TELKOM
NIKA
, Vol. 13, No. 4, Dece
mb
er 201
5, pp. 1298
~1
304
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i4.1900
1298
Re
cei
v
ed Au
gust 23, 20
15
; Revi
sed O
c
t
ober 2
8
, 201
5; Acce
pted
No
vem
ber 1
2
,
2015
Image Restoration Based on Hybrid Ant Colony
Algorithm
Yan Feng*, Hua Lu and Xiliang Zeng
Hun
an Un
ivers
i
t
y
of Internati
o
nal Eco
nomics,
Chan
gsh
a
410
205, Hu
na
n, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
4899
129
3@
q
q
.com
A
b
st
r
a
ct
Imag
e restor
ati
on is th
e pr
oce
ss to eli
m
inate
or re
d
u
ce th
e i
m
a
ge
qu
ality d
egra
datio
n i
n
the d
i
gita
l
imag
e for
m
atio
n, trans
missi
on
and r
e
cor
d
in
g
and
its
pur
pos
e is to pr
ocess
the obs
erve
d
degr
ade
d i
m
ag
e
to make th
e re
stored res
u
lt a
pprox
imate th
e
un-d
egr
ade
d
origi
n
a
l
i
m
a
ge.
T
h
is pa
per, b
a
sed
on th
e b
a
si
c
ant col
ony
alg
o
rith
m a
nd i
n
te
gratin
g w
i
th the ge
netic
a
l
g
o
r
i
thm, pr
op
oses
an i
m
age r
e
storatio
n proc
es
sin
g
meth
od
base
d
on hy
brid
ant colo
ny al
gorith
m
. T
h
is
me
th
o
d
transfor
m
s th
e opti
m
a
l
p
opu
latio
n
infor
m
ati
o
n
of genetic a
l
go
rithm i
n
to the o
r
igin
al
p
hero
m
one co
ncentrat
i
on
matrix of ant
colony
al
gor
i
t
hm an
d
uses
i
t
to
compute the
p
a
ra
meters of d
egra
datio
n fun
c
tion so as to
get a precis
e e
s
timate
of
the origi
n
a
l
i
m
ag
e. By
ana
ly
z
i
n
g
an
d
compari
ng th
e restoratio
n r
e
sults, t
he
me
thod of this p
aper ca
n not
only ov
erco
me
the
influence of noises, but it c
an al
so
make
the im
age s
m
oother with
no
fringe effects in the edges
and
excell
ent visu
al
effects, verifying its practica
b
ility.
Ke
y
w
ords
: Image R
e
storati
o
n, Ant Colony
Algorit
h
m
, Genetic Algor
ith
m
Copy
right
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Image resto
r
ation is to
re
search h
o
w to
rest
o
r
e th
e d
egra
ded i
m
a
ge into real i
m
age, o
r
to re
sea
r
ch
how to
invert the inform
a
t
ion obtain
e
d
into the inf
o
rmatio
n rela
ted to the
re
al
obje
c
tive [1]. Ce
rtain
deg
ree
of de
gra
dation
and
d
i
stortion
a
r
e
inevitable in
the form
ation
,
transmissio
n, storage,
re
cordin
g a
nd
di
splay
of the i
m
age. Sin
c
e
image
quality
degradatio
n
may
be ca
used in
every link of t
he formatio
n
of digital
ima
ge, in many case
s, the ima
ge nee
ds to
be
resto
r
e
d
in order to get hig
h
-qu
a
lity digital image [2].
