journal of global research in computer sciencejournal of ......tamper localization was achieved at...

Volume 1, No. 5, December 2010

Journal of Global Research in Computer ScienceJournal of Global Research in Computer ScienceJournal of Global Research in Computer ScienceJournal of Global Research in Computer Science

RESEARCH PAPER

Available Online at www.jgrcs.info

© JGRCS 2010, All Rights Reserved 27

DATA HIDING OF BINARY IMAGE USING DISCRETE WAVELET

TRANSFORMATION

Vaishali C. Sanap*, Aditi Jahagirdar

and Meenakshi A. Thalor

Department of Information Technology, University Of Pune, Pune ,India

[email protected]

[email protected]

[email protected]

Abstract : Watermarking is a technique of hiding information in image including scanned text, figures, and signatures in such a way that it is difficult to intercept. In

this paper, we proposed a digital image watermarking algorithm based on wavelet transform. Arnold transform and human visual characteristics is used to scramble the

original watermark. In the embedding process two sub-bands which are in the same direction but in different resolution are selected when the carrier image is

decomposed by two-layer wavelet transform. In the process of watermark extraction, the watermarks are detected by using the embedded position and scaling parameter

a after the wavelet decomposition of the watermarked image and the original image.

Keywords: Digital watermark, discrete wavelet Transform, arnold transform, image processing.

�INTRODUCTION

From the implementing point of view, the digital watermarking

algorithm can be divided into two domains, the space domain

and the frequency domain [1].LSB and Patchwork are typical

algorithms in space domain. Draw backs of space domain are:

1) The capacities of the algorithms are not large enough,

especially for small images. For example, a capacity as large as

512 bits is required to incorporate a message authentication

code such as SHA-2 (e.g., SHA-512).

2) The existing large capacity algorithm does not have good

visual quality of the watermarked image and the computational

load is relatively high. Because of this reasons more and more

algorithms have transferred to implement in frequency domain.

DFT domain, DCT (Discrete Cosine Transform) domain and

DWT domain are the familiar three frequency domains. In this

paper, an algorithm of digital image watermarking based on

discrete wavelet transform (DWT) is proposed.[2]

The paper is organized as follows. Discrete wavelet transform

is described in section 2 .In section 3 we will propose an

improved algorithm for watermark embedding and detection. In

section 4 we are concluding the paper.

DISCRETE WAVELET TRANSFORM

�

The Discrete Wavelet Transform (DWT) is a mathematical tool

that efficiently decomposes an image into multi-resolution sub-

bands. It has the characteristics that energy compacts into a few

low transform coefficients after the wavelet transform. Figure 1

shows an image that is divided into ten sub-bands through a

three-scale wavelet transform.

Fig. 1: wavelet decomposition of host image

The image is first decomposed into three high-frequency sub-

band in different directions: the level sub-bands HL, vertical

sub-bands LH, diagonal sub-bands. Process of decomposition is

repeated arbitrary number of times which depends on

application in hand in this paper it is decomposed three times or

three levels

WATERMARKING ALGORITHM

The watermarking algorithm commonly has three parts

watermarking scrambling Embedding algorithm extracting

algorithm.

Watermarking Scrambling

Previously pseudo random sequences are used as watermarking

messages [7]. Limitations of this method are it can only tell

whether the watermark exits while it cannot show what the

watermark is. In order to make sure that even if attackers

intercepted the watermarking messages, they cannot get the

exact secret messages, watermarking messages must be

scrambled. That is to say the watermarking image is

transformed to a shuffling image so that original secret

messages are hidden. Arnold transform is applied to scramble

the watermarking image.Given an N*N image, the Arnold

transform that is applied to every pixel in the image is given

© JGRCS 2010, All Rights Reserved

�

It means Arnold transfo

position (

transform are it is iterative and is periodic iterative means it is

applied to a digital image until the digital image is scrambling.

Arnold transform p

original digital image can be achieved. Its periodicity can be

achieved using following algorithm

Algorithm for Watermark Image Scrambling:

START

1. Input : x,y xn,yn,n,N;

2. Initialize x and y to 1 value.

3. for n =1 to ….

xn=x+y;

yn=x+2*y;

4. if (xn mod N==1 and yn mod N==1) then

5. x=xn mod N;

y=yn mod N;

6. End for

7. Display the value of n .

STOP

�

Assume size of watermark image is 34*34. Using the above

program, periodicity it is executed 18 times. It means we will

get original watermark image after implementing Arnold

transform 18 times. Fig 2 shows result of a

transform.

