Академический Документы
Профессиональный Документы
Культура Документы
Abstract
Images are widely used to convey information in an easy to understand format; however, the conventional
representation of an image as 3 channels per pixel per image creates difficulty in implementing certain
widely used classes of algorithms. In this paper, we summarize two different applications of the fractional
Fourier transform, which transforms the representation of the data in the image and therefore allows us to
perform novel techniques in this new domain. This allows for effective and efficient solutions to problems
that were difficult or inefficient to solve previously.
1 Introduction
ffective and efficient image processing
1.1 Image Encryption This same idea holds for encryption algo-
rithms as well.
Compared to plain text, images have signifi- Additionally, because images contain a sig-
cantly more information, have more local simi- nificant amount of data, conventional encryp-
larities, and are less sensitive to mutations in tion techniques that work well on plain-text can
the data. For example, if we replaced every take significant amounts of time on images.
other letter in a sentence with a random char-
acter, it would be extremely difficult to piece
back together the original meaning, as in the 1.2 Noise Removal
following:
Another difficult image processing problem to
Tsenqjigklbeomnyfkxtjdmbet yvvrftke efficiently solve given the traditional represen-
llzy yoo. tation of an image is noise removal, since noise
The quick brown fox jumped over the only makes sense in relation to neighboring
lazy dog. pixels. One way to deal with noise that has
a constant frequency throughout the image is
However, if we changed every other pixel in the Fourier transform; however, for noise that
an image to a random color, the image would varies in frequency in the image, the fractional
still be easily discernible, as in Figure 1. Fourier transform must be used instead.
1
2 OVERVIEW OF MATHEMATICAL CONCEPTS
In this paper, we will present summaries to the units of are hertz (Hz), or cycles per sec-
solutions for both of these problems: ond, as:
We will give intuitive and in-depth expla-
nations of the more esoteric mathematical
concepts that are key to understanding Z
the algorithm proposed (Section 2). F ( ) = f (t)ei2t dt (1)
Give a brief intuitive explanation of the al-
gorithm before diving into the actual im-
plementation of the algorithm and show-
ing that it solves the problems outlined And its sibling, the (continuous) Inverse
above (Sections 3.1 and 3.2 for image en- Fourier transform:
cryption, Sections 4.1 and 4.2 for noise
removal).
Analyze the cryptographic effectiveness Z
of the algorithm being proposed and f (t) = F ( )ei2t dt (2)
show that it is reasonably difficult to de-
crypt the image without the proper keys
(Section 3.3).
These equations are rather dense, so we
Show the effectiveness of the noise re-
will give an intuitive explanation of the Fourier
moval algorithm, especially in compari-
transform before diving into the actual mathe-
son to the noise removal applied with the
matics surrounding the equation.
ordinary Fourier transform (Section 4.3).
Consider the function f (t) = sin(3t) +
sin(7t) + sin(12t), graphed in Figure 2:
2 Overview of Mathematical
Concepts 3
f (t)
Many mathematical concepts that this paper 2
relies on will be described in detail in the fol-
lowing sections. Only working knowledge of 1
concepts up to calculus is assumed.
t
2 4 6
2.1 Fourier Transform 1
The Fourier transform is a mathematical trans-
2
formation that converts between two represen-
tations of a signal:
The time-domain representation, or the Figure 2: Graph of sin(3t) + sin(7t) + sin(12t).
value of the signal as a function of time.
The frequency-domain representation, or
the amplitude and phase of the signal as
a function of frequency. This function is composed of 3 sin waves
with frequencies of 3 Hz, 7 Hz, and 12 Hz and
Note that phase has the same definition
a 0 phase angle.
here as it does for waves it is the angle offset
at which we begin drawing a wave. Now, consider the same equation graphed
We can express the (continuous) Fourier in the polar coordinate system instead, using t
transform F of a function f : R C, where as the angular coordinate (Figure 3).
2
2.1 Fourier Transform 2 OVERVIEW OF MATHEMATICAL CONCEPTS
3
2.2 Fractional Fourier Transform 2 OVERVIEW OF MATHEMATICAL CONCEPTS
sample the signal. Thus, we arrive at: uous Fourier transform can be thought of as
Z rotating a signal /2 radians from its time- or
F ( ) = f (t)ei2t dt (5) spatial-domain representation to its frequency-
domain representation. Generalizing this to
Note that Equation 5 is almost exactly the any angle of rotation, the fractional Fourier
Fourier transformation from Equation 2, and it transform can transform a function to inter-
follows the intuition discussed before exactly! mediate domains between the time- or spatial-
The only difference occurs in the exponent, domain and frequency domain.
which is negative in the original presentation To put it more formally, let the a-th or-
of the Fourier transform. der FrFT of a function f ( x ) be denoted as
As it turns out, this negative sign is a mat- F a { f ( x )}( x a ). Note that the a-th order FrFT
ter of convention, and can go on either the corresponds to an = a/2 radian rotation
forward or inverse Fourier transform as long in the time-frequency plane. From our above
as it is kept consistent [10]. Intuitively, we can discussion, we can see that F1 is simply the
think of it as following the circles in one direc- continuous Fourier transform. Going further,
tion in the forward transform and the opposite we can see that F2 of a function f (t) would
direction in the inverse transform, reversing rotate the function into the frequency domain
the operation. Because most texts express the and then back into the time domain, mapping
Fourier transform with the negative sign on the function f (t) 7 f (t). F3 would simply
the forward direction, we will do the same. be the Fourier transform of this inverse time
One detail of importance is that because of domain, and F4 would rotate a full circle and
the ei2t component in the Fourier transform, result in the original function.