In the
pa
st d
e
ca
de
s, do
m
e
stic an
d fo
re
i
gn
schola
r
s
have m
ade
e
x
tensive a
nd
in-de
p
th
research
in image res
t
orati
on tec
hnology. Many
one-dimens
ional
s
i
gnal process
i
ng and
estimation
th
eorie
s, i
n
clu
d
i
ng inve
rse fi
ltering,
mi
nim
u
m me
an
error
estimatio
n
and
Baye
sia
n
estimation,
h
a
ve be
en u
s
e
d
in th
e field
of image
re
st
oration, th
us
forming
num
erou
s
re
stora
t
ion
algorith
m
s. In
terms of cert
ain sp
ecifi
c
re
storat
io
n prob
lems, the sch
o
lars u
s
ually i
n
tegrate m
a
n
y
idea
s an
d m
e
thod
s [3]. With the contin
uou
s devel
op
ment of si
gn
al proces
sing
theori
e
s, ma
ny
new p
r
o
c
e
s
sing idea
s an
d re
storatio
n
method
s ke
ep eme
r
gin
g
. In drasti
c contra
st with th
e
typical mathe
m
atical p
r
og
ramming p
r
in
ciple
s
, so
me
bionic intelli
gent optimi
z
ation algo
rith
ms
su
ch as a
n
t colo
ny algorit
hm, genetic
algorith
m
,
artificial neu
ral netwo
rk te
ch
nology, artificial
immune al
g
o
rithm an
d swarm intelli
gen
ce alg
o
ri
thm, have been raise
d
and stu
d
ied
by
simulatin
g
th
e natural e
c
o-sy
stem to
see
k
the
complicated o
p
timization
p
r
oble
m
s. Th
ese
algorith
m
s
h
a
ve greatly
enri
c
he
d mo
dern
optim
i
z
ation te
chnol
ogy and
p
r
ovided fea
s
i
b
le
solutio
n
s to
those o
p
timi
zation p
r
o
b
le
ms which are difficult to be han
dled
by tradition
al
optimizatio
n t
e
ch
nolo
g
y. Ant col
ony alg
o
rithm i
s
a he
uristi
c bi
oni
c
evoluti
ona
ry system ba
se
d
on
popul
ation. By adopting
distribute
d
plus-feed
ba
ck paralleli
zati
on, this algo
rithm is ea
sy to
combi
ne
with
other
metho
d
s a
nd it al
so
has str
ong
robu
stne
ss, h
o
weve
r, this
algorith
m
re
q
u
ire
s
long sea
r
ch time and it is easy to re
sult in pre
-
ma
ture an
d sta
gnation b
eha
viors, sl
owi
n
g its
conve
r
ge
nce
rate. On t
h
e othe
r ha
n
d
, geneti
c
al
gorithm i
s
a
ran
domi
z
ed
adaptive
se
arch
algorith
m
by referring to
the natural
sele
ct
ion a
nd natu
r
al g
enetic m
e
ch
anism
and it
can
comp
ute the
non-lin
ear
multi-dime
nsi
onal dat
a space in a quick an
d effective manner.
Therefore, th
e integratio
n
of these two tec
hniq
u
e
s
or algorith
m
s ca
n elim
inate their o
w
n
sho
r
tco
m
ing
s
and setb
acks and utili
ze their o
w
n adv
antage
s in th
e image resto
r
ation [4, 5].
This p
ape
r firstly sum
m
a
r
ize
s
the th
eory
of ima
ge re
storatio
n system
atically and
analyzes the
basic p
r
in
ci
ple of ant colony algorit
h
m
. Then, it focu
se
s on th
e research o
f
the
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No
. 4, Decem
b
e
r
2015 : 129
8 – 1304
1299
hybrid m
e
tho
d
of ant
colo
n
y
algorithm
a
nd ge
netic
al
gorithm. Fi
nal
ly, it compa
r
e
s
the m
e
thod
of
this paper
with other met
h
ods and verifi
es its
practicability through the experi
m
ental sim
u
l
a
tion
and an
alysi
s
of the image restoration.
2. Basic Prin
ciple of Image Res
t
ora
t
io
n
Subject to the influen
ce
of such fa
ctors
as o
p
tical diffra
c
tio
n
imaging, turbul
ent
disturban
ce,
relative m
o
tion of im
aging o
b
je
ct
ives, non
-ide
al cha
r
a
c
teri
stics of ima
g
ing
appa
ratu
s, limited tran
sm
issi
on me
diu
m
band,
ma
n-ma
de defi
n
ition and
re
quire
ment
s a
nd
rand
om environmental n
o
i
s
e
s
, the image quality de
grad
es at different deg
ree
s
in the image
formation, transmi
ssion a
nd re
co
rdin
g
[6]. The
typ
i
cal de
gra
dat
ions in
clu
de
image blu
rri
n
g
,
distortio
n
a
n
d
noises. T
h
is
pro
c
e
s
s is
ca
lled im
a
ge d
e
g
rad
a
tion, th
e key to i
m
ag
e re
storation
is
to know the
pro
c
e
ss of im
age deg
ra
dat
ion, namely to kno
w
the image de
gra
d
a
tion model, and
to see
k
the o
r
iginal
(cl
ear) image by ad
opting
the in
verse
pro
c
e
s
s on the b
a
si
s of the mod
e
l.
