Flexible Representation of Quantum Images and Its Computational Complexity Analysis

Phuc Quang Le1, Fayang Dong1, Yoshinori Arai2, Kaoru Hirota1

1 Department of Computational Intelligence and Systems Science, 2 Department of Applied Computer Science,
Interdisciplinary Graduate School of Science and Engineering, Tokyo Polytechnic University,
Tokyo Institute of Technology, 1583 Iiyama, Atsugi, Kanagawa 243-0297, Japan
Midori-ku, Yokohama 226-8502, Japan. E-mail: [email protected]
E-mail: {phuclq, tou, hirota}@hrt.dis.titech.ac.jp

Abstract: Flexible Representation of Quantum Images (FRQI) is proposed to provide a representation of images that enables
efficient preparation and image processing operators on quantum computers. FRQI captures colors and their corresponding
positions in an image into a quantum state. The preparation with polynomial simple operations for FRQI turning a quantum
computer from the initialized state to the FRQI state is shown by the proposed Polynomial Preparation theorem. The Quantum
Image Compression (QIC) algorithm reduces simple operations used in the preparation process based on the minimization of
Boolean equations are from same color groups of positions. Quantum image processing operators on FRQI based unitary
transforms dealing with only colors, colors at some positions and the combination by the quantum Fourier transform of both
colors and positions are addressed. The experiments for the storage and retrieval of images using FRQI are implemented in
Matlab. The compression ratio of QIC among same color groups ranges from 6.67% to 31.62% on the Lena image.

1. Introduction
The method to reduce number of simple operations

With quantum computers [5], the research on image that are used in the image state preparation step,
processing and analysis has faced difficulties. Firstly,
Three types of invertible processing operators on
preparation processes turning quantum computers from their FRQI.
initial states to the image state has to be efficient in term of 2. Flexible Representation of Quantum Images and Its
using a polynomial number of simple operations. Secondly, Polynomial Preparation
all of image processing operators on a quantum computer The information for representing an image in natural
are invertible transforms applying on the image state. A way is colors and their corresponding positions of these
representation for images on a quantum computer that colors. In conventional computers, this way is used for
utilizing both the efficient preparation process and the storing images in pixel representation of images. Inspired by
invertible image processing operators is needed to overcome the fact, the quantum flexible representation of image
the difficulties. (FRQI) is proposed. This proposal integrates information of
On a quantum computation model, research on image an image into a quantum state having its formula in (1)
object began with quantum image representations [1], [2] 2 2 n −1
∑ (cos θ i 0 + sin θ i 1 ) ⊗ i ,
1
I =
2n i =0
(1)
and quantum transforms related to signal processing such as
 π
θ i ∈ 0, , i = 1, 2, K , 2 2 n − 1,
quantum Fourier transform [5], quantum discrete cosine  2

transform [3], quantum Wavelet transform [5]. capturing colors and corresponding positions of those colors,
The Flexible Representation of Quantum Image (FRQI) where ⊗ is the tensor product notation, 0 , 1 are 2-D
which captures information about colors and their computational basis quantum states, i , i = 0,1, K ,2 2 n − 1 are
corresponding positions in image into a quantum state is 2 2 n -D computational basis quantum states and θi are
proposed. Then the following computational and image angles encoding colors.
processing aspects on FRQI are studied: There are two parts in the quantum representation of an

The complexity of the preparation process for FRQI, image; (cos θi 0 + sin θi 1 ) and i which encode the colors
and their corresponding positions in the image, respectively. quantity of simple quantum gates in preparation and
An example of 2 × 2 image is shown in Fig. 1. reconstruction of quantum images is the Quantum Image
The quantum state in FRQI form is a normalized state,

i.e. I = 1 as given by I =
1
[(cos θ 0 0 + sin θ 0 1 ) ⊗ 00 + (cos θ1 0 + sin θ1 1 ) ⊗ 01 +
2
+ (cos θ 2 0 + sin θ 2 1 ) ⊗ 10 + (cosθ 3 0 + sin θ 3 1 ) ⊗ 11 ]

∑ (cos2 θi + sin 2 θi ) = 2n
2 2 n −1
1 1
I = 22 n = 1. (2)
2n i =0 Fig. 1 A simple image and its quantum state.
This property is very useful for further unitary operations on Compression (QIC). The preparation and reconstruction of
the image state. quantum images are the same, however, in sense of
Lemma 2.1. Given a set {θi }, i = 0,1,K, 22 n − 1(n ∈ N ) of computation.
th
angles encoding colors at the position i , there is a unitary There are several reasons why image compression must
transform P that turns quantum computers from the be considered in the FRQI. Firstly, studies in classical image