Arnold transform pretreatment eliminates spatial correlation

and disperses the error bits among all pixels to make

watermarking more strongly robust against cropping operation.

Watermark Embedding

To embedded watermark

developed.

into four parts: HH, HL, LH, and LL. The LL band can be

decomposed again into four parts. The algorithm to embed a

watermark into the host is as follows:

Step1:

and W

decompose X into (p/16x q/16) blocks with size 16x16.


It means Arnold transfo

position (x, y) to position (



Arnold transform p




START



3. for n =1 to ….

xn=x+y;

yn=x+2*y;


Break;(get out of for loop).

5. x=xn mod N;

y=yn mod N;

6. End for


STOP





transform.

Fig. 2: Watermarked Image and Scrambled Image




Watermark Embedding

To embedded watermark

developed. A step of wavelet transform decomposes an image




Step1: Let X is the original gray

W is a gray-level image watermark with size


Vaishali C. Sanap


Mod N x, y

It means Arnold transform will shift the value of the pixel at

) to position (x', y’). Characteristics of Arnold



Arnold transform periodic means after some iterative times, an








y=yn mod N;










Watermark Embedding

To embedded watermark a block DWT

step of wavelet transform decomposes an image




the original gray

level image watermark with size


Vaishali C. Sanap et al, Journal of Global Research in Computer Science,

Mod N x, y € {0, 1, 2—N-

rm will shift the value of the pixel at

', y’). Characteristics of Arnold



eriodic means after some iterative times, an


achieved using following algorithm:







transform 18 times. Fig 2 shows result of applying Arnold





a block DWT-based algorithm is





the original gray-level image with size

level image watermark with size


Journal of Global Research in Computer Science,

-1}











pplying Arnold





based algorithm is




level image with size p x q,

level image watermark with size r x s. First,












pplying Arnold




based algorithm is




p x q,

x s. First,

Step 2:

get the confused watermark

Step 3:

with ten sub bands of a pyramid structure by DWT. Sub bands

(LH3, LH2) are selected to be cast. Select the eleme

j) € LH3

its sons' coefficient in LH3.

Summation = |LH2 (2*i

And select LH2(k,l)

in LH3

put LH(k,l) into the vector VecLH2.

Step 4:

embedded into subbands of LH3 and

Watermarks W

For i = 1 to r x s

VecLH3(i)= VecLH3(i)+


End For

Step 5:

modified DWT coefficients and the unchanged DWT

coefficients to form watermarked image.

Fig. 3 shows watermark embedding process for a flower image

.it will apply block DWT to image to get decomposed image of

three levels and ten sub bands .LH3 and LH2 are selected for

embedding .At the same time watermarked Gray level image is

scrambled by A

is redundantly embedded in each block at DWT coefficients

.recursively apply Inverse DWT (IDWT) get watermarked

image which is ready for transmission

Journal of Global Research in Computer Science, 1(5),

Step 2: Processing the watermark

get the confused watermark

Step 3: For each blocks of X, it is decomposed into three levels



€ LH3 with the maximum summation of the absolute value of


Summation = |LH2 (2*i

+|LH2(2*i

+|LH2(2*i ,2*j|

And select LH2(k,l)

LH3(i, j ) 'S sons. Put LH3(i, j )into the vector VecLH3, and


Step 4: In the casting stage, watermarks are redundantly


Watermarks W' are embedded as follows:

For i = 1 to r x s



End For

Step 5: Take the two



Fig. 3: Watermark Embedding





scrambled by Arnold transform .in the casting stage watermark




1(5), December 2010,

Processing the watermark Cy

get the confused watermark w' with size (r X s).