Since the im
age u
s
ually comes
with no
ise
s
, the
noises not only d
egra
de the i
m
age qu
ality, bu
t
also
affect th
e imag
e rest
oration
effect,
therefo
r
e, im
age d
e
-noi
sin
g
is
also a
re
sea
r
ch fo
cu
s
of
this
paper [7].
In the image resto
r
atio
n, it is usu
a
lly assu
me
d that the image d
e
g
rad
a
tion is f
o
rme
d
by
the imag
e bl
urri
ng
cau
s
e
d
by point
sprea
d
fun
c
tio
n
(PSF)
and
the noi
se
p
o
llution a
nd
its
mathemati
c
al
model is:
12
12
12
(,
)
(
)
(
,
)
(
,
)
f
mm
g
v
m
m
mm
(
1
)
In this
formula,
,,
vg
are the origin
al image, the PSF and the a
ddictive noi
ses
respe
c
tively and
is the co
nvolution op
e
r
ator [8]. The
pur
p
o
se of im
age resto
r
ati
on is to obtai
n
a pre
c
ise esti
mation of the origin
al imag
e
v
from the degra
ded im
ag
e
f
.
For the a
bov
e, it can be
kno
w
n that i
m
age re
sto
r
a
t
ion is an
“in
v
erse
pro
b
le
m” in the
mathemati
c
al
pro
b
lem
s
a
n
d
one
impo
rt
ant prope
rty
of the inverse
pro
b
lem
s
i
s
their ill
-conditi
on.
Image resto
r
ation, in a
ccorda
n
ce
with
the
de
gra
d
a
t
ion re
ason
s, analyzes th
e environme
n
tal
factors that indu
ce imag
e
degra
dation
,
build
s the corre
s
p
ondin
g
mathemati
c
al mod
e
l and
resto
r
e
s
th
e i
m
age
acco
rdi
ng to the
inv
e
rse p
r
o
c
e
s
s of imag
e de
grad
ation.
Du
e to noi
se
s, t
he
resto
r
atio
n re
sult of the image may devi
a
te from the origin
al imag
e [9].
3. Basic Prin
ciple of An
t Colon
y
Algorithm
Ant colo
ny a
l
gorithm
se
arche
s
the
opt
imal
solution
s to the
pro
b
lems ad
apti
v
ely by
simulatin
g
th
e ants i
n
the
biologi
cal
wo
rld to se
ar
ch t
he shorte
st p
a
th from thei
r nest to the f
ood
sou
r
ces witho
u
t any visible
sub
s
tan
c
e
s
.
The me
ch
ani
sm of a
n
t col
ony algo
rithm
is: when
an
ant
is sea
r
ching
the
food
sou
r
ce
s,
it will relea
s
e th
e u
n
ique
ph
ero
m
one, m
a
ki
n
g
the
othe
r
ants
within a ce
rta
i
n distan
ce range can pe
rceive and
aff
e
ct their be
h
a
viors. When
more and m
o
re
ants wal
k
a
certain path, more
and
more pheromone will
be left i
n
th
is path,
thus,
increasing the
pheromo
ne in
tensity and the prob
ability for othe
r ants
to choo
se thi
s
path, whi
c
h
will also furth
e
r
increa
se the
pheromo
ne in
tensity [10].
Here, this p
aper will
bri
e
fly introdu
ce
the ba
si
c a
n
t col
ony alg
o
rithm to
be
used in
travelling
sale
sman
proble
m
(
TSP
) with the
TSP
in
n
cities as example.
Th
e
TSP
in
n
cit
i
es i
n
the ant colo
ny algorithm
model is to
find the
sh
ortest pat
h back to the starting poi
nt after
sea
r
ching
n
c
i
ties
onc
e
[11].
Assu
me that there a
r
e
n
cities an
d
m
ants,
(,
1
,
2
,
,
)
ij
di
j
n
is
th
e
d
i
s
t
an
c
e
be
tw
ee
n
the cities
of
,
ij
and
()
ij
t
is the p
hero
m
on
e intensity in the
paths b
e
twe
e
n
the citie
s
of
,
ij
at
time
t
. When
Ant
k
pro
c
ee
d
s
, its path i
s
determi
ned
b
y
the resi
due
pheromo
ne i
n
tensitie
s in
every path a
nd
()
k
ij
pt
is the probability for Ant
k
to move from City
i
to City
j
at time
t
, then
[12]:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Resto
r
ation Base
d o
n
Hybrid Ant
Colo
ny Algo
ri
thm
(Yan Fen
g
)
1300
()
[
(
)]
[
(
)]
,(
)
[(
)
]
[
(
)
]
()
0,
k
ij
ik
k
k
is
is
ij
sJ
i
tt
jJ
i
tt
pt
Ot
h
e
r
w
i
s
e
(
2
)
()
(
1
,
2
,
,
)
k
Ji
i
n
k
t
abu
is
the
c
i
ty s
e
t Ant
k
may
step
on
next and
List
k
t
abu
is
the
Tabu
List of Ant
k
. When
n
cities
have be
en li
sted into the
k
ta
b
u
, Ant
k
has trav
elled th
roug
h
all the cities
and it has fin
i
she
d
one tra
v
erse
afte
r g
o
ing ba
ck to
the starting p
o
int, then every
path Ant
k
ha
s pa
ssed i
s
a
feasibl
e
soluti
on to thi
s
TSP
.