⊗2 n +1 processing point out that there is redundancy in the image

initialized state 0 to the quantum image state,
that can be reduced for compression in quantum image as
2 2 n −1
∑ (cos θ i 0 + sin θ i 1 ) ⊗ i ,
1 by using Hadamard well. Secondly, as shown in section 2, preparing a quantum
I =
2n i =0
image needs a large number of simple gates. For example a
transforms and controlled-rotation transforms.
216 position image needs 232 simple gates for preparation.
Corollary 2.2. The unitary transform P in lemma 2.1 for a Thirdly, in physical experiments, only a limited number of
given {θ i }, i = 0,1,K, 2 2n
− 1(n ∈ N ) can be implemented by simple gates can be prepared. For all of these reasons, the
simple quantum gates, specifically by 2n Hadamard gates reduction of simple gates is necessary for both theoretical

( )
and 2 2 n generalized- C 2 n R y (2θ i ) gates, where R y (2θi ) and practical aspects of the FRQI.
The Quantum Image Compression (QIC) is proposed in
are the rotations about ŷ axis by the angles
order to reduce controlled-rotation gates in same color
2θ i , i = 1, 2,K, 2 2n
−1. groups based on the minimization of their Boolean
Theorem 2.3 (Polynomial Preparation Theorem) Given a expression as shown in Fig. 2. The algorithm starts with the
set {θi }, i = 0,1,K, 22 n − 1(n ∈ N ) of angles encoding colors information about positions in a same color group and ends
at the ith position, there is an efficiently implemented with the minimized form of the Boolean expression.
quantum transformation P turning a quantum system from
A same color
⊗2 n +1
initialized state 0 to the quantum image state using
Build Boolean minterms
polynomial number of simple gates.
3. Quantum Image Compression based on
Build Boolean expression
Minimization of Boolean Expression
Classical image compression techniques such as JPEG Minimize Boolean
[9] reduce the amount of computational resources, used to
restore or reconstruct images. Similarly, quantum image
Output minimized
compression is the procedure that reduces quantum
resources used to prepare or reconstruct quantum images.
End
The main resource in quantum computation is the quantity
of simple quantum gates or simple operations used in the Fig. 2 Quantum Image Compression Algorithm
computation. Therefore, the process of decreasing the
 The result of QIC algorithm is the minimized Boolean cosθ − sin θ 
matrices U = R y (2θ ) =  , where θ is the
expressions that are used to construct a quantum circuit. The  sin θ cosθ 

number of product terms in a minimized Boolean expression shifting parameter. The following calculation produces the
indicates the number of conditioned-rotation gates that is result I S of the application of S on I ,
needed to use for the corresponding group of same color IS = S( I )
positions. The laterals in a product term in each minimized  1 2 2 n −1 
expression point out the condition part of the conditioned- = S n
2 ∑ (cos θ i 0 + sin θ i 1 ) ⊗ i 

 i=0 
rotation gate. The Boolean variables with complement  1 
( )
2 2 n −1

laterals use extra pair of NOT gates for each complement. = U ⊗ I ⊗n  n

2 ∑ (cos θ i 0 + sin θ i 1 ) ⊗ i  (15)

 i =0 
4. Image Processing Operators on Quantum Images 2 2 n −1
∑ U (cos θ i 0 + sin θ i 1 ) ⊗ i
1
=
based on Unitary Transforms 2n i =0
FRQI absorb information on colors and their related 2 2 n −1
∑ (cos(θ i + θ ) 0 + sin (θ i + θ ) 1 ) ⊗ i .
1
=
spiral positions into a quantum state as the fundamental 2n i =0

