Optics and Lasers in Engineering 124 (2020) 105836

Contents lists available at ScienceDirect

Optics and Lasers in Engineering

journal homepage: www.elsevier.com/locate/optlaseng

Image encryption scheme by combining the hyper-chaotic system with

quantum coding
Yuling Luo, Shunbin Tang, Junxiu Liu∗, Lvchen Cao, Senhui Qiu
School of Electronic Engineering, Guangxi Normal University, Guilin, China, 541004

a r t i c l e i n f o a b s t r a c t

Keywords: The application of quantum mechanisms in image processing has attracted wide attention. In this paper, an image
Image processing encryption scheme that is based on hyper-chaos and quantum coding is proposed. First, based on a novel enhanced
Quantum coding quantum representation (NEQR), the bit-level adjacent-exchange operation is implemented to pre-process the
Bit-level
plain-image, which has lower computational complexity compared to conventional bit-level switching operations.
NEQR
Then, the processed image is diffused by the quantified sequences where the self-adapting parameters are applied
Chen’s hyper-chaotic system
to quantify the sequences that are generated by Chen’s hyper-chaotic system. In addition, inspired by the quantum
coding structure, a novel scrambling method that is based on quantum location coding is proposed. This method
is used to further scramble the diffused image to obtain the final cipher-image. The experimental results and
performance analysis demonstrate that the proposed image encryption scheme has high security performance.

1. Introduction In addition, quantum image processing in the field of information se-

curity has also attracted increasingly attention due to the breakthroughs
Image data play an indispensable role in the field of information of quantum teleportation and quantum computing [18]. The fundamen-
transmission and collection. However, due to the strong correlation and tal ideology is to map the classical image types into the quantum do-
high redundancy in plain-images, thay are vulnerable to various types main and execute quantum computing. It is an effective and important
of attacks in the transmission process. Although some traditional en- tool that can meet the high real-time requirements for classical image
cryption algorithms such as the data encryption standard (DES) and processing [18,19]. In 1982, a calculation model based on the princi-
Rivest Shamir Adleman (RSA) have high security and mature verifica- ples of quantum physics was first presented by R. P. Feynman [20].
tion abilities, they are not very suitable for image encryption because Furthermore, the concept of quantum computing was generalized by D.
of the special characteristics of image data [1,2]. Therefore, a number Deutsch, and he pointed out that all information processing tasks can be
of new image encryption techniques have been proposed, among which performed by using the properties of quantum mechanics [21]. Specifi-
the chaos-based encryption schemes have attracted wide attention be- cally, a quantum integer factoring algorithm was designed to find the en-
cause of their good randomness, ergodicity and sensitivity to the ini- cryption keys of the RSA within the polynomial time by Shor [22]. Then,
tial conditions of chaotic systems [3,4]. In addition, a variety of im- a quantum search algorithm for disordered databases was proposed by
age encryption algorithms based on chaotic maps have been presented Grover. Compared with the classical search method, this algorithm sig-
in recent years [5–8]. Especially, the high-dimensional chaotic system- nificantly improves the search efficiency [23]. In addition, the paral-
based encryption schemes have attracted wider attention due to their lelization strategies of quantum reactive scattering codes were studied
larger key space and more complex dynamic characteristics than those from different aspects [24]. Peng et al. proposed a novel quantum in-
of low-dimensional chaotic systems. Specifically, some image encryp- spired feature fusion method based on von Neumann entropy [25]. To
tion schemes that use hyper-chaotic maps with genetic encoding [9– better protect users’ personal information, quantum digital signatures
12] have been designed. Moreover, image encryption schemes based for the access control of sensitive data have been developed [26]. Ex-
on compressive sensing (CS) and chaotic systems were devised as well cept for quantum computing algorithms, several quantum image repre-
[13,14]. Furthermore, several novel hyper-chaos-based image encryp- sentation methods have been applied to the image processing field as
tion methods combined with some physical phenomena have been de- well, such as the Qubit lattice [27], entangled image [28], flexible rep-
veloped [15–17]. resentation of quantum images (FRQI) [29], novel enhanced quantum

∗
Corresponding author.
E-mail address: [email protected] (J. Liu).

https://doi.org/10.1016/j.optlaseng.2019.105836
Received 28 April 2019; Received in revised form 15 August 2019; Accepted 15 August 2019
0143-8166/© 2019 Elsevier Ltd. All rights reserved.
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

representation of digital images (NEQR) [30], quantum representation The remaining chapters of the paper are organized as follows. In
for log-polar images (QUALPI) [31], and n-qubit normal arbitrary super- Section 2, the preliminaries of Chen’s hyper-chaotic system, the quan-
position state (NASS) [32]. Among these methods, the FRQI and NEQR tum states, the NEQR model, the quantum adjacent-exchange operation,
have wider applications to image processing. Specifically, for the FRQI, and the novel scrambling method based on quantum location coding are
the image location information is stored in the basis state, which appears described. The whole encryption scheme and the corresponding decryp-
as a 2n-qubit sequence, and the pixel values of an image are encoded tion process are detailed in Section 3. Section 4 reports the encryption
as the probability amplitude of one qubit state [29]. Moreover, for the experiment results, the performance analysis and comparison. The con-
NEQR, two entangled qubit sequences are used to express the values clusion is provided in Section 5.
and positions of all grey-scale pixels. Therefore, the information can be
accurately recovered by quantum measurements [30].
On the basis of the above quantum image representation models, 2. Preliminaries
some quantum image encryption schemes have been proposed. A colour
image encryption system based on quantum cryptography and double 2.1. Chen’s hyper-chaotic system
random-phase encoding (DRPE) was devised [33]. In [34], a quantum
image encryption scheme that applied the image correlation decomposi- To improve encryption performance, a Chen-based hyper-chaotic
tion, two quantum gates and the Hadamard transform was presented. In system [47] is used in this encryption scheme, which is expressed by
addition, one-dimensional quantum cellular automata was implemented
⎧
in the field of quantum grey-scale image encryption [35]. Moreover, a ⎪𝑥̇ = 𝑎(𝑦 − 𝑥)
quantum image encryption system that was based on XOR operations, ⎪𝑦̇ = 𝑑𝑥 − 𝑥𝑧 + 𝑐𝑦 − ℎ
⎨𝑧̇ = 𝑥𝑦 − 𝑏𝑧 , (1)
chaotic maps and the generalized affine transform was proposed [36]. A ⎪
FRQI-based quantum image encryption scheme was proposed in which ̇
⎪ℎ = 𝑥 + 𝑘
⎩
the Fibonacci transform and geometric transform are used to scramble a
plain-image, and then the image is encrypted by DRPE [37]. Moreover,
where x, y, z, and h are four real state variables and a, b, c, d, and k are
high-dimensional chaotic systems have been widely used in quantum
five control parameters of Chen’s hyper-chaotic maps. The whole system
image encryption [38–42].
is in a hyper-chaotic state and the system has two positive Lyapunov ex-
However, although quantum computing possesses strong computing
ponents when 𝑎 = 36, 𝑏 = 3, 𝑐 = 28, 𝑑 = −16, and 𝑘 ∈ [0.7, −0.7]. It results
power, classical computing is more adaptable to some problems, such
in a shorter prediction time and more complex dynamic characteristics
as recursive problems, and divide-and-conquer ideological problems. In
than the simple one-dimensional chaotic systems [15]. Therefore, this
the future, classical computing should coexist with quantum comput-
system is used to generate four random sequences based on the fourth
ing in information processing. In fact, several encryption schemes that
order Runge-Kutta method. Then, these four random sequences are ap-
combined quantum mechanism with classical computing have been pro-
plied to the encryption operations.
posed. For example, as an excellent key generator, quantum walks (QW)
were applied to an image encryption system [43], chaotic maps based
on quantum mechanism have been utilized, and some novel image en- 2.2. Quantum state
cryption algorithms based on quantum chaos were designed [44,45].
Moreover, an improved classical encryption method that uses a quan- Hilbert space is an infinite dimensional Euclidean space, and a quan-
tum key distribution was published [46]. tum state can be represented by a vector of Hilbert space [48,49]. Quan-
Inspired by these schemes, an image encryption scheme that com- tum states satisfy the principle of superposition states. Therefore, the
bines a hyper-chaotic system with quantum coding is proposed in this basic qubit can be expressed as
work. It combines the quantum image representation model with the ′ ′
classical encryption method. Specifically, the classical digital image is |𝜑⟩ = 𝑎 |0⟩ + 𝑏 |1⟩, (2)
represented by the quantum states based on the novel enhanced quan-
where |0⟩ = [1, 0]𝑇 and |1⟩ = [0, 1]𝑇 are the quantum ground states. The
tum representation (NEQR) model where the grey-scale values of the ′ ′
plural numbers of 𝑎 and 𝑏 are the probability amplitude in the states
quantum states are transformed using quantum adjacent-exchange op- ′ ′
of |0⟩ and |1⟩, respectively, and they satisfy |𝑎 |2 + |𝑏 |2 = 1. The prob-
erations. Then, the transformed image is diffused by the quantified se- ′
ability of the quantum superposition state |𝜑⟩ in |0⟩ is |𝑎 |2 , and the
quences, which are generated by quantifying the sequences from the ′
probability of the quantum superposition state |𝜑⟩ in |1⟩ state is |𝑏 |2 .
chaotic system using the self-adapting parameters. Furthermore, a new
Therefore, a n-qubit system can be expressed as
scrambling method based on quantum location coding is used to scram-
𝑛 −1
ble the diffused image to get the final cipher-image. The main contribu- 2∑
tions of this work are summarized as follows. |𝜇⟩ = 𝐶𝑖 |𝑖⟩, (3)
𝑖=0
(1) A bit-level adjacent-exchange operation is proposed to change the ∑𝑛
where i is a n-bit binary string, Ci is a plural number, and 2𝑖=0−1 (𝐶𝑖2 ) = 1.
grey-scale values of the plain-image based on the NEQR model,
According to Eq (3), 𝐶𝑖2 is the probability that the quantum state |𝜇⟩ col-
which is an extension of quantum computing. The computational
lapses into the quantum state |i⟩ using quantum measurement. There-
complexity is reduced compared to the conventional bit-level
fore, a n-qubit quantum sequence can store 2n n-bits of data at the same
adjacent-exchange operations.
time due to the superposition characteristic of qubits. Consequently, the
(2) A self-adapting parameter generation method is designed and ap-
quantum information system has a more powerful information storage
plied to the image encryption algorithm, which enhances the ca-
capacity compared with the classic computer system.
pabilities to resist differential attacks.
(3) A novel image scrambling method that is based on quantum loca-
tion coding is proposed. It is an extension of the quantum mecha- 2.3. A novel enhanced quantum representation (NEQR) for grey-scale
nism in classical image processing, and it can further disturb the images
diffused data.
(4) Simulation results and comprehensive performance analysis In the NEQR model of grey-scale images, the image manipulation
demonstrate the high security performance of the proposed en- can be easily performed and the computational complexity is lower than
cryption system. the flexible representation of quantum images (FRQI) [30]. A grey-scale
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

