Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Quantum Inf Process (2013) 12:3689–3706

DOI 10.1007/s11128-013-0629-2

A dynamic watermarking scheme for quantum images


using quantum wavelet transform

Xian-Hua Song · Shen Wang · Shuai Liu ·


Ahmed A. Abd El-Latif · Xia-Mu Niu

Received: 3 December 2012 / Accepted: 7 August 2013 / Published online: 24 August 2013
© Springer Science+Business Media New York 2013

Abstract In this paper, a novel watermarking scheme based on quantum wavelet trans-
form (QWT) is proposed. Firstly, the wavelet coefficients are extracted by executing
QWT on quantum image. Then, we utilize a dynamic vector for controlling embedding
strength instead of a fixed parameter for embedding process in other schemes. Analy-
sis and results show that the proposed dynamic watermarking scheme has better visual
quality under a higher embedding capacity and outperforms the existing schemes in
the literature.

Keywords Quantum computation · Quantum image · Quantum wavelet transform ·


Quantum watermarking

1 Introduction

Image is one of the most important information representation styles and widely used in
most of the applications like strategic communication, telemedicine, medical images,
etc. Therefore, the protection of image data from unauthorized access by intercept-
ing the information and then visiting, copying, and destroying it has becomes a hot

X.-H. Song · S. Wang (B) · A. A. Abd El-Latif · X.-M. Niu


Department of Computer Science and Technology, Harbin Institute of Technology,
Harbin 150080, China
e-mail: shen.wang@hit.edu.cn

X.-H. Song · S. Liu


Harbin University of Science and Technology, Harbin 150080, China

A. A. Abd El-Latif
Department of Mathematics, Faculty of Science, Menoufia University,
Shebin El-Koom 32511, Egypt

123
3690 X.-H. Song et al.

problem studied by experts and researchers. The methods for protecting image can
mainly be divided into two categories, one is image cryptosystem and the other is
image steganography. Image cryptosystem is transforming image from an intelligi-
ble form into an unidentifiable form to protect it during the transmission. Image
steganography embeds the image into some media invisibly by hiding the commu-
nication activity to protect message. Besides, image watermarking, as one of the
most important information hiding techniques, aims to protect image copyright via
embedding the watermark signal about the copyright information into the images.
For images in conventional computers, there have been many image watermarking
fruits.
Recently, quantum computation has been applied in many fields of computer sci-
ence. The rapid development of quantum computation and quantum computer raises
people’s interest to research quantum data security. As with classical data on classical
computers, these quantum data such as images will be susceptible to different kinds of
abuses. During the last years, digital image watermarking has found enormous success
as a method for protecting copyright and description of material on classical comput-
ers. In order to guard against their abuse, extending similar techniques to quantum
images appears necessary.
Until the arrival of practical quantum computers, the first stage in this direction
is building the proposal for capturing and storing the images on quantum comput-
ers. Quantum images are representation patterns on quantum computers [1]. Some
methods for representing quantum images have been proposed. The quantum image
is represented by color, a quantum state detected from monochromatic electromag-
netic waves through special machines and position, the storing unit was named Qubit
Lattice [2]. Latorre brought out a novel method which mapped pixels into the Real
Ket of the Hilbert space to complete image compression combined with pixel states
[3]. A flexible representation of quantum image (FRQI) [1] encoded the image color
and position into one quantum state which keeps the classical properties of color and
position. Based on FRQI, a multi-channel representation for quantum images using
RGBα color space is proposed in [4]. Moreover, Zhou et al changed the gray-scale
of the quantum image in FRQI into the qubit binary string when they designed the
quantum image encryption algorithm [5].
With the study of the representations for quantum images, people begin to research
quantum image watermarking strategies. Based on FRQI, Iliyasu et al proposed a
quantum images watermarking and authentication (WaQI) algorithm by means of
restricted geometric transformations [6,7]. The strategy is secure, keyless, blind and
computationally efficient, and can be perfectly used to authenticate the identity of the
owner. However, only under the condition that we know the content of watermark
image, we could know the quantum circuit used to authenticate the identity of the
carrier images owner. It is hereby can only be used to authenticate whether the carrier
image belongs to a certain one. In [8], a watermarking scheme for quantum images
based on QFT and FRQI is proposed. The watermark strategy can find the owner
of the carrier image actively (extract the watermark image without knowing what it
is). But the whole computation process cannot keep the image a normalized quantum
state, because their embedding strength controlling parameter is a fixed parameter.