object for further quantum transforms. The quantum image The quantum image I S has all of its colors coming from
processing operators are unitary operators with image the original image I by shifting the θ angle.
processing purposes. These operators are divided into 3 The changing color of some points in an image depends
categories based on 3 types of unitary transforms, G1 , G2
on the specific positions of the points in the image.
and G3 , applied to FRQI dealing with only colors, colors at
Information about the positions is used as conditions
some specific positions and the combination of colors and
encoded in the matrix C of the conditioned-gate G2 to
positions, respectively.
construct the processing operators. For instance, let us
The unitary transforms of the first group, G1 and the
consider an 2 × 2 image and the changing color at
second one, G2 are in following form
positions 0 , 2 . The matrix C = 0 0 + 2 2 and
G1 = U ⊗ I ⊗ n , (2)
C = 1 1 + 3 3 are used to construct the conditioned-gate
G2 = U ⊗ C + I ⊗ C , (3)
G2 ,
where U are single qubit unitary transforms on colors, C are
G2 = U ⊗ ( 0 0 + 2 2 ) + I ⊗ ( 1 1 + 3 3 ).
matrices regarding eligible positions, C are matrices
The action of this particular G2 on a general 2 × 2 image in
1 0 
regarding ineligible positions, I =   is the identity 1 3
0 1 FRQI form, I = ∑ (cosθi 0 + sin θi 1 ) ⊗ i , is given by
2 i =0
operator and n is the number of qubits encoding positions.
[
G2 I = U ⊗ ( 0 0 + 2 2 )+ I ⊗ ( 1 1 + 3 3 ) I ]
=
1
(cos θ 0U 0 + sin θ 0U 1 )⊗ 0 +
G1 2
+ (cos θ 2U 0 + sin θ 2U 1 ) ⊗ 2 +
1
(16)
2
+ (cos θ 1 0 + sin θ 1 1 ) ⊗ 1 +
1
2
+ (cos θ 3 3 + sin θ 3U 1 ) ⊗ 3 .
1
G2
2
The calculation in (16) shows that the action of operator
U for changing color affects only on the specific
Fig. 3 The circuits of G1 and G2 image processing
positions 0 , 2 .
operations.
The shifting color operator, S is defined as an operator in The number of simple gates used for an operation in G1

the group G1 , S = U ⊗ I ⊗n , by using rotation category is one gate while the number of controlled-rotations
used for an operator in G2 category depends linearly on FRQI based unitary transforms dealing with only colors,
number of positions involving in the operator. colors at some positions and the combination by the
quantum Fourier transform of both colors and positions are

G3 addressed. The experiments for the storage and retrieval of

images using FRQI are implemented in Matlab. The
compression ratio of QIC among same color groups ranges
Fig. 4 The circuit of the operator G3 based on QFT
5. Experiments on Quantum Images from 6.67% to 31.62% on the Lena image.
The essential requirements for a representation of Reference
classical or quantum images are simplicity and efficiency in [1] S. E. Venegas-Andraca, J. L. Ball: Storing Images in
storage and retrieval the images. The storage of quantum Entangled Quantum Systems, quant-ph/0402085, 2003.
image is preparation process that is ensured by the proposed [2] J. I. Latorre: Image compression and entanglement,
EPT in section 2. The measurement of the quantum image arXiv:quant-ph/0510031, 2005.
state produces a probability distribution that is used for the [3] A. Klappenecker, M. Rötteler, Discrete Cosine
retrieval of the image, shown in fig. 5. Transforms on Quantum Computers, Proceedings of
the 2nd International Symposium on Image and Signal
Processing and Analysis, pp. 464-468, 2001.

[4] A. Fijany, C. P. Williams, Quantum Wavelet

Fig. 5 Coefficients of quantum state and probability Transforms: Fast Algorithms and Complete Circuits,
distribution arXiv:quant-ph/9809004, 1998.
In experiment, compression ratios among groups are
estimated based on analysis of minimizations of Boolean [5] M. A. Nielsen, I. L. Chuang: Quantum Computation

expressions derived from the 256 × 256 gray image of Lena and Quantum Information, Cambridge Univ. Press,

as in the Fig. 6. The minimization step in QIC algorithm is 2000.

done by Logic Friday free software. The compression ratios
[6] A. Barenco, C. H. Bennett, R. Cleve, D. P. DiVincenzo,
range from 6.67% to 31.62% between groups.
N. Margolus, P. Shor, T. Sleator, J. A. Smolin, H.
Weinfurter: Elementary gates for quantum computation,
Phys. Rev. A 52, 3457, 1995.

[7] C. Lomont: Quantum convolution and quantum

correlation algorithms are physically impossible,
quant-ph/0309070, 2003.

Fig. 6 Compression ratios of same color groups in Lena image [8] R. K. Brayton, A. Sangiovannni-Vincentelli, C.
6. Discussion
McMullen, G. Hachtel, Logic Minimization Algorithms
FRQI captures colors and their corresponding positions
for VLSI Synthesis, Kluwer Academic Publishers, 1984.
in an image into a quantum state. The preparation with
polynomial simple operations for FRQI turning a quantum [9] JPEG homepage, http://www.jpeg.org/

computer from the initialized state to the FRQI state is

shown by the proposed Polynomial Preparation theorem.
The Quantum Image Compression (QIC) algorithm reduces
simple operations used in the preparation process based on
the minimization of Boolean equations are from same color
groups of positions. Quantum image processing operators on