ij
is the
heu
ri
stic fa
ctor, in
dicating
the expe
ctati
on fo
r Ant
k
to move from
City
i
to
City
j
, whi
c
h
is u
s
u
a
lly the
re
ciprocal
of
ij
d
,
,
are the
rel
a
tively importa
n
t
pro
c
e
dures
of
phe
rom
o
n
e
an
d h
e
u
r
ist
i
c fa
ctor in th
e eq
uation
respe
c
tively, whe
n
all th
e
ants
have fini
she
d
o
ne tra
v
erse,
glob
al
upd
ate will
be
con
duct
e
d
in
the pheromo
ne in every p
a
th according
to Formula (3):
1
(1
)
(
1
)
(
)
=
ij
ij
ij
m
k
ij
ij
k
tt
(
3
)
In this
formula,
01
（）
is the volatilization coefficient,
1-
is the durability
c
oeffic
i
ent,
ij
is the
phe
ro
mone i
n
cre
m
ent after this ite
r
ation
and
k
ij
is th
e
re
sidu
al
pheromo
ne o
f
the
th
k
Ant in this iteratio
n. Acco
rdi
ng to different phe
romone u
pdat
e strate
gies,
the schemati
c
of ant colo
n
y
algor
ithm is indicate
d in Figure 1.
(a)
(b)
Figure 1. Sch
e
matic of ant
colo
ny algorit
hm
From the a
b
o
ve, it can be se
en tha
t
ant co
lony
algorithm h
a
s such feature
s
of
parall
e
lism,
p
o
sitive fee
d
b
a
ck a
nd
stro
ng gl
obal
mi
nimum
se
arching
ca
pa
city. The
ants will
select the path according to it
s pheromone
concentrat
ion and they
will choose
the
path with
hi
gh
con
c
e
n
tration
.
When the
r
e
are mo
re an
d more a
n
ts
passin
g
a pa
th, the resid
u
al phe
romo
n
e
concentration of this pa
th
will increase
gradually and the proba
bili
ty for the ant
s to
choose this
path al
so
in
creases,
su
gge
sting th
e p
o
si
tive feedba
ck
of info
rmatio
n. In the
mea
n
whil
e, with
th
e
increa
se of time, the pheromone inte
nsit
y will decre
a
s
e gradu
ally [13, 14].
4. Main Procedures o
f
Image Re
sto
r
ation Base
d on H
y
brid Ant Colon
y
Alg
o
rithm
Set the size of an image
as
LM
N
,
M
and
N
are the width
and hei
ght of
the
image, the grayscale is
K
,
the ch
romo
so
me is rep
r
e
s
ented by
L
two-dime
nsi
onal
array, every
element of which
co
rre
sp
o
nds to a g
ene
, use int
ege
r
codi
ng, name
l
y the real value of gray
sca
l
e
to represent a gene.
In the experiment, limite the image a
s
256-col
o
r two
-
dime
nsi
onal
grayscal
e image an
d
cod
e
the ch
romosome a
s
the two-dim
ensi
onal matr
ix with the grayscale of every pixel as the
element. Fo
r example, if
the imag
e si
ze i
s
M
N
, then the ge
nic val
ue of the
ch
romosome
A
Ant
nest
F
o
od
B
A
Ant
nest
F
o
od
B
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No
. 4, Decem
b
e
r
2015 : 129
8 – 1304
1301
,
(
0
,
1
,
2
,
,
1,
0
,
1,
2
,
,
1
)
ij
xi
N
j
M
is
the gr
ayscale value in t
he
th
i
ro
w a
nd
th
j
colum
n
in
the ima
ge. Si
nce
the
grayscale val
ue i
s
the
inte
ger within
[0,25
5
], the al
gorit
hm of thi
s
pa
pe
r
adopt
s intege
r codi
ng to pe
rform
cro
s
sov
e
r and m
u
tation ope
ration
s.