the result F ( ) = | F ( )|ei ( ) is a complex num- The mathematical definition is convention-
ber, where the magnitude | F ( )| is the ampli- ally expressed as:
tude of the signal at the given frequency and
( ) is the phase angle. Z
Note that above, we only go into the math- F a { f ( x )}( x a ) = Ka ( x, x a ) f ( x ) dx (6)
ematics of the Fourier transform. Computa-
4
2.3 Two-Dimensional Fractional Fourier Transform 3 IMAGE ENCRYPTION
5
3.2 Detailed Description 3 IMAGE ENCRYPTION
Then, apply a FrFT of order ai . Note Next, we present the mathematical defini-
that this is in reality two FrFT, one tion of the random phase mask described ear-
4. in the x direction and one in the y lier, which is simply Equation 10, for some
direction, and each with their own function n( x ) that generates random numbers
fractional orders aix and aiy . uniformly distributed in [0, 1) based on its in-
In reality, we could continue this loop fur- put parameter.
ther and further, but the diminishing returns
in security and increasing computation time ei2n( x) (10)
make this option unattractive.
We can see that if we multiply f ( x ) by this
The decryption process would be essen-
phase mask, we get the polar representation
tially the same procedure in the opposite direc-
of a complex number in the complex plane,
tion, that is:
with an amplitude of f ( x ) and a phase angle
1. For i = 3...1 of 2n( x ). Because the phase angle is ran-
2. Apply a FrFT of order ai . dom, any correlations between adjacent pixels
Reverse the jigsaw transformation is erased, and the image can be considered
3.
with index bi . white noise after this step.
4. Use the amplitude at each location as the Now, denote the Jigsaw transform with in-
color of the pixel in the decrypted image. dex b as Jb {} and its inverse as Jb {}. If we
Note that we do not need to know the ran- apply it to the image multiplied by the random
dom phase mask used during the encryption phase mask, we have Equation 11.
phase to decrypt the image since the ampli-
tude of the image at a pixel is the color of that Jb1 { f ( x )ei2n( x) } (11)
pixel, and the phase mask does not change the
amplitude of the pixel. However, the phase Now, apply the FrFT of order a1 . Recall
mask is still necessary to whiten the image to that this is actually two transformations, one
be encrypted and remove the effect that the in the x direction and one in the y direction
sharp color differences between jigsaw pieces with two different fractional orders a1x and a1y .
would have on the FrFT result. Without this This results in Equation 12.
phase mask, the encryption result would look
like Figure 6f, which has undesirable patterns F a1 { Jb1 { f ( x )ei2n(x) }} (12)
in the noise. Now, repeating this procedure twice more,
Therefore, the decryption process requires we get the final encrypted image g( x ):
the knowledge of 9 keys in total: 3 indices for
the 3 jigsaw transformations, and 6 fractional
orders, 1 in x and 1 in y for each FrFT. g( x ) = F a3 { Jb3 {F a2 { Jb2 {F a1 { Jb1 { f ( x )ei2n( x) }}}}}}
(13)
We can see the result of this equation in
3.2 Detailed Description Figure 6c when a1x = a1y = a2x = a2y = a3x =
a3y = 0.5.
We will go over the algorithm outlined above in Decryption following our intuition is given
more detail and firmly develop a mathematical by Equation 14.
basis for our claims.
First, to make this section more concise, de-
note the image with a function of one input f ( x ) = Jb1 {F a1 { Jb2 {F a2 { Jb3 {Fa3 { g( x )}}}}}}
f ( x ). Note that all of the below results would (14)
still apply given the conventional representa- Note that as before, we can leave out the
tion of an image as a function of two inputs random phase mask from our decryption step
since the FrFT naturally extends to two dimen- since we can just take the amplitude of the
sions, as discussed previously. function as the color of each pixel.
6
3.3 Results 3 IMAGE ENCRYPTION
Figure 5
One final point to note here is that this al-
As we can see from the attempted decryp- gorithm can be implemented very efficiently
tion in Figure 6e and from the graph above, an in hardware, and in fact, the original paper
error of 0.05 in a fractional order still results in proposes a schematic for a possible optical im-
a completely encrypted image. An individual plementation [3]. Because the Jigsaw transfor-
attempting to brute force the keys would have mation portions of this algorithm are very easy
to go at a resolution of at least 0.05 through the to compute in software, and the difficult FrFT
range of fractional orders [0, 4), which means can be computed with optical hardware at es-
they would have to test 4/0.05 = 80 values sentially instantaneous speeds, this algorithm
for each of the 6 fractional orders, resulting in has the potential to perform extremely well,
806 = 262, 144, 000, 000 combination of values even on large images.
7
4 NOISE REMOVAL
(c) (d)
(e) (f)
8
4.3 Results 4 NOISE REMOVAL
9
REFERENCES
ordinary Fourier transform, resulting in an al- Additionally, the fractional Fourier trans-
most 10-fold improvement. form has been adapted to a variety of other
fields, such as digital watermarking, image
compression, and more [8].
References
[1] Erik D Demaine and Martin L Demaine.
Jigsaw puzzles, edge matching, and poly-
omino packing: Connections and complex-
ity. Graphs and Combinatorics, 23(1):195208,
2007.
Thus, we can see that filtering with the frac- [5] M Alper Kutay and Haldun M Ozaktas.
tional Fourier transform produces significant Optimal image restoration with the frac-
benefits in image quality over filtering with the tional fourier transform. JOSA A, 15(4):825
standard Fourier transform for many types of 833, 1998.
noise with no additional computational cost.
While not depicted, the original paper tested [6] Ervin Sejdic, Igor Djurovic, and LJubia
a couple more of image degradation, all with Stankovic. Fractional fourier transform
superior results for the FrFT method over the as a signal processing tool: An overview
ordinary Fourier transform method [6]. of recent developments. Signal Processing,
91(6):13511369, 2011.
10