TELKOM
NIKA
, Vol.12, No
.4, Dece
mbe
r
2014, pp. 87
5~8
8
2
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v12i4.534
875
Re
cei
v
ed Au
gust 26, 20
14
; Revi
sed O
c
t
ober 2
9
, 201
4; Acce
pted
No
vem
ber 1
5
,
2014
Image F
u
zzy Enhancement Based on Self-Adaptive Bee
Colony Algorithm
Meng Lei*
1
, Yao Fan
2
Coll
eg
e of Information En
gi
ne
erin
g, T
i
bet Universit
y
for Nati
ona
lities,
Xi
an Ya
ng 7
1
2
082, Sha
n
x
i, C
h
in
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 1502
09
814
@
qq.com
1
, fan
y
a
o11
24
198
3@1
63.com
2
A
b
st
r
a
ct
In the i
m
a
ge
a
c
quisiti
on
or tra
n
smissio
n
, the
imag
e may
b
e
da
ma
ged
an
d distorted du
e
to
vari
ous
reaso
n
s; theref
ore, in
order t
o
sati
sfy peo
pl
e
’
s visua
l
effects, these i
m
a
ges
w
i
th degra
d
i
n
g qu
ality
must
be
process
ed to
me
et practic
a
l
nee
ds.
Integr
ating
artifici
al
bee c
o
l
ony
alg
o
rith
m a
nd fu
zz
y
s
e
t, this pa
p
e
r
introd
uces fu
zz
y
e
n
tropy i
n
t
o
the
se
lf-ada
ptive fu
zz
y
e
n
hanc
e
m
ent of
im
ag
e so as t
o
real
i
z
e
th
e
self-
ada
ptive p
a
ra
meter s
e
lecti
o
n. In the me
a
n
w
h
ile, bas
ed
on the exp
o
nenti
a
l pro
pert
i
es of infor
m
a
t
ion
incre
a
se, it pro
poses
a n
e
w
definiti
on of fu
zz
y
e
n
tropy
an
d
uses artific
i
al
bee c
o
lo
ny a
l
g
o
rith
m to re
ali
z
e
the self-ad
apti
v
e contrast en
hanc
e
m
ent u
n
der the maxi
mum e
n
tropy cri
t
erion. T
he ex
peri
m
e
n
tal res
u
lt
show
s that the
met
hod
pro
p
o
s
ed i
n
this
pa
p
e
r can
in
cr
eas
e the
dyna
mic
rang
e co
mpres
s
ion
of the i
m
a
ge,
enh
anc
e the v
i
sual
effects of
the i
m
a
ge, e
n
h
ance t
he i
m
ag
e det
ails, h
a
ve
so
me c
o
lor fi
d
e
lity ca
pacity
a
n
d
effectively over
come the defic
i
enci
e
s
of traditi
ona
l i
m
ag
e en
hanc
e
m
ent
me
thods.
Ke
y
w
ords
: image e
n
h
ance
m
ent, bee co
lony
algor
ith
m
, fu
zzy set
1. Introduc
tion
Image enh
an
ceme
nt is m
a
inly aimed to improve th
e visual qu
ality of image. Image
enha
ncement
sele
ctively highlight
s the i
n
tere
sting
ch
ara
c
teri
stics
or supp
re
sse
s
(cove
r
s) so
me
unne
ce
ssary
characte
ri
stics in the i
m
age to
make the im
age match visual re
sp
o
n
se
cha
r
a
c
teri
stics a
nd g
e
t a
more
practi
cal imag
e o
r
t
r
an
sform
into
an im
age
m
o
re
suita
b
le f
o
r
human o
r
m
a
chi
ne to perform analytical pro
c
e
ssi
ng
by adding some inform
ation or chan
gi
ng
data. Imag
e e
nhan
cem
ent
doe
sn’t a
naly
z
e th
e
re
a
s
o
n
s
to
imag
e d
e
grad
ation
and
the
pro
c
e
s
se
d
image may not be closed t
o
the original image [1],[2].
After years’
rese
arch, im
a
ge e
nhan
ce
ment
technol
ogy ha
s m
a
d
e
si
gnificant
prog
re
ss
and it
ha
s fo
rmed
multipl
e
theo
reti
cal
algo
rithm
s
by no
w. Accordin
g to
the
differe
nt sp
ace
s
whe
r
e en
han
ceme
nt is located, it can b
e
divi
ded into
the algorithm
base
d
on sp
atial domain
and
the algo
rithm
based on freque
ncy do
main [3]. The
forme
r
alg
o
rithm di
re
ctly operate
s
o
n
th
e
image grayscale while
the
latter co
nducts
certain correction on
the transform
a
tion
coeffici
ent
value of ima
ge within
ce
rtain imag
e tra
n
sfor
mation
domain,
whi
c
h is an in
dire
ct enha
ncem
ent
algorith
m
. It shoul
d be
pointed o
u
t that thes
e traditional ima
ge enh
an
ce
menttech
nolo
g
ies
haven’t con
s
i
dere
d
the fuzzine
s
s of image, on the
co
ntrary, it only simply
chan
ges the contrast
or
sup
p
re
sse
s
n
o
ise of t
he e
n
tire im
age [4]. It
wea
k
e
n
s t
h
e
image
detai
ls in th
e n
o
i
s
e
sup
p
re
ssion;
it inevitably causes
seri
ou
s neg
at
ive effects a
nd it ha
s ce
rtain limit
ations [5].