The fitness fu
nction of every individual (chrom
osome
)
is:
2
ˆ
ˆ
()
de
f
ii
Ef
g
h
f
(
4
)
In this
formula,
ˆ
i
f
is th
e ima
ge represent
ed by Individ
ual
i
,
g
is the d
egra
ded
imag
e to
be ob
serve
d
,
h
is the degrad
ation pro
c
e
ss and
is co
nvolution.
Therefore, th
e su
per-resol
uti
on re
sto
r
at
ion of the im
age h
a
s b
e
come the o
p
timization
probl
em of this algo
rithm.
The ant
colo
ny algorith
m
fails to find
an
optim
al sele
ction of the
p
a
ram
e
ter in
a
d
vance,
makin
g
it fail
to achi
eve th
e optimal
perf
o
rma
n
ce of a
n
y pro
b
lem
s
. Gene
rally, after a
nalyzin
g t
he
specific
probl
ems, ant col
o
ny al
gorithm
will continue to adjust t
he param
e
ters
in the
experience,
hopin
g
to g
e
t the optim
al
solutio
n
. Moreover, th
e
hy
brid
algo
rith
m of ant
col
o
ny algo
rithm
and
geneti
c
algo
ri
thm can furth
e
r improve the comp
ut
atio
n efficien
cy.
First, pre
p
rocess the imag
e
and the
n
take the ima
ge
as the i
n
itial
popul
ati
on a
n
d
pe
rform
co
mbined
ope
ration of ant
colony
algorith
m
and
genetic al
gorithm. Its co
mputation ide
a
is as follo
ws:
(1)
Dete
rmin
e the initial p
a
ram
e
ters of
t
he geneti
c
a
l
gorithm a
nd
ant colo
ny al
gorithm
and initialize the feasi
b
le p
opulatio
n.
(2) O
p
e
r
ate the gen
etic al
gorithm a
nd seek
the evol
u
t
ionary feasi
b
le solutio
n
s.
Comp
ute th
e individual
fitness i
n
th
e pop
ulation
,
sele
ct the
geni
c valu
e in the
chromo
som
e
acco
rdi
ng t
o
the
fitness of
t
he i
n
d
i
vidual, pe
rfo
r
m
crossove
r a
nd
mutati
on
operation
s
o
n
the
geni
c v
a
lue in
the
chrom
o
some
,
cal
c
ulate
the
individual
st
atus i
n
the
n
e
w-
gene
ration
p
opulatio
n, re
cord the
cu
rrent ev
ol
utio
nary p
opulati
on an
d eval
uate the
cu
rrent
optimal indivi
dual.
(3) Tran
sform
the o
p
timal
popul
ation inf
o
rmatio
n
of t
he g
eneti
c
al
gorithm
into t
he initia
l
pheromo
ne concentratio
n
matrix of
the ant colo
ny algorithm.
Empty the
Tabu
list which
stores th
e
acce
ssed
no
de
s, set the
node
l
i
st which h
a
ve
been a
c
cessed by the ants as all set
s
, determi
ne th
e
starting point
of the ants, u
pdate the nod
es
whi
c
h are
all
o
we
d
to be
acce
ssed,
th
e
Tabu
list a
nd th
e un
acce
sse
d
no
de
s, sel
e
ct the
next
node
to b
e
acce
ssed
acco
rding to
the
st
atus t
r
an
sf
er
prob
ability formula, after th
e tra
n
sfe
r
no
de,
update the n
ode
s whi
c
h a
r
e allo
wed to
be accesse
d
, the
Ta
bu
list and
the unacce
ssed no
de
s
and jud
ge wh
ether the ant
s have compl
e
ted the traverse.
(4) Sele
ct the
phe
rom
one
update
me
ch
anism,
upd
ate the
phe
rom
one
co
ncentration of
every node,
evaluate the curre
n
t paths, update t
he pheromo
ne concentratio
n
of every path,
pre
s
e
r
ve the optimal path the ant se
arch, compa
r
e
a
ll the feasible
paths an
d o
u
tput the optimal
path.