123
A dynamic watermarking scheme for quantum images 3691

Some analysis and improvement about the watermarking scheme using QFT has been
proposed in [9].
In this paper, a novel dynamic watermarking scheme for quantum images using
quantum wavelet transform is proposed. The used wavelet is the fast algorithm and
complete circuit of quantum Daubechies D (4) wavelet transform [10]. The proposed
scheme utilizes a dynamic vector to control embedding strength instead of a fixed
parameter for the embedding process in other schemes. The matrix is an optimal
solution of an optimization equation which is built to make the watermarked image
having better visual quality. Simulations conducted show that the proposed scheme
has better visual quality under a bigger capacity compared to other schemes.
The outline of this work is organized as follows. Section 2 gives a brief background
of quantum computation, the FRQIs and quantum Daubechies D (4) wavelet transform.
FRQI and QWT are the bases of the proposed quantum watermarking scheme. Section
3 introduces the proposed dynamic watermarking scheme. Section 4 is devoted to the
analysis and results. Finally, in Sect. 5, the conclusion is drawn.

2 Background on quantum computation, representation of quantum images


and quantum wavelet transform

In this section, we introduce an overview of quantum computation, especially the


quantum network for adder operation, the flexible representation of quantum images
[1] and quantum Daubechies wavelet transform [10] as the basis for the proposed
scheme.

2.1 Background on quantum computation

A quantum computer is a physical machine that can accept input states which represent
a coherent superposition of many different possible inputs and subsequently evolve
them into a corresponding superposition of outputs. Computation, i.e. a sequence of
unitary transformations, affects simultaneously each element of the superposition,
generating a massive parallel data processing albeit within one piece of quantum
hardware [12]. A quantum bit, or qubit, is a unit vector in a two-dimensional com-
plex vector space for which a particular basis, denoted by |0, |1, has been fixed. A
quantum circuit provides a visual representation of how a complicated multi-qubit
quantum computation can be decomposed into a sequence of simpler, usually 1-qubit
and 2-qubit, quantum gates. In general, a given unitary matrix, which specifies some
desired quantum computation, will admit many different, but equivalent, decompo-
sitions depending on the set of primitive quantum gates used, and the skill of the
quantum circuit designer in composing those gates in an intelligent way. Some nota-
tions for the elementary quantum gates and their corresponding matrices are shown
in Fig. 1.
In our watermarking scheme given below, quantum networks for arithmetic opera-
tion should be designed. As we all known, for classical computation, it is easy to imple-
ment arithmetic operations. However, any unitary operation in quantum computer is
inverse. That is why quantum networks effecting elementary arithmetic operations

123
3692 X.-H. Song et al.

Fig. 1 Notations and corresponding matrices for basic gates

Fig. 2 Basic sum and carry operations for plain addition network. a The sum operation. b The carry
operation (from [13])

such as addition, multiplication, and exponentiation cannot be directly deduced from


their classical Boolean counterparts.
An explicit construction of several elementary quantum networks, i.e. plain adder,
adder modulo N , controlled-multiplier modulo N and exponentiation modulo N is
designed in [13]. A quantum network is a quantum computing device consisting of
quantum logic gates whose computational steps are synchronized in time. The out-
puts of some of the gates are connected by wires to the inputs of others. Inputs are
encoded in binary form in the computational basis of selected qubits often called a
quantum r egister , or simply a r egister . The addition of two registers |a and |b
can be written as

|a, b → |a, a + b. (1)

To prevent overflows, the second register (initially loaded in state |b) should be
sufficiently large, i.e. if both a and b are encoded on n qubits, the second register should
be of size n + 1. In addition, the network described here also requires a temporary
register of size n − 1, initially in state |0, to which the carries of the addition are
provisionally written. The basic subnetworks of carry and sum operations for plain
addition network are demonstrated in Fig. 2. And the operation of the full addition
network is illustrated in Fig. 3. If we reverse the action of the above network (i.e. if we
apply each gate of the network in the reversed order) with the input (a, b), the output
will produce (a, a − b) when a ≥ b. when a < b, the output is (a, 2n+1 − (b − a)),
where n + 1 is the size of the second register.