For each blocks of X, it is decomposed into three levels



with the maximum summation of the absolute value of


Summation = |LH2 (2*i-1,2*j-1|

+|LH2(2*i-1,2*j|+|LH2(2*I ,2*j

+|LH2(2*i ,2*j|

And select LH2(k,l) € LH2 with the maximum absolute value

'S sons. Put LH3(i, j )into the vector VecLH3, and


In the casting stage, watermarks are redundantly


are embedded as follows:

VecLH3(i)= VecLH3(i)+� X W’(i/s+1, i mod s +1)

VecLH2(i)= VecLH3(i)+� X W’(i/s+1, i mod s +1)

Take the two-dimensional Inverse DWT (IDWT) of the








rnold transform .in the casting stage watermark




December 2010, 27-29

Cy with Arnold transform we

with size (r X s).





1,2*j|+|LH2(2*I ,2*j-1|

€ LH2 with the maximum absolute value




embedded into subbands of LH3 and LH2

are embedded as follows:

X W’(i/s+1, i mod s +1)

X W’(i/s+1, i mod s +1)

dimensional Inverse DWT (IDWT) of the












28

ith Arnold transform we

with size (r X s).



(LH3, LH2) are selected to be cast. Select the element LH3 (i,


1|




LH2 for robustness.

X W’(i/s+1, i mod s +1)

X W’(i/s+1, i mod s +1)



�








ith Arnold transform we



nt LH3 (i,





for robustness.











�

Watermark Extracting Method

In this method, the watermark

embedded position and scaling parameter

decomposition of the

image as shown in fig 4.

Step l:

original image

each block with DWT into three levels of ten subbands

respectively.

Step 2:

get the vector VecLH3 and VecLH2 from the subbands (LH3,

LH2)

get VecLH3' and VecLH2' from the watermarked image.

Step 3:

�for i=1 to r x s

W”

W”

W” = (W”3+w”2)/(2x

End for

Step 4:

extracted watermarks

Step 5:

extracted watermarks by the standard correlation coefficient

�



In this method, the watermark


decomposition of the


Fig 4: Watermark Extraction

Step l: First, we decompose a watermarked image

original image X into ( p/16xq/16) blocks, and then decompose


respectively.

Step 2: Using the algorithm in the embedding method, we can


LH2) of the original image blocks. Then we correspondingly


Step 3: Average and scaling down the watermarks

for i=1 to r x s

W”3 (i/s +1, I mod s+1) = VecLH3’(i)

W”2 (i/s +1, I mod s+1) =

W” = (W”3+w”2)/(2x

End for

Step 4: Processing

extracted watermarks

Step 5: Measure the similarity of original watermarks and


Vaishali C. Sanap



In this method, the watermarks are detected by using the


decomposition of the� watermarked image and the original



First, we decompose a watermarked image

into ( p/16xq/16) blocks, and then decompose


Using the algorithm in the embedding method, we can


of the original image blocks. Then we correspondingly


Average and scaling down the watermarks

(i/s +1, I mod s+1) = VecLH3’(i)

(i/s +1, I mod s+1) = VecLH2’(i)

W” = (W”3+w”2)/(2x �)

Processing W" with Arnold transform, we get the

extracted watermarks W'.

Measure the similarity of original watermarks and


Vaishali C. Sanap et al, Journal of Global Research in Computer Science,

s are detected by using the

embedded position and scaling parameter a after the wavelet

watermarked image and the original


First, we decompose a watermarked image








(i/s +1, I mod s+1) = VecLH3’(i)-VecLH3(i)

VecLH2’(i)-VecLH2(i)

with Arnold transform, we get the





after the wavelet


First, we decompose a watermarked image X' and the








VecLH3(i)

VecLH2(i)



extracted watermarks by the standard correlation coefficient as



after the wavelet�


the








as

At receiver side this algorithm will decompose original image

and watermarked image using DWT, applying watermark

embedding algorithm to get

watermark, applying Arnold transform to get original

watermark and using correlation

watermarks are detected by using the embedded position and

scaling parameter

watermarked image and the original image.

CONCLUSION

Watermarking was done in wavelet domain. Conventional

watermarking attacks were not possible. The resolution of

tamper localization was achieved at pixel level. The

watermarked image’s quality was still maintained while

providing pixel

�REFERENCES

[1] P. Meenakshi Devi, M. Venkatesan and K. Duraiswamy,

“Fragile Watermarking Scheme for Image Authentication

with Tamper Localization Using Integer Wavelet

Transform”.