The flowch
art
of the image
resto
r
atio
n al
gorithm
b
a
se
d on hybri
d
a
n
t colony alg
o
rithm is
as follo
ws Fig
u
re 2 [15, 16]
:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Resto
r
ation Base
d o
n
Hybrid Ant
Colo
ny Algo
ri
thm
(Yan Fen
g
)
1302
Inp
u
t deg
r
ad
ed image
De
fine the
fit
n
ess
functio
n,
the population s
i
ze,
the
maximum
iterations
Max
an
d var
i
ou
s
p
a
ram
e
ter valu
es
Select the
optimal indivi
dual in the population
t
o
duplicate to the ne
xt generation
Perform s
e
lec
t
ion,
crossover, mutation
ope
ration
a
nd calculate t
h
e fitness
Select N relative
l
y optimal
individuals
fr
om the n
e
wly-p
r
o
d
uc
ed po
pu
la
tio
n
Determine the ne
xt pixe
l
point to be
accessed according to the acc
essibility of
the re
s
o
urce
pixel points,
update the
p
h
er
om
one
o
f
the pix
e
l
p
o
in
ts and
upd
ate
the lis
t of
various
pixel points
Reca
l
c
ulate the acce
ssibility
Whet
her t
h
e l
i
st
of
th
e pixel points
to be
acce
ssed i
s
empty?
Evaluate
the
c
u
rrent es
tima
t
i
o
n of
origi
n
al
image main struc
t
ure and edge,
ex
tract the
optimal data
Outp
ut res
t
ore
d
image
Extrac
te
of the im
age matrix
data,
rand
om
ly pro
d
u
ce th
e po
pu
latio
n
of
a
gr
ou
p of
f
easib
le so
lu
tion
s
Co
de the chrom
o
some
as the two-
di
mensional matrix
w
i
t
h
t
h
e gr
a
y
s
c
al
e of
e
v
ery pixe
l
as the element
Calculate
th
e r
e
si
dual
imag
e, plus p
i
xel
by
pixel,
each
pi
xel value
is the weighted av
e
r
age
of all
the points
of
neig
hbo
r
h
o
o
d
pixe
ls
valu
e and
th
e
a
n
d p
o
s
i
tio
n
Get initial image
res
t
or
ation by combining
image com
pon
ents wi
th
t
e
xtur
e comp
onents
Reg
u
lar
i
ze r
e
si
d
u
al
image
Get
fina
l res
t
ored image
by non
local mean filter processi
ng
Y
Whether meet the loo
p
termina
t
ion ?
Initialize vari
ous parameters
and t
r
ansform the
previous
relatively optimal
individu
als
int
o
the ph
ero
m
one
distribution in the
ea
rly stag
e of
the population algorithm
Initialize the lis
t which
has been obtained
by the artificial an
t, the u
n
-a
ccessed
a
n
d
accessed li
st
N
N
Y
Figure 2. Image re
storation
flowch
art ba
sed o
n
hybrid
ant colony al
gorithm
5. Experimental Simulation and An
aly
s
is
In orde
r to prove the effect
iveness of th
e algo
rithm o
f
this pape
r, we have te
st
ed it with
binary
imag
e. In the
re
sto
r
ation, we
sim
u
late the
o
p
e
r
ation
s
of
hyb
r
id ant colo
ny
algo
rithm.
Fi
rst,
we
experi
m
e
n
t on th
e bi
n
a
ry ima
ge
a
nd the
blu
r
operators use
Ga
ussia
n
blur ope
rato
rs
of
10*10.
Use t
h
ree
bina
ry i
m
age
s, ther
e
f
ore, all the v
a
lue
s
a
r
e 1
o
r
0. As i
ndi
ca
ted in Fig
u
re
(3),
(a),
(c) a
nd
(e) a
r
e th
ree
different d
egree of
d
egrad
ed ima
g
e
s
, while
(b
), (d
) and
(f)
are
the
resto
r
e
d
imag
es by the alg
o
rithm of this
pape
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No
. 4, Decem
b
e
r
2015 : 129
8 – 1304
1303
(a)
Deg
r
ad
ed
image
(b)
Re
store
d
image of (a
)
(c) De
gra
ded
image
(d)
Re
store
d
image of (c)
(e) Deg
r
ad
ed
image
(f) Re
sto
r
ed i
m
age of (e
)
Figure 3.