⨉
 
⨉

 
⨉

Fig. 1. An example image with the size of 2 × 2.

⨉
image I with the size of 2n × 2n can be represented as ⨉

|𝐼⟩ =
1 ∑ ∑
𝑛
2 −1 2 −1 𝑛

|𝑓 (𝑌 , 𝑋)⟩ ⊗ |𝑌 𝑋⟩
⨉
2𝑛 𝑌 =0 𝑋=0
2𝑛 −1 2𝑛 −1
, (4) ⨉
1 ∑ ∑ 0 𝑞−2 𝑞−1
= |𝐶 … 𝐶 𝐶 ⟩ ⊗ |𝑌 𝑋⟩
2𝑛 𝑌 =0 𝑋=0 𝑌 𝑋 𝑌𝑋 𝑌𝑋
⨉
where ⊗ is the tensor product, 𝐶𝑌0 𝑋 𝐶𝑌1 𝑋 … 𝐶𝑌𝑞−2 𝐶 𝑞−1 ∈ [0, 1] is the binary
𝑋 𝑌𝑋
sequences, and the pixel values of the grey-scale image range from 0 to
2q . We take an image with the size of 2 × 2 as an example to detail the
NEQR model, which is shown in Fig. 1. The pixel values are in the range Fig. 2. The corresponding quantum circuit of S1 .
of [0, 255], and they are expressed as
1
|𝐼⟩ = (|0⟩ ⊗ |00⟩ + |100⟩ ⊗ |01⟩
2
+|200⟩ ⊗ |10⟩ + |255⟩ ⊗ |11⟩)
. (5)
1
= (|00000000⟩ ⊗ |00⟩ + |01100100⟩ ⊗ |01⟩
2
+|11001000⟩ ⊗ |11111111⟩ + |255⟩ ⊗ |11⟩)

In the NEQR model, the grey-scale information of all pixels is stored

in the base vector of the qubit sequence. Thus, the information can be ac-
curately recovered. To reconstruct the classical image from the quantum
image model, each pixel needs to be constructed separately. Through
quantum measurements [30], the classical grey-scale information can
be accurately extracted from the quantum image model.

2.4. Quantum grey-scale image adjacent-exchange operation

In this work, a quantum adjacent-exchange operation is designed to

change the grey-scale information of a quantum image. An image with Fig. 3. The corresponding quantum circuit of S2 .
the size of 2n × 2n is taken as an example. Here, the quantum adjacent- This exchange sub-operation can achieve the adjacent exchange
exchange operation can be achieved using adjacent-swap gate SYX . It of binary bits for the grey-scale values of quantum images. The sub-
consists of S1 and S2 , which are denoted by operation is used to change quantum image |I⟩ as
𝑛 𝑛
1 ∑ ∑ 0
2 −1 2 −1
𝑆1 = 𝑆𝑊 𝐴𝑃 ⊗4 , (6) 𝐴𝑌 𝑋 |𝐼⟩ = 𝐴𝑌 𝑋 ( |𝐶 … 𝐶𝑦𝑥
6
𝐶𝑦𝑥
7
⟩|𝑦𝑥⟩)
2𝑛 𝑦=0 𝑥=0 𝑦𝑥
𝑆2 = 𝐸 ⊗ 𝑆𝑊 𝐴𝑃 ⊗3 ⊗ 𝐸, (7)
𝑛 𝑛
1 ∑ ∑ 0
2 −1 2 −1
where SWAP is the quantum swap gate, E is the unit matrix, the grey- = ( |𝐶 … 𝐶𝑦𝑥
6
𝐶𝑦𝑥
7
⟩|𝑦𝑥⟩
2𝑛 𝑦=0 𝑥=0 𝑦𝑥 𝑦𝑥≠𝑌
scale range is from 0 to 255, and 𝑆 𝑊 𝐴𝑃 ⊗3 = 𝑆 𝑊 𝐴𝑃 ⊗ 𝑆 𝑊 𝐴𝑃 ⊗ 𝑋
, (10)
𝑆𝑊 𝐴𝑃 . The pixel values of f(Y, X) are changed using the adjacent-swap +𝐴𝑌 𝑋 |𝐶𝑌0 𝑋 … 𝐶𝑌6 𝑋 𝐶𝑌7 𝑋 ⟩|𝑌 𝑋⟩)
gate SYX , and the details are as follows. 𝑛 𝑛
1 ∑ ∑ 0
2 −1 2 −1
= ( |𝐶 … 𝐶𝑦𝑥
6
𝐶𝑦𝑥
7
⟩|𝑦𝑥⟩
𝑔(𝑌 , 𝑋) = 𝑆1 𝑆2 |𝑓 (𝑌 , 𝑋)⟩ = 𝑆𝑌 𝑋 |𝑓 (𝑌 , 𝑋)⟩ 2 𝑦=0 𝑥=0 𝑦𝑥 𝑦𝑥≠𝑌
𝑛
𝑋
. (8)
= |𝐶𝑌1 𝑋 𝐶𝑌3 𝑋 𝐶𝑌0 𝑋 𝐶𝑌5 𝑋 𝐶𝑌2 𝑋 𝐶𝑌7 𝑋 𝐶𝑌4 𝑋 𝐶𝑌6 𝑋 ⟩ +|𝐶𝑌1 𝑋 𝐶𝑌3 𝑋 𝐶𝑌0 𝑋 𝐶𝑌5 𝑋 𝐶𝑌2 𝑋 𝐶𝑌7 𝑋 𝐶𝑌4 𝑋 𝐶𝑌6 𝑋 ⟩|𝑌 𝑋⟩)