2.2 Flexible representation of quantum image

Inspired by the pixel representation for images in classical computers, a flexible rep-
resentation for quantum images on quantum computers is proposed in [1], in which,

123
A dynamic watermarking scheme for quantum images 3693

Fig. 3 Plain adder network. Note the position of a thick black bar on the right- or left-hand side of basic
carry and sum networks. A network with a bar on the left side represents the reversed sequence of elementary
gates embedded in the same network with the bar on the right side (from [13])

the quantum image corresponding to a classical image sized 2n × 2n is defined by a


quantum encoding state about image color and position, i.e.

2 −12n
1 
|I (θ ) = n (cos θi |0 + sin θi |1) ⊗ |i,
2
i=0
 π
θi ∈ 0, , i = 0, 1, . . . , 22n − 1, (2)
2
wherein, cos θi |0 + sin θi |1 encodes the color information and |i encodes about
the corresponding position of the quantum images. The position information includes
two parts: the vertical and horizontal coordinates. Considering a quantum image in
2n-qubits system,

|i = |y|x = |yn−1 yn−2 . . . y0 |xn−1 xn−2 . . . x0 ,


x, y ∈ {0, 1, . . . , 2n − 1}, (3)
|y j , |x j  ∈ {|0, |1}, j = 0, 1, . . . , n − 1,

here |y = |yn−1 yn−2 . . . y0  encodes the first n-qubits along the vertical location
and |x = |xn−1 xn−2 . . . x0  encodes the second n-qubits along the horizontal axis.
Besides, the FRQI state is a normalized state, i.e.

22n −1
1 
I (θ ) = n  (cos2 θi + sin2 θi ) = 1. (4)
2
i=0

Furthermore, the authors proposed a polynomial preparation theorem that proves the
existence of a unitary preparation process which can use a polynomial number of
simple operators to transform quantum computers from the initial state to FRQI state.

123
3694 X.-H. Song et al.

2.3 Quantum wavelet transform

In classical computation, wavelet transform, every bit of which is useful as Fourier


transform, researches the multi-scales structure of signal and is important in image
compression and processing. In [10], efficient and complete quantum circuits for
Haar and Daubechies D (4) wavelets are proposed. Specially, the quantum circuit of
Daubechies D (4) wavelet can be implemented by means of QFT circuit designed in
[11].
The essence of designing a quantum circuit is factorizing the transform matrix into
elementary unitary matrices. Therefore, to achieve a feasible and efficient quantum
(4)
implementation, a suitable factorization of D2n need to be developed, where, n is the
size of qubits. The circuit of Daubechies fourth-order wavelet kernel with dimension
2n can be summarized briefly as follows [10].
Suppose two unitary matrices C0 and C1 as
 
c −c2 1 c0 /c3 1
C0 = 2 3 , C1 = , (5)
c2 c3 2 1 c1 /c2
√ √ √ √ √ √
0 = (3
wherein, c√ √ + 3)/4 2, c1 = (3 − 3)/4 2, c2 = (1 − 3)/4 2 and
(4)
c3 = (1+ 3)/4 2. The Daubechies fourth-order wavelet kernel D2n can be divided
into

D2(4)
n = (I2n−1 ⊗ C 1 )Q 2n (I2n−1 ⊗ C 0 ) (6)
 
c c3 01
where, C0
= N C0 = 2 2 , N= and Q 2n is the downshi f t per m-
c3 −c2 10
utation matri x given by
⎛ ⎞
0 1
⎜0 0 1 ⎟
⎜ ⎟
⎜0 0 0 1 ⎟
⎜ ⎟
Q 2n =⎜. .. .... ⎟. (7)
⎜ .. . . . ⎟
⎜ ⎟
⎝0 0 ··· 0 0 0⎠
1 0 ··· 0 0 0

