Digital Image hiding using curvelet transform

YongHong ZHANG
Department of Mathematics and Sciece of Information, Weinan Teacher' s University, Shanxi Weinan ,china 714000. Email: yongh_zhang@163.com

Abstract-This paper presents a digi tal image hiding tech nology by using the curvelet transform. Firstly, apply Arnold transform to original image; Secondly, apply curvelet Transform to the original image and the open image, gaining their curvelet coefficients; Thirdly, interpolate their curvelet coefficients; Fi nally, reconstruct the image by using Inverse curvelet Transform, and thus get the result image. Simulation results show that this approach is easy to use and safety. Index Terms-curvelet Transform; image hiding; Arnold trans form;fusion .

and location parameters. Unlike the wavelet transform, it has directional parameters, and the curvelet pyramid contains elements with a very high degree of directional specicity.In ad dition, the curvelet transform is based on a certain anisotropic scaling principle which is quite different from the isotropic scaling of wavelets. The elements obey a special scaling law, where the length of the support of a frame elements and the width of the support are linked by the relation width Basically, curvelet transform extends the ridgelet transform

length 2.
I. INTRODUCTION

Embedding images into other images has applications in data hiding and digital watermarking. During the last few years, much progress has been made in developing watermark ing techniques that are robust to signal processing operations, such as compression[11.Image encryption schemes have been increasingly studied to meet the demand for real-time secure image transmission over the Internet and through wireless networks. Traditional image encryption algorithm such as data encryption standard has the weakness of low-level efficiency when the image is large[2-31. The technique makes use of curvelet transform which rep resents the latest research result on multi-resolution analysis [4-51. By combining the advantages of the two methods, image edge information is captured more accurately than conven tional spectral methods such as wavelet and Gabor filters. Curvelet was originally proposed for image denoising and has shown promising performance. As it captures edge and linear information accurately, it has also shown promising results in character recognition recently [61. Ni and Leng attempted an initial application of curvelet on color image retrieval, but it was not implemented properly and no meaningful result was reported. In their work, no benchmark image database is used, no retrieval accuracy was reported and there was no comparison with other techniques [71. This paper presents an extended image encrypted algorithm, the algorithm applies extended discrete chaotic dynamic sys tem and curvelet transform to image encryption. Section II presents the principle of image encryption algorithm based on chaotic system. Section III presents the Encryption and decryption Algorithm. Security analysis and conclusion will be discussed in Sections IV and Section V, respectively. II. ALGORITHM PRINCIPLES

to multiple scale analysis. Therefore, lets start from the defi nition of ridgelet transform. Given an image function the continuous ridgelet transform is given as[81:

f (x,y),

1Rj(a,b,8) = 8
Where E

JJ 'I!a,b,O(X,y)f(x,y)dxdy
-! .T. 2 'I'

(1 ) and

a > 0 is the scale, b E R is the translation [0,211'] is the orientation. The ridgelet is defined as:

'I!a,b,O(x,Y) = a

x cos 8 + y sin 8 - b ) a

(2)

The digital curvelet transform is taken on a 2-D Cartesian grid

f[m,n],0 :s: m < M,O:S: n < N

CDa,b,8 (x,y) =

O<m<M On<N

L J[m,n]'I!b,O[m,n]

(3)

The digital curvelet transform is implemented using the fast discrete curvelet transform. Basically, it is computed in the spectral domain to employ the advantage of FFT. Given an image, both the image and the curvelet are transformed into Fourier domain, then the convolution of the curvelet with the image in spatial domain becomes the product in Fourier domain. Finally the curvelet coefficients are obtained by applying inverse Fourier transform on the spectral product. But due to the frequency response of a curvelet is a non rectangular wedge, the wedge needs to be wrapped into a rectangle to perform the inverse Fourier transform. The wrapping is done by periodic tiling of the spectrum using the wedge, and then collecting the rectangular coefficient area in the center. Through this periodic tiling, the rectangular region collects the wedges corresponding portions from the surrounding periodic wedges. The Curvelet transform is a higher dimensional generaliza tion of the Wavelet transform designed to represent images

A. DESCRIPTION OF CURVELET TRANSFORM

The curvelet transform, like the wavelet transform, is a multi-scale transform,with frame elements indexed by scale 978-1-4244-8728-81111$26.00 2011 IEEE

488

at different scales and different angles. Curvelets enjoy two unique mathematical properties, namely: Curved singularities can be well approximated with very few coefficients and in a non-adaptive manner - hence the name "curvelets." Curvelets remain coherent waveforms under the action of the wave equation in a smooth medium.