Figs. 2 and 3 are the corresponding quantum circuit of S1 and S2 . and

𝑛 𝑛
1 ∑ ∑ 0
2 −1 2 −1
Then, the quantum exchange sub-operation AYX is constructed using
𝐴𝑌1 𝑋1 𝐴𝑌 𝑋 |𝐼⟩ = 𝐴𝑌1 𝑋1 ( |𝐶 … 𝐶𝑦𝑥
6
𝐶𝑦𝑥
7
⟩|𝑦𝑥⟩
the quantum adjacent-swap gate SYX . 2 𝑦=0 𝑥=0 𝑦𝑥 𝑦𝑥≠𝑌
𝑛
𝑋
𝑛 −1 2𝑛 −1
2∑ ∑ +𝐴𝑌 𝑋 |𝐶𝑌0 𝑋 … 𝐶𝑌6 𝑋 𝐶𝑌7 𝑋 ⟩|𝑌 𝑋⟩)
𝐴𝑌 𝑋 = ( |𝑦𝑥⟩⟨𝑦𝑥| ⊗𝐼) + |𝑌 𝑋⟩⟨𝑌 𝑋| ⊗ 𝑆𝑌 𝑋 , (9) 𝑛 𝑛
1 ∑ ∑ 0 .
2 −1 2 −1
𝑦𝑥≠𝑌 𝑋 (11)
𝑦=0 𝑥=0
= ( |𝐶 … 𝐶𝑦𝑥
6
𝐶𝑦𝑥
7
⟩|𝑦𝑥⟩
2𝑛 𝑦=0 𝑥=0 𝑦𝑥 𝑦𝑥≠𝑌 𝑋,𝑌 𝑋
where the exchange sub-operation AYX is a unitary matrix, which satis- 1 1

fies 𝐴𝑌 𝑋 𝐴+ = 𝐸, and 𝐴+ is the Hermitian conjugate of unitary matrix +|𝐶𝑌1 𝑋 𝐶𝑌3 𝑋 𝐶𝑌0 𝑋 𝐶𝑌5 𝑋 𝐶𝑌2 𝑋 𝐶𝑌7 𝑋 𝐶𝑌4 𝑋 𝐶𝑌6 𝑋 ⟩|𝑌 𝑋⟩
𝑌𝑋 𝑌𝑋
AYX . +|𝐶𝑌1 𝐶𝑌3 𝐶𝑌0 𝐶𝑌5 𝐶𝑌2 𝐶𝑌7 𝐶𝑌4 𝐶𝑌6 ⟩|𝑌1 𝑋1 ⟩)
1 𝑋1 1 𝑋1 1 𝑋1 1 𝑋1 1 𝑋1 1 𝑋1 1 𝑋1 1 𝑋1
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

In Eqs. (10) and (11), 𝐴𝑌1 𝑋1 is the exchange sub-operation for |f(Y1 , ′
Therefore, the new quantum state |𝑢𝑖 ⟩ can be described as
X1 )⟩. It is apparent that each exchange sub-operation can only change
′
the corresponding pixel. Therefore, the quantum image adjacent- |𝑢𝑖 ⟩ = 𝑅𝑙𝑖 𝑅𝑖 |𝑢𝑖 ⟩
exchange operation can be described as
= 𝑅𝑙𝑖 ((𝑐𝑜𝑠𝛽𝑖 cos(𝛼𝑖 ∕2) − 𝑠𝑖𝑛𝛽𝑖 sin(𝛼𝑖 ∕2))|0⟩
∏ 𝑛 −1 ∏2𝑛 −1
𝐴|𝐼⟩ = 2𝑌 =0 𝑋=0 𝐴𝑌 𝑋 |𝐼⟩ +(𝑐𝑜𝑠𝛽𝑖 sin(𝛼𝑖 ∕2) + 𝑠𝑖𝑛𝛽𝑖 cos(𝛼𝑖 ∕2))|1⟩)
𝑛 −1 2𝑛 −1
2∑
1 ∑ = 𝑅𝑙𝑖 (𝑐𝑜𝑠(𝛽𝑖 + 𝛼𝑖 ∕2)|0⟩ + 𝑠𝑖𝑛(𝛽𝑖 + 𝛼𝑖 ∕2)|1⟩) , (19)
= 𝑆𝑌 𝑋 |𝐶𝑌0 𝑋 … 𝐶𝑌6 𝑋 𝐶𝑌7 𝑋 ⟩|𝑌 𝑋⟩
2𝑛 𝑌 =0 𝑋=0 = 𝑅𝑙𝑖 (𝑐𝑜𝑠ℎ𝑖 |0⟩ + 𝑠𝑖𝑛ℎ𝑖 |1⟩)
𝑛 −1 2𝑛 −1
2∑
1 ∑ = 𝑐𝑜𝑠(𝛾𝑖 + ℎ𝑖 )|0⟩ + 𝑠𝑖𝑛(𝛾𝑖 + ℎ𝑖 )|1⟩
= |𝐶𝑌1 𝑋 𝐶𝑌3 𝑋 𝐶𝑌0 𝑋 𝐶𝑌5 𝑋 𝐶𝑌2 𝑋 𝐶𝑌7 𝑋 𝐶𝑌4 𝑋 𝐶𝑌6 𝑋 |𝑌 𝑋⟩. (12)
2𝑛 𝑌 =0 𝑋=0 = 𝑐𝑜𝑠𝜙𝑖 |0⟩ + 𝑠𝑖𝑛𝜙𝑖 |1⟩
𝑛 −1 2𝑛 −1
2∑
1 ∑
= |𝑔(𝑌 , 𝑋)⟩|𝑌 𝑋⟩ where 𝛼 i and 𝛾 i are the processed chaotic sequences from Eq. (24), and
2𝑛 𝑌 =0 𝑋=0 𝜙i is a new angle after the above unitary transform. Therefore, a permu-
= |𝐺⟩ tation matrix d is obtained as

According to the quantum grey-scale image adjacent-exchange op- 𝑑𝑖 = 𝑎𝑏𝑠(𝑟𝑜𝑢𝑛𝑑(𝐹 −1 (𝜙𝑖 ))). (20)
eration, the inverse operation can be described as ′
∏ 𝑛 −1 ∏2𝑛 −1 + Finally, 𝐼 is scrambled by the following exchange operations:
𝐴−1 |𝐺⟩ = 2𝑌 =0 𝑋=0 𝐴𝑌 𝑋 |𝐺⟩ ′ ′
𝑛 𝑛 𝐼 (𝑖) ⟺ 𝐼 (𝑑𝑖 ), (21)
1 ∑ ∑ 0
2 −1 2 −1
= 𝑛 |𝐶 … 𝐶𝑦𝑥 6
𝐶𝑦𝑥
7
⟩|𝑌 𝑋⟩
2 𝑌 =0 𝑋=0 𝑦𝑥 Here, 𝑖 ∈ [1, 2, … 𝑀 𝑁 ].
. (13)
2𝑛 −1 2𝑛 −1
1 ∑ ∑
= 𝑛 |𝑓 (𝑌 , 𝑋)⟩|𝑌 𝑋⟩ 3. Image encryption and decryption scheme
2 𝑌 =0 𝑋=0