Re
store
d
imag
e based on
Hybrid Ant Col
o
ny Algorithm
It is clea
r fro
m
the expe
ri
mental result t
hat this al
g
o
rithm m
a
kes the re
sto
r
ed
image
s
mother
with
no fring
e
effe
cts a
nd excel
l
ent visual
eff
e
cts, imp
r
ove
s
the subje
c
ti
ve effects to a
certai
n exten
t; reduce
s
th
e ringin
g
ph
enome
nom n
ear the imag
e edge
s, make
s the edg
es
sha
r
pe
r
and
t
he o
u
tline
cle
a
re
r a
n
d
re
du
ce
s the
bl
urri
ness
of the
lo
cal
area
s. It sugge
sts that t
he
post-processi
ng algo
rithm in this pape
r can im
p
r
ove
the geomet
ry regula
r
ity of
the edge
s an
d
texture of the re
store
d
i
m
age a
nd e
nhan
ce
s subj
ective and o
b
jective qu
ali
t
y of the restored
image.
6. Conclusio
n
Having
wide
appli
c
ation fi
elds, imag
e restoration te
chn
o
logy ha
s beco
m
e on
e of the
resea
r
ch hot spot
s in the image field at
home and a
b
roa
d
. This p
aper h
a
s inte
grated the a
n
t
colo
ny algo
rithm and
gen
e
t
ic algo
rithm
and a
ppli
ed t
o
the imag
e
resto
r
atio
n proce
s
sing. Thi
s
hybrid alg
o
rit
h
m ca
n se
arch the
soluti
ons in
m
any
global poi
nts sim
u
ltane
o
u
sly and it h
a
s
parall
e
lism,
p
o
sitive fee
d
b
a
ck m
e
chani
sm
and
st
ron
g
unive
rsality. The
expe
ri
ment result h
a
s
sho
w
n th
at the meth
od o
f
this pa
pe
r
has better ef
fect in
re
stori
ng the i
m
ag
e re
sol
u
tion
and
quality.
Referen
ces
[1]
Pabl
o Ruiz, Hi
ram Madero-O
r
ozco, Javier
Mateos
, Osslan Osiris Verga
r
a-Vill
eg
as, Ra
fael Moli
na,
Agge
los K. Ka
tsagge
los. Co
mbini
ng P
o
iss
on Si
ngu
lar
Int
egra
l
an
d T
o
tal Variati
on Pri
o
r Mode
ls i
n
Image Restor
a
t
ion.
Sign
al Pro
c
essin
g
. 201
4; 103(
10): 29
6-3
08.
[2]
Stanle
y
J
Ree
v
es. Image R
e
storati
on: F
u
n
d
a
menta
l
s of
Image Restorati
o
n.
Acade
mic P
r
ess Li
brary
in Sig
nal Proc
e
ssing
. 20
14; 4: 165-
192.
[3]
Moez Kall
el, R
a
ja
e Abou
laic
h
,
Abderrahm
an
e Hab
bal, Mah
e
r Moakh
e
r. A Nash-gam
e Appro
a
ch to
Joint Image R
e
storatio
n an
d
Segmentati
o
n
.
Applie
d Math
ematica
l
Mode
llin
g
. 201
4; 38
(11): 303
8-
305
3.
[4]
Richar
d F
r
ans,
Yo
yon
g
Arfia
d
i. Sizi
ng, Sh
a
pe,
a
nd T
opol
og
y Optimiz
a
ti
ons of
Roof T
r
usses Us
in
g
H
y
brid Ge
netic
Algorithms.
Pr
oced
ia En
gin
e
e
rin
g
. 201
4; 95
: 185-19
5.
[5]
Umera F
a
ro
oq
, Muhummad
T
a
riq Siddiq
u
e
.
A Com
parati
v
e Stud
y
on U
s
er Inte
rfaces of Interactiv
e
Genetic Alg
o
rit
h
m.
Procedi
a Co
mp
uter Scie
nce
. 201
4; 32: 45-5
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Resto
r
ation Base
d o
n
Hybrid Ant
Colo
ny Algo
ri
thm
(Yan Fen
g
)
1304
[6]
Z
hai Xu
emi
ng, Z
hang Do
ng
ya
,
De
w
e
n
W
a
n
g
.
F
eatur
e E
x
tra
c
tion a
nd
Clas
s
ificatio
n of El
e
c
tric Po
w
e
r
Equi
pment
Images
Bas
ed
o
n
C
o
rner
Inva
riant M
o
ments
.
T
E
LKOMNIKA
Indo
nes
ian
Journ
a
l of
Electrical E
ngi
neer
ing
. 2
012;
10(5): 10
51-
10
56.
[7]
Ren
bo L
uo, W
enzhi L
i
ao,
Yougu
o Pi.