Moreover, matrix Q 2n can be factored as Q 2n = F 2n P2n T2n P2n F ∗2n . Where F 2n is


the classical Cooley–Tukey FFT factorization for a 2n-dimensional vector and F ∗2n is
the conjugate transpose of F 2n [11], P2n is bit-reversal permutation matrix, A quantum
description of P2n is given as

P2n : |an−1 an−2 . . . a1 a0  → |a0 a1 . . . an−2 an−1 , (8)

and T2n is a diagonal matrix given as T2n = Diag{1, ω2n , ω22n , . . . , ω22n −1 }, ω2n =
n

2iπ √
e− 2n , i = −1.

123
A dynamic watermarking scheme for quantum images 3695

Fig. 4 Quantum circuit of Daubechies fourth-order wavelet kernel (from [10])

Finally, the product matrix P2n T2n P2n as a whole can be factorized into

n−1 n−i
P2n T2n P2n = (I2n−1 ⊗ G(w22n )). . .(I2n−i ⊗G(w22n )⊗ I2n−1 ) . . . (G(w2n ) ⊗ I2n−1 ),


1 0
wherein, G(ω2kn )
= Diag{1, ω2kn }
= , k = 0, 1, . . . , n − 1. Therefore,
0 ω2kn
the whole quantum circuit of Daubechies fourth-order wavelet kernel is depicted in
Fig. 4, which represents a complete gate-level implementation of D2(4) n .

3 Proposed scheme

In this paper, a new watermarking scheme is implemented for quantum images. In


the embedding algorithm, we design a dynamic embedding-strength-controlling vec-
tor α , which is confirmed by solving an optimization equation to ensure the best
image quality. The concrete procedures of the embedding and the extraction are given
hereafter.

3.1 Watermark image’s embedding procedure

The proposed watermarking embedding procedure is given by the following steps and
can be shown in Fig. 5.
1. Quantum images preparation Given a carrier quantum image sized 2n × 2n ,
whose FRQI is:

2 −1 2n
1 
|C = n (cos θi |0 + sin θi |1) ⊗ |i
2
i=0
2n −1
2
= |ci  ⊗ |i, (9)
i=0

123
3696 X.-H. Song et al.

Fig. 5 General framework for the embedding procedure of the proposed quantum image watermarking
scheme

and a watermark image sized 2n × 2n , whose FRQI is

2 −1 2n
1 
|W  = n (cos ϕ j |0 + sin ϕ j |1) ⊗ | j
2
j=0
22n −1

= |w j  ⊗ | j. (10)
j=0

2. Quantum wavelet transform Firstly, execute QWT on the carrier quantum image
|C, getting its wavelet coefficients QW T (|C) shown as follows

2 −1 2n
1 
QW T (|C) = n QW T ((cos θi |0 + sin θi |1) ⊗ |i)
2
i=0
2n −1
2
= |wci  ⊗ |i. (11)
i=0

3. Watermark processing The quantum watermark image |W  is processed by the


following unitary transforms,
⎛ 2n −1

2
Pi = ⎝ I ⊗ | j j|⎠ + Ph(αi ) ⊗ |i i|,
j=0, j =i
 
1 0 eiαi 0
I = , Ph(αi ) = , (12)
0 1 0 eiαi

where, I is a 2-D identity matrix and Ph(αi ) is a phase gate.


Therefore, applying Pk and Pl Pk to quantum watermark image |W  gives us

123
A dynamic watermarking scheme for quantum images 3697

⎛2n −1

2
1
Pk (|W ) = Pk ⎝ n (cos ϕ j |0 + sin ϕ j |1) ⊗ | j⎠
2
j=0
⎡ 2n
2 −1
1
= n⎣ (cos ϕ j |0 + sin ϕ j |1) ⊗ | j
2
j=0, j =k

+eiαk (cos ϕk |0 + sin ϕk |1) ⊗ |k⎦ , (13)

Pl Pk (|W ) = Pl (Pk (|W ))