Step4. Interpolating curvelet coefficient gained through step3 by using fusion parameter a ; go to step2 until Transform, and thus get the result image lows:

times;

Step5. Reconstruct the image by using Inverse curvelet

RI .

The corresponding resuming algorithm is described as fol

B.

Arnold transform
Arnold transform is chosen as pretreatment method for open

B.

Resuming algorithm:
Step 1. Input end image , open image a, t and the times of

image as it is simple and periodic. Arnold transform is a kind of transform in the traversing theory, called as Arnolds cat mapping. For a digital square image, Arnold transformation is defined as:

iteration ;

xn+1 Yn+l

(xn + Yn)mod t (xn + 2Yn)mod t

(4)

of the original image and the Curvelet coefficient of the open image

EI and the open image 0PI , gaining the Curvelet coefficient

OPI; times; RIA RIA,

Step2. Apply Discrete Curvelet Transform to the end image

Where,

is the width (height) of the open image. Every

Step3.Interpoiating curvelet coefficient gained throng step3 by using fusion parameter a;go to step2 until Transform, and thus get the Resuming image and gain Resuming image Step4.Reconstruct the image by using Inverse Curvelet Step5. Apply inverse Arnold transform to the image

pixel in the image is transformed using the above formula. After traversing all of pixels in the image, a scrambled image will be gained. Transforming repeatedly an image can generate different results until reaching the requirement. Due to the Arnold transform periodicity, the original image can be recovered. C. Algorithm principles Based on the statement in section II, we can give the the Algorithm principles. The first, apply Arnold transform to original image , and gain the image ; The second,apply digital curvelet transform to the original image and the open image and gain their coefficient; interpolate the coefficients respectively; Finally, reconstruct the image by using Inverse digital curvelet transform , and thus get the result image. Resuming algorithm is the inversion of hiding procedure.

RI

IV. EXPERIMENTAL RESULTS

A. Example of digital image hiding and resume

Fig.2.show the example of digital image hiding and re sume,among that the lena.jpg is the original image,and the Cameraman.jpg is the open image.where the fusion parameter a
=

0.1, times

end imag

open Image

(a)

(b)
open Imago gray histogram

Fig.I. Algorithm principles

III. HIDING AND RESUMING ALGORITHM Based on the above analysis, we can get the corresponding hiding algorithm as follows:

(c)

(d)
"00 Image gray hlstogrnm

A. Hiding algorithm:
Stepl. Input original image , open image a and the times of iteration

times; ORI,
and and

Step2. Apply Arnold transform to original image gain the image the open image

0 RIA ORIA 0PI,

gaining their curvelet coefficients ;

Step3. Apply curvelet Transform to the image

(e)

(f)

489

andy are grey-scale values of two adjacent pixels in the image (seeTablel).
. . ' Tab.I. Corre atlOn coeffi Clents 0f two ad' l1acent plxe Is m two images original image hiding image Horizontal 0.9782 0.00083 Vertical 0.00072 0.9736 Diagonal 0.9657 0.00069

Fig.2.Hiding image and resuming image by using algorithm of this paper (a) original image (b) original image's gray histogram (c) open image (d) open image's gray histogram (e) end image (f) end image's gray histogram (g) resumed image (h) resumed image's gray histogram

(g)

(h)

D.