Discr
imin
ativ
e Superv
i
se
d
Neig
hb
orho
o
d
Preservi
ng
Embed
din
g
F
e
ature E
x
tractio
n
for H
y
p
e
r s
pectral
imag
e
Classific
a
tio
n
.
T
E
LKOMNIKA Indon
esi
a
n
Journ
a
l of Elec
trical Eng
i
ne
eri
n
g
. 201
4; 12(6
)
: 4200-4
2
0
5
.
[8]
A Bouh
ami
d
i,
R Enkh
bat, K J
b
ilo
u. Co
nd
itio
nal Gr
a
d
ie
nt T
i
khon
ov Meth
o
d
for A C
onve
x
Optimizatio
n
Probl
em in Im
age R
e
storati
o
n.
Journ
a
l of
Co
mp
utation
a
l
and Ap
pli
ed
Mathe
m
atics
. 201
4;
255(
1):
580-
592.
[9]
Xi
ao-Gu
an
g Lv
, Yong-Z
h
ong
Song, S
hun-
Xu W
ang,
J
i
an
g
Le. Imag
e R
e
storation
w
i
th
A Hig
h-ord
e
r
T
o
tal Variation
Minimiz
a
tio
n
Method.
Ap
pli
ed Mathe
m
atic
al Mode
lli
ng
. 20
1
3
; 37(16): 8
210
-822
4.
[10]
Miros
ł
a
w
T
o
mera. Ant Colony
Opti
m
i
zati
o
n
Alg
o
rithm A
ppli
ed to S
h
i
p
Steerin
g Co
n
t
rol.
Proced
ia
Co
mp
uter Scie
nce
. 201
4; 35: 83-9
2
.
[11]
T
i
anjun Li
ao, T
homas Stützl
e, Marco A.
Montes de Oca, Marc
o Dori
go. A Unified
Ant Colo
n
y
Optimization Algorit
hm for
Continuous Optimization.
Euro
pea
n J
ourn
a
l
of Operati
o
n
a
l
Res
earch
.
201
4; 234(
3): 597-6
09.
[12]
RM Rizk-A
lla
h,
Elsa
ye
d
M Za
ki, Ahme
d A
h
med E
l
-Sa
w
y.
H
y
brid
izin
g A
n
t Col
o
n
y
O
p
ti
mizatio
n
w
i
t
h
Firefly
A
l
gorithm for Unconstr
ain
ed Optimiz
a
tion Pro
b
lems.
Appl
ied
Math
ematics a
nd
C
o
mputati
o
n
.
201
3; 224(
1): 473-4
83.
[13]
Pour
ya
Hose
in
i
,
Mahrokh G. S
h
a
y
este
h. Effici
ent Co
ntrast E
nha
nceme
n
t of
Images Us
in
g
H
y
brid
Ant
Colo
n
y
Optimi
sation, Ge
netic
Algor
ithm, a
n
d
Simu
late
d A
nne
ali
ng.
D
i
git
a
l Si
gn
al Pr
oc
essin
g
. 20
13;
23(3): 87
9-8
9
3
.
[14]
M Verda
g
u
e
r,
N Cl
ara, O Gu
tiérrez, M Poc
h
. App
licati
on
of An
t-Co
lon
y
-
O
ptimizatio
n A
l
gorit
hm fo
r
Improved M
a
n
agem
ent of F
i
r
s
t F
l
ush Effe
cts in Urb
an W
a
ste
w
a
t
er S
y
ste
m
s.
Science
o
f
the T
o
tal
En
vi
ronm
e
n
t
. 2
014; 48
5(1): 14
3-15
2.
[15]
Ali Ro
ozb
eh N
i
a
, Mohamm
ad
Hemmati F
a
r, Se
y
ed T
aghi A
k
hava
n
Ni
aki.
A F
u
zz
y
Ve
nd
or Mana
ge
d
Inventor
y of Multi-Item Econo
mic Or
der Quantit
y
Mod
e
l u
n
der Shorta
ge: An Ant Colo
n
y
Optimizatio
n
Algorit
hm.
Internatio
nal J
ourn
a
l of Producti
o
n
Econo
mics
. 201
4; 155(
9): 259-2
71.
[16]
Rachi
d
He
dj
a
m
, Mohamed
Cheri
e
t. Histor
i
cal
D
o
cume
nt Image Resto
r
ation Us
ing
Multisp
e
ctra
l
Imaging Sy
stem.
Pattern Recogn
ition
. 2
013;
46(8): 229
7-2
312.
Evaluation Warning : The document was created with Spire.PDF for Python.