⎡ 2n
2 −1
1 ⎣ 
= n (cos ϕ j |0 + sin ϕ j |1) ⊗ | j
2
j=0, j =k,l

+e iαk
(cos ϕk |0 + sin ϕk |1) ⊗ |k

+eiαl (cos ϕl |0 + sin ϕl |1) ⊗ |l⎦ . (14)

From (14), it is clear that


⎛ ⎞
22n
 −1
P(|W ) = ⎝ P j ⎠ (|W )
j=0
2 −1 2n
1  iα j
= n e (cosϕ j |0 + sin ϕ j |1) ⊗ | j
2
j=0
= |P W . (15)

α = (α0 , α1 , . . . , α22n −1 ) is a dynamic embedding strength controlling vector instead


of a fixed parameter as in [8].
4. Embedding watermark The processed quantum watermark image |P W  is
embedded into the wavelet coefficients QW T (|C) according to the following model

QW T (|C W ) = QW T (|C) + |P W 
2n −1
2
= |wmci  ⊗ |i. (16)
i=0

QW T (|C) represents the QWT qubits of the carrier image, QW T (|C W ) repre-
sents the qubits after embedding watermark.
Quantum image is the representation pattern of image in quantum computer, so
the whole implementation process about it must satisfy the principal of quantum
mechanics. A quantum network of the embedding process on a quantum image sized
of 2 × 2 is demonstrated in Fig. 6 and can be understood as follows. At first, two

123
3698 X.-H. Song et al.

Fig. 6 Quantum network of quantum image watermark embedding

registers are needed to encode the carrier image and the watermark image. Then,
QWT operation is executed on the quantum carrier image |C and a sequence of phase
gates Ph(α) = P decided by embedding strength α is implemented on the watermark
image. Finally, the adder quantum network shown in Fig. 3 is implemented on the
carrier image and the watermark image.
Note: When we extract the watermark image, we should execute inverse QWT to
QW T (|C W ), that is

in QW T {QW T (|C W )} = in QW T {QW T (|C) + |P W }


= |C + in QW T {|P W }. (17)

where, in QW T is the inverse QWT operator. If we want to make the watermarked


quantum image more similar to the original carrier quantum image |C, we hope the
part in QW T {|P W } have a smaller magnitude. Therefore, the vector α should satisfy
the following programming equation, which is constituted by a target function, aims to
make the visual quality of the carrier quantum image change less and a constraint con-
dition, to make the QWT coefficients still a normalized quantum state after embedding
watermark,

min in QW T {|P W }


s.t. QW T (|C) + |P W  = 1, (18)

where, in QW T is the inverse QWT transform, the solution α is obtained by solving


the optimization problem which is actually a quadratic equation with 22n unknowns
under an equality constrain. The optimization problem here is solved by genetic algo-
rithm.Genetic algorithm is a random and fast method for solving large programming
equation. To simplify the problem, the elements of α is divided into k groups, k is
an even. Through selection, crossover and inheritance steps, the optimal solution of
α is obtained. The dynamic parameter α depends on both the carrier image and the
watermark image, thus, it can be considered as the key for recovering the watermark.

123
A dynamic watermarking scheme for quantum images 3699

Fig. 7 General framework for the extraction procedure of the proposed quantum image watermarking
scheme

5. Inverse quantum wavelet transform Executing the inverse QWT on QW T (|C W ),
we get the watermarked quantum image |C W .

3.2 Watermark image’s extraction procedure

At the extraction phase, we use the original carrier image and the same key vector α to
extract the watermark image. Because all the transforms used in quantum computation
are unitary described by unitary matrices, so the embedding procedure is completely
reversible. The main extraction procedure is shown in Fig. 7.

1. Quantum wavelet transform Execute the QWT on the watermarked quantum


image, we obtain the quantum state QW T (|C W ).
2. Extracting watermark Through Eqs. 15 and 16, we have

|W  = P −1 (QW T (|C W ) − QW T (|C)), (19)

where, P −1 is the inverse transform of the controlled phase P. Thus, we can


recover the watermark image |W .

4 Analysis and results

In this section, we introduce several analysis in terms of visual quality, computational


complexity, and payload capacity.
Visual quality : This is the difference between the carrier image and its watermarked
version. PSNR is a well known measure of watermarked image quality as defined in
Eq. 20.