= |𝐼⟩ 3.1. Image encryption scheme

2.5. Scrambling method based on quantum location coding Without the loss of generality, the size of the grey-scale image I is
supposed to be M × N, and the overall framework of the image encryp-
In this work, a new scrambling method based on quantum location tion scheme is shown in Fig. 4, which can be divided into the following
information coding is designed. First, the M × N plain-image I(x, y) is steps.
′
reshaped into a one-dimensional sequence 𝐼 as
• Step 1. The classical image I is represented as the quantum image |I⟩
𝑖 = (𝑦 − 1) × 𝑀 + 𝑥; 𝑥 ∈ [1, 𝑀], 𝑦 ∈ [1, 𝑁], (14)
according to the NEQR model. In addition, according to Section 2.4,
where 𝑖 ∈ (1, 2, … , 𝑀 𝑁 ). Eq. (14) can be understood as sampling im- the grey values of a quantum image are changed by a quantum
age from top to bottom and from left to right. Moreover, the inverse adjacent-exchange operation AYX , which is described as
operation can be described as
𝑀−1
∏ 𝑁−1
∏
{
𝑥 = 𝑖 − (𝑦 − 1) × 𝑀 |𝐼1 ⟩ = 𝐴|𝐼⟩ = 𝐴𝑌 𝑋 |𝐼⟩. (22)
, (15) 𝑌 =0 𝑋=0
𝑦 = 𝑐𝑒𝑖𝑙(𝑖∕𝑀)
• Step 2. The classical image I1 is extracted from the quantum mea-
where (x, y) represents the pixel coordinates of the grey-scale image, i
′ surement, and then I1 is reshaped into a one-dimensional sequence
can be regarded as the ith item in the row vector of 𝐼 (𝑀 𝑁 entries), and
I2 according to Eq. (14). Furthermore, the self-adapting parameters
ceil(A) returns the nearest integer that is greater than or equal to A.
can be obtained as
Second, the mapping function between the sequences and angle is
{
established, as shown in Eq. (16), and it is used to construct the corre- 𝑠0 = 𝑠𝑢𝑚(𝐼2 )
sponding quantum state. , (23)
𝑠𝑎𝑑 𝑗 𝑢𝑠𝑡 = 𝑚𝑜𝑑(𝑠0 , 256)
𝐹 ∶ 𝑁𝑢𝑚𝑏𝑒𝑟 ⇌ 𝐴𝑛𝑔𝑙𝑒, (16)
where mod(s0 , 256) returns the modulus after the division of s0 by
where 𝑁𝑢𝑚𝑏𝑒𝑟 = (1, 2, … , 𝑖), and 𝐴𝑛𝑔𝑙𝑒 = (𝛽1 , 𝛽2 , … 𝛽𝑖 ) are or- 256.
dered sequences. 𝛽𝑖 = (𝜋𝑖)∕𝑀 𝑁 , 𝐹 (𝑖) = 𝛽𝑖 , and 𝐹 −1 (𝛽𝑖 ) = 𝑖, where
𝑖 ∈ (1, 2, … , 𝑀 𝑁 ). Moreover, Number is expressed as a form of a • Step 3. The initial values of a chaotic system and its control parame-
quantum state (|𝑢1 ⟩|𝑢2 ⟩, … |𝑢𝑖 ⟩), which is ters are set. After that, according to Eq. (1), four chaotic sequences
|𝑢𝑖 ⟩ = 𝑐𝑜𝑠𝛽𝑖 |0⟩ + 𝑠𝑖𝑛𝛽𝑖 |1⟩. (17) can be obtained based on iterative computations. Moreover, these
chaotic sequences are quantified as
′
Then, the relationship between Number and the coordinate of 𝐼 is
established. Here, the quantum state |u1 ⟩ denotes the first pixel coordi- ⎧𝑥 = 𝑓 𝑙𝑜𝑜𝑟(𝑚𝑜𝑑(𝑥 × 1014 , 256))
⎪ 1
nate (1,1), and the quantum state |ui ⟩ denotes the ith pixel coordinate ⎪𝑥 = 𝑓 𝑙𝑜𝑜𝑟(𝑚𝑜𝑑(𝑦 × 1014 , 256))
(𝑖 − (𝑐𝑒𝑖𝑙(𝑖∕𝑀) − 1) × 𝑀, 𝑐𝑒𝑖𝑙(𝑖∕𝑀)). ⎪ 2
Afterwards, the quantum state is changed by applying two kinds of ⎨𝛼 = 𝑚𝑜𝑑((𝑧 × 1014 − 𝑓 𝑙𝑜𝑜𝑟(𝑧 × 1014 )) + 𝑠 , 𝜋 ) , (24)
⎪ 0
4
qubit rotation gates, which are described as ⎪ 𝜋
⎪𝛾 = 𝑚𝑜𝑑((ℎ × 1014 − 𝑓 𝑙𝑜𝑜𝑟(ℎ × 1014 )) + 𝑠0 , )
⎡𝑐𝑜𝑠 𝛼𝑖 𝛼𝑖 ⎤ [ ] ⎩ 4
−𝑠𝑖𝑛 𝑐𝑜𝑠𝛾𝑖 −𝑠𝑖𝑛𝛾𝑖
⎢ 2 2⎥
𝑅𝑖 = ⎢ 𝛼 𝛼𝑖 ⎥, 𝑅𝑙𝑖 = 𝑠𝑖𝑛𝛾 . (18) where floor(A) denotes rounds the elements of A to the nearest inte-
⎢ 𝑠𝑖𝑛 𝑖 cos ⎥ 𝑖 cos 𝛾𝑖 gers less than or equal to A.
⎣ 2 2 ⎦
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Fig. 4. Flowchart of the encryption and decryption process.

• Step 4. The sequences x1 and x2 are utilized to perform the XOR

operation with I2 . The process can be written as
(a) (b) (c)

⎧𝐼2 (1) = 𝑠𝑎𝑑 𝑗 𝑢𝑠𝑡 ⊕ 𝐼2 (1) ⊕ 𝑥1 (1)

⎪
⎨𝐼2 (𝑖 + 1) = 𝑚𝑜𝑑(𝐼2 (𝑖) + 𝐼2 (𝑖 + 1), 256) ⊕ 𝑥1 (𝑖) , (25)
⎪
⎩𝐼2 (𝑖 + 1) = 𝐼2 (𝑖 + 1) ⊕ 𝑥2 (𝑖)

where 𝑖 ∈ [1, … 𝑀𝑁 − 1], and ⊕ denotes the XOR operation.

•
(d) (e) (f)
Step 5. Scramble the generated matrix I2 in Step 4 using the scram-
bling algorithm that is described in Section 2.5, and the cipher-text
C can be obtained. Taking into account the security of transmis-
sion, the cipher-text C can be reshaped into M × N cipher-image by
Eq. (15), and then it can be expressed as quantum state |C⟩ using the
NEQR.

3.2. Image decryption process (g) (h) (i)

The decryption process can be described by the following steps.

• Step 1. The control parameters (a, b, c, d, and k) and initial values

of the chaotic system (x0 , y0 , z0 , and h0 ) are utilized to generate
the chaotic sequences. Then, the generated chaotic sequences are
quantified according to Step 3 in the encryption process. In addition,
the encrypted image was reshaped into one-dimensional cipher-text (j) (k) (l)
C by Eq. (14).