Security and complexity

V. CONCLUSION

In this paper, an image hiding scheme by using Curvelet Transform is proposed. From the experimental results, it is concluded that it outperforms existing schemes, both in terms of speed and security. Having a high through put, the proposed system is ready to be applied I fast real time hiding applications. Moreover, the fusion parameter a can use chaotic sequence.
ACKNOWLEDGMENT

B. Key sensitivity
High key sensitivity is required by secure cryptosystems, which means that the cipher text cannot be decrypted correctly although there is only as light difference between hiding or resuming keys. This guarantees the security of a cryptosystem against brute-force at tacks to some extent. From the simula tion results of Fig.3 which we resumed the image by using inaccuracy key a
=

The work is supported by Scientific Research Funds of Education Department of Shaanxi Province (Grant Number: 20 1OJK535) and key Funds of Weinan Teacher's University (Grant Number: 1 1YKFOI5) and key disciplinary fields of background mathematics .
REFERENCES [1] Petitcolas, F.A.P., Anderson RJ., and Kuhn, M.G.: Information Hiding-a Survey [JI. Proceedings of the IEEE, 1999,87(7),1062-1078. [2] Chen GR, Mao Y B. A symmetric image encryption scheme based on 3D chaotic cat maps [J]. Chaos,Solitons and Fractals, 2004, 21:749-761. [3] Chiaraluce F, Ciccarelli L, et al. A new chaotic algorithm for video encryption [J]. IEEE Trans Consum Electron 2002, 48:838-843. [4] J. Starck, E. J. Cands, and D. L. Donoho, The Curvelet Transform for Image Denoising, IEEE Transactions on Image Processing, 11(6):670684, 2002. [5] Minh N. Do, Directional Multiresolution Image Representations, PhD Thesis, EPFL, 2001. [6] E. Candes et al, Fast Discrete Curvelet Transforms, Multiscale Modeling and Simulation, 5(3): 861-899, 2006. [7] A, Majumdar, Bangia Basic Character Recognition Using Digital Curvelet Transform, Journal of Pattern Recognition Research, 1: 17-26, 2007. [8] L. Ni and H. C. Leng, Curvelet Transform and Its Application in Image Retrieval, 3rd International Symposium on Multispectral Image Processing and Pattern Recognition, Proceedings of SPIE, Vol. 5286, 2003. [9] J. Starck, E. J. Cands, and D. L. Donoho, The Curvelet Transform for Image Denoising, IEEE Transactions on Image Processing, 11(6):670684, 2002. [10] MATTHEWS R. On the derivation of a chaotic encryption algorithm[J]. Cryptologia, 1989,8 (1):29-42. [11] Kowalczyk R, Forys RU. Qualitative analysis on the initial value problem to the logistic equation with delay[J]. Math Compute Model, 2002, 35:1. [12] Li C, Liao X, Yu J. Hopf bifurcation in a prototype delay system [J]. Chaos Solitons and Fractals 2004, 19:779-787. [13] D. Zhao, G. Chen and W. A. Liu, A chaos-based robust wavelet-domain watermarking algorithm, Chaos, Solitons Fractals, 22(10):47-54, 2004. [14] Les Piegl, Wayne Tiller. The NURBS Book [M]. Springer. 1995 [15] Cohen DS, Rosenblat S. A delay logistic equation with variable growth rate[J]. Society for Industrial and Applied Mathematics,1982, 42(3), 608624. [16] Sadkowski A. On the application of the logistic differential equation in electro chemical dynarnics[J]. Electro Anal Chern. 2000, 486: 92-4. [17] T.Xiang, X.Liao, G.Tang, etc. A novel block crypto system based on iterating a chaotic map[J]. Physics Letters A 349. 2006, 109-115.

0.1000000000000l,we can see that the

effect of hiding and resuming using our method is good.

wrong key resume image with wrong key
wrong kOy Imago oray histogram with wrong k(ly

Fig.3 Resuming image by using inaccuracy key: (a) resuming image by using inaccuracy key(a
=

0.10000000000001,

times

5) (b) resuming

image's gray histogram by using inaccuracy key

C. Correlation of Two Adjacent Pixels

To test the correlation between two adjacent pixels in original image and encrypted image, the following procedure was carried out. Randomly select 5000 pairs of two adjacent (in vertical, horizontal, and diagonal direction) pixels from original images and encrypted image, and calculate the cor relation coefficients two formulas:

[11J, respectively by using the following

1

cov(x, y)
rxy
=

i=l cov(X, y) ,E(x) JD(x)JD(y)

n

L (Xi - E(X))(Yi - E(y))

=

D(x)
W here, of is

L (Xi - E(x))2
i=l

.!. t Xi i=l
n

E(x) is the estimation of mathematical expectations x,D(x) is the Estimation of variance of X, and cov(x, y) the estimation of covariance between X and y . where X

490