M AX I
P S N R = 20 log10 √ , (20)
MSE

where (MSE), Eq. 21, is the mean squared error for two m × n monochrome images,
I and its watermarked version K is defined as

1 
m−1 n−1
MSE = [I (i, j) − K (i, j)]2 . (21)
mn
i=0 j=0

here, M AX I is the maximum possible pixel value of the image.

123
3700 X.-H. Song et al.

Fig. 8 A part of images used in simulations

Complexity : The complexities of the common classical transforms and their quan-
tum counterparts will analyzed to show how the quantum transforms are more efficient
than their classical counterpart.

123
A dynamic watermarking scheme for quantum images 3701

Table 1 The PSNR Values of


Carrier images Watermark images PSNR (dB)
watermarked images
Lena Baboon 77.05
Lena Cameraman 77.69
Lena Pills 74.08
Cameraman Lena 80.84
Cameraman Baboon 75.89
Cameraman Pills 74.76
Baboon Lena 77.95
Baboon Cameraman 75.39
Baboon Pills 76.13
Pills Lena 77.68
Pills Baboon 75.16
Pills Cameraman 75.03
Einstein Jetplane 77.84
Einstein baboon 78.43
Einstein Cameraman 76.52
Jetplane Einstein 76.09
Jetplane Lena 79.22
Jetplane Cameraman 81.77

Payload capacity : A good image hiding scheme can embed a large quantity of secret
data within a carrier image [17]. We compute the maximum capacity of the proposed
scheme and compare it with other schemes.

4.1 Visual quality

Herein, we make the simulations of the input quantum carrier images and watermark
images on a classical computer due to the condition that the physical quantum com-
puter is not in our grasp right now. The simulations are demonstrated with a classical
computer with Intel(R) Xeon(R) CPU E5620 2.40 GHz, 2.39 GHz(4 processors) 8 GB
RAM and 64-bit operating system. The simulations are based on linear algebra with
complex vectors as quantum states and unitary matrices as unitary transforms using
Matlab 7.12.0(R2011a). The final step in quantum computation is the measurement,
which converts the quantum information into the classical form as probability distri-
butions. Analyzing the distribution gives information for revealing structures in these
images.
Because of the limit of classical computer, we use images of size 64×64 to simulate
our algorithm. A part of carrier quantum images are the resized images of standard
512 × 512 Lena, Baboon, Cameraman and Pills, which are shown in Fig. 8. When
solving the programming equation using genetic algorithm, we set the groups number
k = 8.

123
3702 X.-H. Song et al.

Fig. 9 The upperle f t is the original carrier image, the upperright is the watermark image, the le f tdown
is the watermarked image and the rightdown is the extracted watermark image

The PSNR values of some watermarked images and the carrier images are shown
in Table 1. From Table 1, we can conclude that the smallest PSNR in our scheme,
74.08 dB, is obviously higher than the PSNRs in [7,8], the biggest PSNRs in [7] and
[8] are 45.3864 and 64.82 dB, respectively. Moreover, the watermarking capacity in
our scheme can reach the size of the carrier image, which is four times higher than
the capacity in [8], which is half size of the carrier image. Because of the unitary
property of quantum computation, the extracted watermark image is completely same
as the embedded image. Parts of experimental results are shown in Figs. 9 and 10.
The quantum image state has 8,192 complex numbers as its coefficients, which decide
the probability distribution of quantum image state. Furthermore, the difference of
the probability distributions reflects images’ similarity. Comparison about the prob-
abilities of the first 100 computational basis between the original carrier quantum
image and the watermarked quantum image in Figs. 9 and 10 is shown in Figs. 11 and
12, respectively. Through comparison about the two scatter diagrams, we conclude
that most of the probabilities amplitudes between the original carrier image and the
watermarked image change imperceptibly.