• Step 2. Based on the above steps, C is inversely scrambled to obtain

I2 using the permutation matrix from Section 2.5. After that, the
reverse XOR operation is performed on I2 as
{
𝐼2 (𝑖 + 1) = 𝐼2 (𝑖 + 1) ⊕ 𝑥2 (𝑖)
, (26)
𝐼2 (𝑖 + 1) = 𝑚𝑜𝑑(𝐼2 (𝑖 + 1) ⊕ 𝑥1 (𝑖) − 𝐼2 (𝑖), 256) Fig. 5. Experimental results: (a), (d), (g), and (j) are the plain-images; (b), (e),
(h), and (k) are the corresponding encrypted images; (c), (f), (i), and (l) are the
corresponding decrypted images with the correct keys.
𝐼2 (1) = 𝑠𝑎𝑑 𝑗 𝑢𝑠𝑡 ⊕ 𝐼2 (1) ⊕ 𝑥1 (1), (27)
From these decryption steps, it can be seen that the decryption pro-
where 𝑖 ∈ [𝑀 𝑁 − 1, 𝑀 𝑁 − 2, … , 1]. cess is the inverse process of the encryption method, which is the shown
in Fig. 4. It can be found that the decrypted image from the decryption
• Step 3. The new generated I2 is reshaped into I1 with the size process is consistent with the plain-image.
of M × N, and I1 is denoted using the NEQR model. Then, the
plain quantum image |I⟩ is obtained based on the inverse quantum 4. Numerical simulation and performance analysis
adjacent-exchange operation.
The numerical simulation is executed in MATLAB R2014a. Four
𝑀−1
∏ 𝑁−1
∏ 256 × 256 images are tested on a PC with Windows 10, 8.00 GB of RAM,
|𝐼⟩ = 𝐴 |𝐼1 ⟩ =
−1
𝐴+ |𝐼 ⟩.
𝑌𝑋 1
(28)
𝑌 =0 𝑋=0
and an Intel(R) Core(TM) i7 7700HQ 2.80 GHz processor. The initial
values of Chen’s hyper-chaotic system are 𝑥0 = 0.1314, 𝑦0 = 0.5214, 𝑧0 =
• Step 4. The plain-text I can be extracted from the quantum image |I⟩. 0.3689, and ℎ0 = 0.8417 and the control parameters of the chaotic system
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Fig. 6. Correlation distributions of “Lena”. (a), (c),

and (e): Correlation distributions of the plain-image
in the horizontal, vertical, and diagonal directions, re-
spectively; (b), (d), and (f): Correlation distributions
of the encrypted image in the horizontal, vertical, and
diagonal directions, respectively.

are 𝑎 = 36, 𝑏 = 3, 𝑐 = 28, 𝑑 = −16, and 𝑘 = 0.2. The corresponding exper- Table 1
imental results are shown in Fig. 5. Comparisons of correlation coefficients.

Image Horizontal Vertical Diagonal

Plain-image “Lena” 0.9417 0.9696 0.9149

4.1. Computational complexity of the quantum adjacent-exchange Cipher-image in [40] −0.0013 0.0190 −0.0073
operation Cipher-image in [36] −0.0072 0.0086 −0.0028
Cipher-image in [37] −0.0013 0.0147 0.0240
The size of the image is assumed to be 2n × 2n . The adjacent exchange Cipher-image in [50] −0.0226 0.0041 0.0368
Cipher-image in this work 0.0018 0.0040 −6.0096e-04
of binary bits is performed by using a quantum adjacent-exchange op-
eration. The complexity of the operations is mainly determined by SYX ,
and the Controlled-NOT (CNOT) gate is selected as the basic unit gate.
The adjacent-swap gate SYX includes seven swap gates, and each swap seen that the complexity of the proposed quantum adjacent-exchange
gate is composed of three CNOT gates [39]. Hence, the number of CNOT operation is lower than that of the classical computation.
gates in a quantum adjacent-exchange operation is 21. The characteris-
tics of quantum parallelism are considered, and the total computational 4.2. Statistical analysis
complexity is O(21n).
However, the classical digital image contains 22n pixels, and the com- 4.2.1. Correlation of adjacent pixels
plexity of completing a one bit exchange instruction is 3. If the image The “Lena” image with the size of 256 × 256 is taken as an example
adjacent-exchange operation is implemented using classical computing, for this test. The correlation between adjacent pixels is an important
the computational complexity requires O(21 × 22n ). Therefore, it can be index to measure whether this method can resist statistical attacks. To
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Table 2 Table 6
Correlation coefficients of different plain-images and corresponding cipher- The NPCR and UACI of the “Lena” image using
images. different methods.

Horizonal Vertical Diagonal Method NPCR(%) UACI(%)

Images
Plain Cipher Plain Cipher Plain Cipher Ref. [53] 99.4746 33.3774
Ref. [54] 99.6017 28.1370
“Lena” 0.9417 0.0018 0.9696 0.0040 0.9149 −6.0096e-4
Ref. [55] 99.6051 33.4000
“Peppers” 0.9670 0.0054 0.9733 0.0045 0.9395 −0.0038
Ref. [56] 99.5865 33.2533
“Baboon” 0.7145 -0.0019 0.6410 0.0021 0.6110 −0.0046
Proposed algorithm 99.6086 33.4507
“Clock” 0.9536 0.0072 0.9739 0.0011 0.9427 −0.0017

Table 3
Table 7
The variances of the different images.
The NPCR and UACI for different plain-images using the
Name of Variance proposed method.
cipher-images
Plain Cipher Images “Lena” “Peppers” “Baboon” “Clock”
“Lena” 30665.3828 222.9218 NPCR(%) 99.6086 99.6094 99.5850 99.6310
“Peppers” 51750.2734 244.4844 UACI(%) 33.4507 33.4736 33.5038 33.4203
“Baboon” 44006.7891 236.2656
“Clock” 282061.6328 214.7891

Four 256 × 256 images of “Lena”, “Peppers”, “Baboon” and “Clock” are
Table 4
tested, and the corresponding histograms of the plain-images and the
Information entropies of the original images and en-
crypted images.
corresponding encrypted images are shown in Fig. 7.
In addition, the distribution uniformity of a histogram can be mea-
Information entropy sured by the variance [51]. The smaller the variance is, the more uni-
Test cipher-images
Plain Cipher form the distribution will be. The variance is defined as

