi
, V
wi
, S*
wi
, matrices.D*
i
= U
wi
S*
wi
V
wiT
(13)4.
 
Extract the possibly corrupted watermark (W*matrix) from the D
i
matrices.(D*
i
-S
i
)/k=W*
i
(14)
4 EXPERIMENTAL RESULTS
In this section several experiments arecarried out to compare between the methods of Liu et al. and the proposed approach. The256x256 cameraman image is used to bewatermarked. Figure 1 shows the original image,the watermark, the watermarked image, and theextracted watermark using Liu method. A singlewatermark is used. Figure 2 shows the originalimage, the block based watermark, thewatermarked image and extracted watermark.The block of extracted watermarks which givesmaximum correlation with the original watermark  block is magnified in the figure. The correlationcoefficients between the original transmittedwatermark block and the watermark extractedfrom each block in the image using the proposedmethod are indicated in Fig.(2-f) . The size of each block used in our experiments is 16
×
16.Different block sizes can be used but this size ismoderate having small complexity. Figure (2-f)indicates that the correlation coefficient is higher than 0.5 for all extracted watermarks. Thisensures the ability of the proposed algorithm toextract the watermarks perfectly in the absence of any attacks. Notice also that there is nodifference between the original image and thewatermarked image using the human eye,enforcing the fidelity of this method.Applying some attacks such as Gaussiannoise, blurring, cropping, JPEG compression,rotation and resizing on the watermarked images.Figures (3) and (4) show the attackedwatermarked images for Liu method and the proposed method, respectively. The major  problem encountered with attacks is the processof watermark extraction which is studied inFigs.(5) and (6).The first attack applied is Gaussian noisewith zero mean and 0.01 variance. The secondattack is blurring using a low pass filter of 3x3window. The third attack is cropping half of thewatermarked image. The fourth attack is JPEGcompression. The fifth attack is rotation by 15degree. The sixth attack is resizing from size256
×
256 to 128
×
128 and then to 256
×
256. Figure(5) shows the extracted watermark and thecorrelation coefficient between each extractedwatermark and the original watermark for themethod of Liu. The results reveal that the valueof the correlation coefficient is less than 50% for extracted watermarks under attacks except for thecompression attack.Figure (6) shows the extracted watermarksfor the proposed algorithm after applying thesame attacks we applied on Liu method. Theextracted watermark giving the maximumcorrelation coefficient with the originalwatermark block is zoomed out in the figure, andthe maximum correlation coefficient value isshown. In all cases, there is some blocks withcorrelation coefficient higher than 50% ensuringthe existence of the watermark. Table (1) givescorrelation coefficient results after applyingGaussian noise attacks with different values of noise variance. The table gives the highestcorrelation and number of extracted watermark  blocks with correlation coefficients higher thanthe predetermined threshold for 0.5 and 0.4thresholds. Similarly, Table (2) gives correlationcoefficient results after applying lowpass filteringattacks with filters of different window sizes.Correlation (1) refers to the maximum correlationobtained by the proposed method and correlation(2) refers to the correlation obtained by Liumethod. These results reveal the ability of the proposed algorithm to extract watermarks even inthe presence of severe attacks.Figure (7) shows the relation betweendifferent values of noise variance and the number of successfully extracted blocks using 0.5 and 0.4thresholds, respectively. Notice that the number 

Activity (3)

You've already reviewed this. Edit your review.
1 hundred reads
rarariz liked this
demane liked this

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->