123
A dynamic watermarking scheme for quantum images 3703

Fig. 10 The upperle f t is the original carrier image, the upperright is the watermark image, the le f tdown
is the watermarked image and the rightdown is the extracted watermark image

4.2 Complexity

A comparison of classical and quantum transforms complexities including Fourier


transform (FT), cosine transform (CT), wavelet transform (WT), Walsh–Hadamard
transform (WHT) and fractional Walsh transform (FWT) is shown in Table 2. Here,
the size of the vector is supposed to N = 2n and the complexities of the quantum
transforms are computed by the numbers of the used basic quantum gates. From
Table 2, we can see that the quantum transforms have lower computational complexity
over their classical counterparts.
In our watermarking scheme, the number of the basic quantum gates needed in the
whole watermarking process for 2n × 2n sized image is as follows: preparing quantum
carrier image and watermark image needs 2O(24n ) simple quantum gates, QWT needs
O(4n 2 ), watermark processing needs O(2n) and finally embedding watermarking step
includes 4n + 2 CARRY operations and 2n + 1 SUM operations, the CARRY needs
3 basic gates and SUM needs 2, so the total gates of the embedding phase is

123
3704 X.-H. Song et al.

−3
x 10
9.5

Probability 9

8.5

7.5

7
0 10 20 30 40 50 60 70 80 90 100
The first 100 Computational Basis

−3
x 10
9.5

9
Probability

8.5

7.5

7
0 10 20 30 40 50 60 70 80 90 100
The first 100 Computational Basis

Fig. 11 The probabilities of the first 100 computational bases. Upper The original Lena. Lower The
watermarked Lena after embedding Cameraman

0.015
Probability

0.01

0.005
0 10 20 30 40 50 60 70 80 90 100

The first 100 Computational Basis


0.015
Probability

0.01

0.005
0 10 20 30 40 50 60 70 80 90 100
The first 100 Computational Basis

Fig. 12 The probabilities of the first 100 computational bases. Upper The original Baboon, Lower The
watermarked Baboon after embedding Pills

123
A dynamic watermarking scheme for quantum images 3705

Table 2 Complexities of
Classical transforms Complexity Quantum transforms Complexity
quantum and classical
transforms [10,11,14–16]
FFT O(n2n ) QFT O(n 2 )
(4) (4)
DWT (D2n ) O(2n ) QWT (D2n ) O(n 2 )
DCT O(n22n ) QCT O(2n )
WHT O(2n ) QWHT O(1)
DFWT O(3n) QFWT O(n)

Table 3 Comparison of the


Method Watermark capacity (Bits/pixel)
watermarking capacity for
256 × 256 Lena image
Zhang et al. [17] 1.3
Yaghmaee et al. [18] 0.4–1.12
WaQI [7] 0.5–1.75
Zhang et al. [8] 2
Proposed scheme 8

2 · 24n + 4n 2 + 2n + 3(4n + 2) + 2(2n + 1) = 24n+1 + 4n 2 + 18n + 8

Here, we neglect the complexity of the preparation, FRQI, that is O(24n ). Thus,
the complexity of the watermarking scheme is O(4n 2 ), which indicates the efficiency
of the proposed watermarking scheme.

4.3 Payload capacity

In our scheme, the maximum payload can be same as the size of the carrier image. If
we convert the payload to classical parameter, the capacity of the proposed scheme is
8 bits/pixel, which is far larger than most of the classical and quantum watermarking
schemes. Table 3 shows the comparison between watermark capacity of the pro-
posed scheme and some other important watermarking methods for 256 × 256 Lena
image.
From the comparison in Table 3, it is demonstrated that the proposed scheme out-
performs the common classical watermarking schemes and the existing quantum ones.

5 Conclusion

In this paper, we have proposed a new dynamic watermarking scheme for quantum
images using QWT and FRQI, which can find the owner of the carrier image actively. In
the new scheme, we utilized a dynamic diagonal vector for controlling the embedding
strength instead of a fixed parameter used in other schemes for embedding process.
The dynamic embedding vector is decided by both the carrier quantum image and the
watermark image, which is only known by the authorized owner, so it can be considered
as the key for the extraction phase. We have shown how to implement the embedding
and the extraction processes by the principles of quantum mechanics. Results show

123
3706 X.-H. Song et al.

that the proposed scheme has better quality, low computational complexity and higher
capacity compared to other schemes proposed recently.