1 ∑
“Lena” 7.5683 7.9976 255
“Peppers” 7.3146 7.9973 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒 = (ℎ𝑖𝑠𝑡𝑖 − 𝑎𝑣𝑔)2 , (30)
“Baboon” 7.3670 7.9974
256 𝑖=0
“Clock” 6.7057 7.9976
1 ∑
255
𝑎𝑣𝑔 = ℎ𝑖𝑠𝑡𝑖 , (31)
256 𝑖=0
Table 5
Information entropy of different methods for “Lena”. where histi represents the histogram of an image, and avg is the average
Method in [10] in [37] in [52] in [50] Proposed method value of a histogram. The variances of the “Lena”, “Peppers”, “Baboon”
and “Clock” images are listed in Table 3. From this table and Fig. 7, it
Entropy 7.9972 7.9627 7.9973 7.9973 7.9976
can be seen that the variances of the cipher-images are less than those
of the plain-images, and all the histogram distributions of cipher-images
are uniform, which indicates that the attacker is unable to get useful
ensure the accuracy of the test results, a large number of test samples are information from the histogram analysis.
necessary. Therefore, 10,000 pairs of the adjacent pixels in the horizon-
tal, vertical and diagonal directions from the plain-images and the cor-
responding encrypted images are randomly selected. The corresponding 4.2.3. Information entropy analysis
correlation coefficients can be calculated as Information entropy is a measure of the randomness. For an image,
∑𝑁 it can be used to express the uncertainty of the image information. Sup-
(𝑥𝑖 − 𝑥̄ )(𝑦𝑖 − 𝑦̄) pose that an image has 2N grey values 𝑚𝑖 (𝑖 = 0, 1, 2, … , 2𝑁 − 1). The in-
𝐶𝑜𝑟𝑥𝑦 = √ 𝑖=1 , (29)
∑𝑁 formation entropy H(m) can be expressed as
𝑖=1 (𝑥𝑖 − 𝑥
̄ )2 (𝑦𝑖 − 𝑦̄)2
∑𝑁
∑ 1 ∑𝑁 𝐻(𝑚) = − 2𝑖=0−1 𝑝(𝑚𝑖 )𝑙𝑜𝑔2 𝑝(𝑚𝑖 )
where 𝑥̄ = 𝑁1 𝑁 𝑖=1 𝑥𝑖 , 𝑦̄ = 𝑁 𝑖=1 𝑦𝑖 , and xi and yi are the values of two , (32)
∑2𝑁 −1
𝑖=0 𝑝(𝑚𝑖 ) = 1, 𝐻(𝑚) ≤ 8𝑏𝑖𝑡𝑠
adjacent pixels. The correlation distribution of the adjacent pixels in
the three directions of the plain-image and the corresponding encrypted
image are shown in Fig. 6. The corresponding correlation coefficients of where p(mi ) is the probability distribution of mi . For an image with 256
this method and other related works are listed in Table 1. Additionally, levels grey-scale, the ideal value of the information entropy is 8. The
the correlation coefficients of different plain-images are listed in Table 2. closer the entropy is to 8, the stronger the randomness. The informa-
From Fig. 6, Tables 1 and 2, it can be found that the correla- tion entropies of the “Lena”, “Peppers”, “Baboon” and “Clock” images
tion coefficients of the adjacent pixels in the plain-images are approx- are listed in Table 4, and the entropies of the “Lena” image using dif-
imately equal to 1, the correlation coefficients of the adjacent pixels ferent algorithms are shown in Table 5. Table 4 shows that the infor-
in the encrypted images are close to 0, and the correlation coefficients mation entropies of the cipher-images are close to 8, and Table 5 shows
in the three directions are lower than those of other research works that the proposed method has the highest entropy compared with other
[36,37,40,50]. Therefore, the correlation of adjacent pixels is reduced, approaches [9,37,50,52]. These results demonstrate that the proposed
which further illustrates that the encryption scheme has certain resis- work has the ability to resist statistical attacks.
tance to statistical attacks.
4.3. Differential attacks analysis
4.2.2. The histogram analysis
A histogram can reflect the pixel distribution, which makes it a com- If a slight change of the plain-image results in a great difference in
mon tool to evaluate the performance of an encryption scheme. A uni- the cipher-image, it can illustrate that this scheme can effectively resist
formly distributed histogram can effectively resist statistical attacks. differential attacks. Two indexes that can assess the capability are the
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Fig. 7. Histogram analysis. (a), (c), (e), and (g) are the histograms of
“Lena”, “Peppers”, “Baboon” and “Clock”, respectively; (b), (d), (f),
and (h) are the histograms of the corresponding cipher-images.

Fig. 8. NPCR and UACI test: (a) and (b) are the NPCR and
UACI values at different locations, respectively.
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Fig. 9. MSE curves: (a)-(d) are the decrypted

MSE curves with the different secret keys x0 ,
y0 , z0 and h0 , respectively.

number of pixel’s change rate (NPCR) and the unified averaged changing 4.4. Key space analysis
intensity (UACI), which are denoted as
To resist brute-force attacks, the key space must be large enough.
∑
𝑖,𝑗 𝐷(𝑖, 𝑗) When the key space is greater than 2100 , it has a certain ability to resist
𝑁𝑃 𝐶𝑅 = × 100%, (33) brute-force attacks [15]. The secret keys of the encryption algorithm in
𝑀 ×𝑁
this work are x0 , y0 , z0 , h0 and k, and its key space can be expressed
1 ∑ |𝑒1 (𝑖, 𝑗) − 𝑒2 (𝑖, 𝑗)| ∏
𝑈 𝐴𝐶𝐼 = [ ] × 100%. (34) as 𝑆𝑝 = 5𝑖=1 𝑆𝑃𝑖 , where 𝑆𝑃𝑖 is the ith sub-key space of all the secret
𝑀𝑁 𝑖,𝑗 255 keys. Here, the secret key x0 is taken as the example to calculate its key
space. x0 and 𝑥0 + Δ are utilized as the initial values of Chen’s hyper-
′
The size of the plain-image is assumed to be M × N, e1 is the en- chaotic system, and then two different chaotic sequences X and 𝑋 with
crypted image of the original plain-image, e2 is the encrypted image of lengths of N are generated. The mean absolute error (MAE) between two
the plain-image with a one bit change, and D(i, j) is defined as sequences are calculated as

{ 𝑁
0, 𝑒1 (𝑖, 𝑗) = 𝑒2 (𝑖, 𝑗)
′ 1 ∑ ′
𝑀𝐴𝐸 (𝑋 , 𝑋 ) = |𝑋 − 𝑋 |. (36)
𝐷(𝑖, 𝑗) = . (35) 𝑁 𝑖=1 𝑖
1, 𝑒1 (𝑖, 𝑗) ≠ 𝑒2 (𝑖, 𝑗)
If Δ satisfies 𝑀𝐴𝐸 = 0, the key space of x0 is equal to 1/Δ. It can
In this test, the pixel’s lowest bit of each position in the “Lena” image be found that Δ is close to 10−15 , and so that the corresponding secret
is successively changed. Thus, the NPCR and UACI at different locations key space of x0 is 252 [57]. Similarly, the secret key spaces of y0 , z0 , h0 ,
can be counted, and the following curves are obtained and shown in and k are all 252 . Therefore, the total secret key space of this scheme is
Fig. 8. 252×5=260 > 2100 , and this key space is large enough to withstand brute-
Fig. 8 and Table 6 show that the encrypted image has great differ- force attacks.
ences when the plain-image has a slight change. The average NPCR and
UACI using the proposed method are 99.6086% and 33.4507%, respec- 4.5. Key sensitivity analysis
tively, which are higher than those of other methods [53–56].
Moreover, the NPCR and UACI the “Peppers”, “Baboon”, “Clock” im- A good encryption scheme should be sensitive to each secret key so
ages are shown in Table 7. The average NPCR and UACI of the five that it can withstand brute-force attacks. In other words, a slight change
images are very close to the expected values of 𝑁𝑃 𝐶𝑅𝑖𝑑𝑒𝑎𝑙 = 99.6049% in each encryption key will cause serious distortions in the decrypted
and 𝑈 𝐴𝐶𝐼𝑖𝑑𝑒𝑎𝑙 = 33.4635%. Therefore, the proposed method has a good image. To test the key sensitivity, the NPCR, UACI, mean square er-
capacity to resist differential attacks. ror (MSE) and peak signal-to-noise ratio (PSNR) are used. The MSE and
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

Fig. 10. The decrypted images with the

(a) (b) (c) (d) ′
incorrect initial keys: (a)𝑥0 = 𝑥0 + 10−15 ,
′ ′
(b)𝑦0 = 𝑦0 + 10 ,
−15
(c)𝑧0 = 𝑧0 + 10−15 ,
′
(d)ℎ0 = ℎ0 + 10−15 .

(a) (b) (c) (d)

(e) (f) (g) (h)

(i) (j) (k) (l)

(m) (n) (o) (p)

Fig. 11. Occlusion attacks analysis: (a) is the cipher-image with 6.25% occlusion, (b) is the corresponding decrypted image; (c), (e), (i), and (m) are the cipher-
images with 12.5% occlusion, (d), (f), (j), and (n) are the corresponding decrypted images; (g) and (k) are the cipher-images with 25% occlusion, (h) and (l) are the
corresponding decrypted images; (o) is the cipher-image with 50% occlusion, (p) is the corresponding decrypted image.