Tiruchengode, India.

[2] Hu

for Binary Images in Morphological TransformDomain

High

vol. 10, no. 3, April 2008.

[3] H. Yang and Alex C. Kot, “

Binary

Preserving”

475

[4] Min Wu, and Bede Liu,

Authentication and Annotation”,

Multimedia,

[5] Y. C. Tseng and H.

Images,”

873

[6] Y. Liu, J. Mant, E. Wong and S. H. Low, “

Detection of Text Documents Using Transform

Techniques”

Conf. on Security and Watermarking of Multimedia

Contents

[7] Anfeng Hu, Ning Chen, “

for Color Image Based on Wavelet

Transform”

Central South University, hangsha, 410083, China

Journal of Global Research in Computer Science, 1(5),







scaling parameter a


CONCLUSION





providing pixel-level tampering accu

REFERENCES

P. Meenakshi Devi, M. Venkatesan and K. Duraiswamy,



Transform”. K S Rangasamy College of Technology,


Huijuan Yang, Alex C. Kot


High-Capacity Approach” IEEE

vol. 10, no. 3, April 2008.

H. Yang and Alex C. Kot, “

Binary Images Authentication by Connectivity

Preserving”,IEEE Trans. On Multimedia

475-486, April 2007.

Min Wu, and Bede Liu,


Multimedia, vol. 6, no. 4, pp. 5

Y. C. Tseng and H.

Images,” IEEE Trans. on Computers

873-878, July 2002.

Y. Liu, J. Mant, E. Wong and S. H. Low, “


Techniques”, Proc. of SPIE, Electronic Imaging (EI’99)


Contents, vol. 3657, pp. 317

Anfeng Hu, Ning Chen, “


Transform”School of Information Science and Engineering,


1(5), December 2010,







a after the wavelet decomposition of the






level tampering accu




K S Rangasamy College of Technology,


ijuan Yang, Alex C. Kot, “Orthogonal Data Embedding


Capacity Approach” IEEE

vol. 10, no. 3, April 2008.

H. Yang and Alex C. Kot, “Pattern

Images Authentication by Connectivity

IEEE Trans. On Multimedia

486, April 2007.

Min Wu, and Bede Liu, “Data Hiding in Binary Images for


vol. 6, no. 4, pp. 5

Y. C. Tseng and H.-K. Pan,

IEEE Trans. on Computers

878, July 2002.



Proc. of SPIE, Electronic Imaging (EI’99)


, vol. 3657, pp. 317-328, San Jose, CA, 1999.

Anfeng Hu, Ning Chen, “A Blind Watermarking Algorithm


School of Information Science and Engineering,


December 2010, 27-29



embedding algorithm to get W” which is scrambled


watermark and using correlation to compare images. The


after the wavelet decomposition of the






level tampering accuracy.





“Orthogonal Data Embedding


Capacity Approach” IEEE transactions on multimedia,

Pattern-Based Date Hiding for


IEEE Trans. On Multimedia, vol. 9,no.3, pp.

“Data Hiding in Binary Images for

Authentication and Annotation”, IEEE Trans. on

vol. 6, no. 4, pp. 528-538, August 2004.

K. Pan, “Data Hiding in 2

IEEE Trans. on Computers, vol. 51, no. 7, pp.





328, San Jose, CA, 1999.

A Blind Watermarking Algorithm

for Color Image Based on Wavelet Transform and Fourier



29



W” which is scrambled


to compare images. The












for Binary Images in Morphological TransformDomain-A

transactions on multimedia,

Based Date Hiding for


, vol. 9,no.3, pp.


IEEE Trans. on

538, August 2004.

“Data Hiding in 2-D Color

, vol. 51, no. 7, pp.

Y. Liu, J. Mant, E. Wong and S. H. Low, “Marking and

Detection of Text Documents Using Transform-domain



328, San Jose, CA, 1999.


Transform and Fourier


Central South University, hangsha, 410083, China.



W” which is scrambled


to compare images. The












A

transactions on multimedia,

Based Date Hiding for

Images Authentication by Connectivity-

, vol. 9,no.3, pp.


IEEE Trans. on

D Color

, vol. 51, no. 7, pp.

Marking and

domain




Transform and Fourier