Acknowledgments This work is supported by the National Natural Science Foundation of China
(61100187, 11201100), the Fundamental Research Funds for the Central Universities (Grant No. HIT.
NSRIF. 2010046), Heilongjiang Province Educational Department Funds of China (12521107) and the
Youth Foundation at the Harbin University of Science and Technology (2011YF009).

References

1. Le, P.Q., Dong, F., Hirota, K.: A flexible representation of quantum images for polynomial preparation,
image compression and processing operations. Quantum Inf. Process. 10(1), 63–84 (2010)
2. Venegas-Andraca, S.E., Bose, S.: Storing, processing and retrieving an image using quantum mechan-
ics. Proc. SPIE Conf. Quantum Inf. Comput. 5105, 137–147 (2003)
3. Latorre, J.I.: Image compression and entanglement. arXiv:quant-ph/0510031 (2005).
4. Sun, B., Le, P.Q., Iliyasu, A.M., Adrian Garcia, J., Yan, F., Dong, J.F., Hirota, K.: A multi-channel
representation for images on quantum computers using the RGBα color space. Proceedings of the
IEEE 7th International Symposium on Intelligent, Signal Processing, pp. 160–165 (2011).
5. Zhou, R.G., Wu, Q., Zhang, M.Q., Shen, C.Y.: Quantum image encryption and decryption algo-
rithms based on quantum image geometric transformations. Int. J. Theor. Phys. doi:10.1007/
s10773-012-1274-8 (2012)
6. Le, P.Q., Iliyasu, A.M., Dong, F., Hirota, K.: Fast geometric transformations on quantum images.
IAENG Int. J. Appl. Math. 40(3), 113–123 (2010)
7. Iliyasu, A.M., Phuc, Q.L., Dong, F., Hirota, K.: Watermarking and authentication of quantum images
based on restricted geometric transformations. Inf. Sci. 186, 126–149 (2011)
8. Zhang, W.W., Gao, F., Liu, B., Wen, Q.Y., Chen, H.: A watermark strategy for quantum images based
on quantum Fourier transform. Quantum Inf. Process. doi:10.1007/s11128-012-0423-6 (2012)
9. Yang, Y.G., Jia, X., Xu, P., Tian, J.: Analysis and improvement of the watermark strategy for quantum
images based on quantum Fourier transform. Quantum Inf. Process. 12(8), 2765–2769 (2013)
10. Fijany, A., Williams, C.P.: Quantum wavelet transform: fast algorithm and complete circuits. LNCS
1509, 10–33 (1999)
11. Barenco, A., Ekert, A., Suominen, K.-A., Torma, P.: Approximate quantum Fourier transform and
decoherence. Phys. Rev. A 54, 139 (1996)
12. Deutsch, D.: Quantum theory, the Church-Turing principle and the universal quantum computer. Proc.
R. Soc. Lond. A400, 97–117 (1985)
13. Vedral, V., Barenco, A., Ekert, A.: Quantum networks for elementary arithmetic operations. Phys. Rev.
A 54(1), 147–153 (1996)
14. Klappenecker, A., Rotteler, M.: Discrete cosine transforms on quantum computers. In: IEEE Proceed-
ings of the 2nd International Symposium on Image and Signal Processing and, Analysis, 464–468
(2001).
15. Pang, C.Y., Zhou, Z.W., Guo, G.C.: Quantum discrete cosine transform for image compression. arXiv,
preprint quant-ph/0601043 (2006).
16. Labunets, V., Rundblad, E.L., Egiazarian, K., Karen, Astola, J.: Fast classical and quantum fractional
Walsh transforms. In: IEEE Proceedings of the 2nd International Symposium on Image and Signal
Processing and, Analysis, 558–563 (2001).
17. Zhang, F., Zhang, X., Zhang, H.: Digital image watermarking capacity and detection error rate. Pattern
Recognit. Lett. 28, 1–10 (2008)
18. Yaghmaee, F., Jamzad, M.: Estimating watermarking capacity in gray scale images based on image
complexity. EURASIP J. Adv. Signal Process. doi:10.1155/2010/851920 (2010)

123

You might also like