PSNR are defined as In the case of wrong keys with only a 10−15 deviation from the cor-
rect keys x0 , y0 , z0 , and h0 , and the corresponding decrypted images are
∑𝑀 ∑ 𝑁 obtained. For comparing these decrypted images with the plain-images,
1
𝑀𝑆𝐸 = [𝐷 (𝑥, 𝑦) − 𝑃 (𝑥, 𝑦)]2 , (37)
𝑀 × 𝑁 𝑥=1 𝑦=1 𝑒 the MSE curve can be gained and shown in Fig. 9, and the respective de-
crypted images are shown in Fig. 10. In addition, the mean NPCR, UACI,
MSE, and PSNR are listed in Table 8. From Fig. 9, Fig. 10 and Table 8,
it can be seen that the proposed encryption scheme is highly sensitive
255 × 255 to the secret keys. In [53], the mean NPCR is 99.5383%, which is less
𝑃 𝑆𝑁𝑅 = 10𝑙𝑜𝑔10 , (38)
𝑀𝑆𝐸 than the average value of 99.5979% in the proposed method. Therefore,
the performance of proposed method in key sensitivity is superior to the
where De is a decrypted image with the wrong secret keys that have only work of [53].
slight differences from the correct keys, and P is the plain-image.
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

(a) (b) (c)

(d) (e) (f)

(g)
MSE

k /dbW

Fig. 12. Mixed cipher-images, decrypted images and the corresponding MSE curve under different Gaussian white noises: (a): 𝑘 = 5𝑑𝑏𝑊 , (b): 𝑘 = 10𝑑𝑏𝑊 , (c):
𝑘 = 20𝑑𝑏𝑊 , (d)-(f): The corresponding decrypted images, and (g): MSE curve of different noise intensities.

Table 8 4.6.1. Occlusion attack analysis

The key sensitivity of the proposed scheme based on slightly mod- The test results of this work’s method for the “Lena” image are shown
ified keys. in Fig. 11, and it can be seen that the decrypted images can still be
Incorrect keys NPCR(%) UACI(%) MSE PSNR identified, even though there are different degrees of occlusion attacks.
′ Thus, the encryption scheme can be shown to have a certain resistance
𝑥0 = 𝑥0 + 10 −15
99.5972 30.7214 9132.3218 8.5236
′
𝑦0 = 𝑦0 + 10−15 99.5880 30.6063 9078.8581 8.5505
to occlusion attacks with different degrees and locations.
′
𝑧0 = 𝑧0 + 10−15 99.5956 30.7216 9135.2858 8.5236
′
ℎ0 = ℎ0 + 10−15 99.6109 30.6703 9103.4879 8.5387 4.6.2. Noise attacks analysis
Gaussian random noise and salt-and-pepper noise are two common
signal noises and they are often used to test the ability to resist noise
attacks. Suppose that Gaussian noise is added to the encrypted image C,
4.6. Occlusion and noise attacks analysis which can be described as

𝐶 ′ = 𝐶 + 𝑘𝐺, (39)
In the process of information transmission, when network congestion
′
occurs, noise and congestion are unavoidable. Therefore, the abilities to where 𝐶 is the mixed signal that is polluted by noise, k is the noise
resist occlusion attacks and noise attacks are also important for encryp- intensity and G is the Gaussian noise with zero mean and a standard
tion scheme [58]. deviation of 1. The decrypted image of the mixed signal and the corre-
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

(a) (b) (c)

(d) (e) (f)

(g)
MSE

Density /%

Fig. 13. Salt-and-pepper noise attack analysis: (a)-(c) are the mixed encrypted images with salt-and-pepper noise under densities of 5%, 10%, and 20%, (d)-(f) are
the corresponding decrypted images, and (g) MSE curve of different noise densities.

sponding MSE curve based on different k are shown in Fig. 12. Similarly, mechanics due to the quantum no-cloning theorem and the principle of
salt-and-pepper noise is also used for testing, and the corresponding re- quantum uncertainty [35,39]. The information should be measured if
sults and MSE curve based on different noise densities are shown in the decipherer wants to get it in the quantum state. According to quan-
Fig. 13. tum uncertainty theory, once an unknown quantum state is measured,
From Figs. 12 and 13, it can be seen that the encryption scheme it will irreversibly collapse into an observable eigenstate, and that is the
has strong robustness to resist noise attacks. I.e. the decrypted image reason why legitimate users can detect whether the quantum informa-
can still be identified, even though the image signal is subject to noise tion is attacked from outside or not. Therefore, the principle of quantum
attacks during network transmission. mechanics guarantees the security and robustness of the proposed en-
cryption scheme.

4.7. Discussion
5. Conclusions
Apart from ensuring the basic security of image data, the encrypted
image can still be vulnerable to various attacks during transmission. An image encryption scheme, which is based on Chen’s hyper-chaotic
To improve this drawback, the encrypted image can be represented as a system and quantum coding, is proposed in this work. The NEQR model
NEQR quantum state, as shown in this work. When the encrypted image is applied to map classical digital image data into quantum states, which
is stored and transmitted in the form of a quantum state, it is impossi- can guarantee that the classical images can be accurately extracted. Fur-
ble to accurately reproduce any unknown quantum state in quantum thermore, we combine self-adapting parameters with the new proposed
Y. Luo, S. Tang and J. Liu et al. Optics and Lasers in Engineering 124 (2020) 105836

confusion method based on quantum location coding to encrypt the im- [21] Deutsch D. Quantum theory, the Church-Turing principle and the universal quantum
age, which improves the security of the algorithm. In addition, the quan- computer. Proc R Soc A 1985;400(400):97–117.
[22] Shor P. Algorithms for quantum computation: discrete logarithms and factoring.
tum adjacent-exchange operation reduces the computational complex- In: Proceedings 35th annual symposium on foundations of computer Science; 1994.
ity. The numerical simulation and performance analysis illustrate that p. 124–34.
this encryption scheme has high security, good key sensitivity, larger [23] Grover LK. A fast quantum mechanical algorithm for database search. Annu Acm
Sympos Theory Comput 1996;78(1):212–19.
key space, uniform histogram, and good robustness to resist noise, oc- [24] Laganà A, Pacifici L, Bellucci D. Parallelization strategies for quantum reactive scat-
clusion and differential attacks. Future work includes the development tering codes. Future Generation Computer Systems 2004;20(5):829–40.
of potential applications for multimedia data security using this work. [25] Peng W, Deng H. Quantum inspired method of feature fusion based on von neumann
entropy. INFORMATION FUSION 2014;18(1):9–19.
[26] Qiu L, Cai F, Xu G. Quantum digital signature for the access control of sensitive data
Declaration of Competing Interest in the big data era. Future Gener Comput Syst 2018;86(1):372–9.
[27] Venegas-Andraca SE, Bose S. Storing, processing, and retrieving an image using
quantum mechanics. Proc SPIE 2003;5105(8):1085–90.
The authors declare that they do not have any financial or nonfinan-
[28] Venegas-Andraca SE, Ball JL. Processing images in entangled quantum systems.
cial conflict of interests. Quantum Inf Process 2010;9(1):1–11.
[29] Le PQ, Dong F, Hirota K. A flexible representation of quantum images for polynomial
Acknowledgements preparation, image compression, and processing operations, 10. Kluwer: Academic
Publishers; 2011. p. 63–84.
[30] Zhang Y, Lu K, Gao Y, Wang M. NEQR: A novel enhanced quantum representation
This research is supported by the National Natural Science Founda- of digital images. Quantum Inf Process 2013;12(8):2833–60.
tion of China under Grants 61801131 and 61661008, the Guangxi Nat- [31] Zhang Y, Lu K, Gao Y, Xu K. A novel quantum representation for log-polar images.
Quantum Inf Process 2013;12(9):3103–26.
ural Science Foundation under Grants 2017GXNSFAA198180, the fund- [32] Li H, Zhu Q, Li M, Ian H. Multidimensional color image storage, retrieval, and com-
ing of Overseas 100 Talents Program of Guangxi Higher Education un- pression based on quantum amplitudes and phases. Inf Sci 2014;273(1):212–32.
der Grant F-KA16035, 2018 Guangxi One Thousand Young and Middle- [33] Yang Y, Jia X, Sun S, Pan Q. Quantum cryptographic algorithm for color im-
ages using quantum Fourier transform and double random-phase encoding. Inf Sci
Aged College and University Backbone Teachers Cultivation Program, 2014;277(1):445–57.
and the Science and Technology Major Project of Guangxi under Grant [34] Hua T, Chen J, Pei D, Zhang W, Zhou N. Quantum image encryption algorithm based
AA18118004. on image correlation decomposition. Int J Theor Phys 2015;54(2):526–37.
[35] Yang YG, Tian J, Lei H, Zhou YH, Shi WM. Novel quantum image encryption using
one-dimensional quantum cellular automata. Inf Sci 2016;345(1):257–70.
Supplementary material [36] Liang HR, Tao XY, Zhou NR. Quantum image encryption based on generalized affine
transform and logistic map. Quantum Inf Process 2016;15(7):2701–24.
[37] Wang H, Wang J, Geng YC, Song Y, Liu JQ. Quantum image encryption based
Supplementary material associated with this article can be found, in
on iterative framework of frequency-Spatial domain transforms. Int J Theor Phys
the online version, at doi:10.1016/j.optlaseng.2019.105836 2017;56(10):3029–49.
[38] Li XZ, Chen WW, Wang YQ. Quantum image compression-Encryption scheme based
References on quantum discrete cosine transform. Int J Theor Phys 2018;57(9):2904–19.
[39] Zhou N, Chen W, Yan X, Wang Y. Bit-level quantum color image encryption scheme
[1] Xu L, Li Z, Li J, Hua W. A novel bit-level image encryption algorithm based on with quantum cross-exchange operation and hyper-chaotic system. Quantum Inf Pro-
chaotic maps. Opt Lasers Eng 2016;78(21):17–25. cess 2018;17(6):137.
[2] Li H, Wang Y, Zuo Z. Chaos-based image encryption algorithm with orbit per- [40] Gong LH, He XT, Cheng S, Hua TX, Zhou NR. Quantum image encryption algorithm
turbation and dynamic state variable selection mechanisms. Opt Lasers Eng based on quantum image XOR operations. Int J Theor Phys 2016;55(7):3234–50.
2019;115(1):197–207. [41] Ran Q, Wang L, Ma J, Tan L, Yu S. A quantum color image encryption scheme based
[3] Hu G, Xiao D, Wang Y, Li X. Cryptanalysis of a chaotic image cipher using latin on coupled hyper-chaotic lorenz system with three impulse injections. Quantum Inf
square-based confusion and diffusion. Nonlinear Dyn 2017;88(2):1305–16. Process 2018;17(8):1–30.
[4] Niyat AY, Moattar MH, Torshiz MN. Color image encryption based on hybrid hyper- [42] Li P, Zhao Y. A simple encryption algorithm for quantum color image. Int J Theor
-chaotic system and cellular automata. Opt Lasers Eng 2017;90(1):225–37. Phys 2017;56(6):1961–82.
[5] Hua Z, Zhou Y, Huang H. Cosine-transform-based chaotic system for image encryp- [43] Yang Y, Pan Q, Sun S, Xu P. Novel image encryption based on quantum walks. Sci
tion. Inf Sci 2019;480(1):403–19. Rep 2015;5(1):1–9.
[6] Li C, Lin D. Cryptanalyzing an image encryption algorithm based on autoblocking [44] El-latif AAA, Li L, Wang N, Han Q, Niu X. A new approach to chaotic image en-
and electrocardiography. IEEE Multimedia 2018;25(4):46–56. cryption based on quantum chaotic system, exploiting color spaces. Signal Process
[7] Wang X, Zhu X, Wu X, Zhang Y. Image encryption algorithm based on multiple mixed 2013;93(11):2986–3000.
hash functions and cyclic shift. Opt Lasers Eng 2018;107(1):370–9. [45] Liu H, Jin C. A novel color image encryption algorithm based on quantum chaos
[8] Chai X, Gan Z, Chen Y, Zhang Y. A visually secure image encryption scheme based sequence. 3D Res 2017;8(1):1–13.
on compressive sensing. Signal Process 2017;134(1):35–51. [46] Wang B, Xu J, Song H. Research on the improved algorithm for image quantum
[9] Chai X, Gan Z, Lu Y. A novel color image encryption algorithm based on genetic encryption in multimedia networks. Comput Electr Eng 2017;62(1):414–28.
recombination and the four-dimensional memristive hyperchaotic system. Chin Phys [47] Li D, Liu Y, Gong L. Color image encryption algorithm based on Chua’s circuit and
B 2016;25(10):1–13. Chen’s hyper-chaotic system. J Inf Comput Sci 2015;12(3):1021–8.
[10] Zhang X, Wang H. A novel image encryption algorithm based on genetic recombi- [48] Williams CP, Gries D. Explorations in quantum computing. Springer; 1998. ISBN
nation and hyper-chaotic systems. Nonlinear Dyn 2016;83(1):333–46. 9781846288869.
[11] Wang X, Zhang Y, Bao X. A novel chaotic image encryption scheme using DNA se- [49] Rieffel E, Polak W. An introduction to quantum computing for non-Physicists. Com-
quence operations. Opt Lasers Eng 2015;73(1):53–61. puter Science 2000;32(3):300–35.
[12] Enayatifar R, Gadelha F, Siarry P. Index-based permutation-diffusion in multiple-im- [50] Xu L, Gou X, Li Z, Li J. A novel chaotic image encryption algorithm using block
age encryption using DNA sequence. Opt Lasers Eng 2019;115(1):131–40. scrambling and dynamic index based diffusion. Opt Lasers Eng 2017;91(1):41–
[13] Gong L, Deng C, Pan S, Zhou N. Image compression-encryption algorithms by com- 52.
bining hyper-chaotic system with discrete fractional random transform. Opt Laser [51] Ahmad J, Ahmed F. Efficiency analysis and security evaluation of image encryption
Technol 2018;103(1):48–58. schemes. Int J Video Image Process Netw Secur 2010;12(4):18–31.
[14] Tong X, Zhang M, Wang Z, Ma J. A joint color image encryption and compression [52] Chai X, Yang K, Gan Z. A new chaos-based image encryption algorithm with dynamic
scheme based on hyper-chaotic system. Nonlinear Dyn 2016;84(4):2333–56. key selection mechanisms. Multimed Tool Appl 2017;76(7):9907–27.
[15] Luo Y, Zhou R, Liu J. A novel image encryption scheme based on Kepler’s third law [53] Luo Y, Du M, Liu J. A symmetrical image encryption scheme in wavelet and time
and random Hadamard transform. Chin Phys B 2017;26(12):1–14. domain. Commun Nonlinear Sci Numer Simul 2015;20(2):447–60.
[16] Chai X, Gan Z, Yuan K, Lu Y, Chen Y. An image encryption scheme based on three-di- [54] Liu H, Wang X, Kadir A. Image encryption using DNA complementary rule and
mensional Brownian motion and chaotic system. Chin Phys B 2017;26(2):99–113. chaotic maps. Appl Soft Comput J 2012;12(5):1457–66.
[17] Wang X, Wei N, Zhang D. A novel image encryption algorithm based on chaotic [55] Zhang Y, Wang X. A new image encryption algorithm based on non-adjacent coupled
system and improved gravity model. Opt Commun 2015;338(1):209–17. map lattices. Appl Soft Comput J 2015;26(1):10–20.
[18] Beach G, Lomont C, Cohen C. Quantum image processing (QuIP). In: Applied im- [56] Wang X, Liu L, Zhang Y. A novel chaotic block image encryption algorithm based
agery pattern recognition workshop; 2004. p. 1–6. on dynamic random growth technique. Opt Lasers Eng 2015;66(66):10–18.
[19] Nagy M, Akl SG. Quantum computation and quantum information. Int J Parallel [57] Lambi D. Cryptanalyzing a novel pseudorandom number generator based on pseu-
Emergent Distrib Syst 2006;21(1):1–59. dorandomly enhanced logistic map. Nonlinear Dyn 2017;89(1):2255–7.
[20] Feynman RP. Simulating physics with computers. Int J Theor Phys [58] Luo Y, Cao L, Qiu S, Lin H, Harkin J, Liu J. A chaotic map-control-based and the
1982;21(6–7):467–88. plain image-related cryptosystem. Nonlinear Dyn 2016;83(4):2293–310.