So far, th
e i
m
age
enh
an
cement
ba
sed
on fu
zzy the
o
ry h
a
s a
c
hie
v
ed si
gnifican
t
re
sults
and the
main
advantag
e o
f
image fu
zzy
enha
ncemen
t is
that it ca
n preserve th
e imag
e deta
ils
[6]. The pa
ra
meter
sele
cti
on such a
s
membe
r
ship
and e
nha
nce
d
ope
rato
rs i
n
the ima
ge f
u
zzy
enha
ncement
play an im
portant
signif
i
cant o
n
the
enhan
ce
me
nt effects
while artifici
al bee
colo
ny alg
o
rit
h
m ha
s th
e
advantag
es
of simpl
e
co
mputation, e
a
se
to
reali
z
e an
d fe
w
control
para
m
eters.
Con
s
id
erin
g the paralleli
sm of artifi
cial bee colony al
gorithm, a
s
the fitness fun
c
tion
of bee colony
algorithm, th
e new
definiti
on of fuzz
y e
n
tropy propo
sed in this p
a
p
e
r ha
s favora
ble
robu
stne
ss, introdu
ce
d fidelity and enh
ances the
st
ability of the
algorithm a
nd the ability to
maintain
det
ails. Thi
s
pa
per fi
rstly
systemati
c
ally i
n
trodu
ce
s th
e ba
si
c ide
a
, differen
c
e
s
and
appli
c
ation
chara
c
te
risti
c
s of the
co
mm
on meth
od
s
of image
en
h
ancement. T
hen, it inte
grates
artificial
bee
colo
ny alg
o
rit
h
m a
nd fu
zzy set
s
and
introdu
ce
s fu
zzy
entropy
into
the self-a
dapt
ive
fuzzy e
nha
ncement of im
age
so a
s
to reali
z
e
th
e
self-a
daptiv
e paramete
r
sele
ction. In
the
meanwhile,
it rai
s
es a new
definition of fuzzy
en
tropy b
a
sed
on the
in
dici
al respon
se
of
informatio
n in
cre
a
se a
nd it
reali
z
e
s
the
self
-a
daptive contrast enh
ancem
ent of
image by
usi
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No.4, Decemb
er 201
4: 875
– 882
876
artificial be
e colo
ny algorit
hm in the maximum ent
ropy crite
r
ion.
Finally, it realize
s
the se
lf-
adaptive fuzzy enhancement of image th
rough sim
u
lat
i
on experi
m
ents.
2. Image Enhanceme
n
t
Techniqu
es
Image e
nhan
ceme
nt is
a
method to
hig
h
light some i
n
formatio
n in
an ima
ge a
n
d
wea
k
e
n
or get ri
d of some u
nne
ce
ssary information a
c
cording to som
e
spe
c
ific
re
quire
ment
s. Its
purp
o
se i
s
t
o
en
han
ce
the
cla
r
ity an
d contrast
of imag
e in
certain
spe
c
ific a
ppli
c
ation
s
t
o
improve th
e i
m
age
quality
and m
a
ke th
e processe
d
results m
o
re
con
s
i
s
tent wit
h
hum
an visu
a
l
sen
s
o
r
y syste
m
or ea
sier to
be recogni
ze
d by machin
e
s
[7].
The
cu
rre
nt common
-
u
s
ed
enh
an
ceme
nt tech
niqu
e i
s
divid
ed i
n
to
tech
niqu
e b
a
se
d o
n
spatial do
mai
n
and tech
ni
que ba
sed o
n
transfo
rmat
ion domai
n.The forme
r
techni
que di
re
ctly
pro
c
e
s
ses in
the spa
c
e of the image wh
ile the la
tter pro
c
e
s
ses in
the transfo
rm
ation domai
n of
the image.T
h
e co
mmon
-
u
s
ed tra
n
sfo
r
m
a
tion spa
c
e i
s
the frequ
en
cy dom
ain
sp
ace, n
a
mely t
h
e
Fouri
e
r t
r
an
sf
orm. T
he
en
han
ceme
nt
method
s b
a
sed o
n
spatial
domai
n in
clu
de: the
gray
scale
transfo
rmatio
n to e
nha
nce imag
e th
ro
ugh
per pi
xe
l point
s, the
histo
g
ram transfo
rmatio
n
to
cha
nge th
e i
m
age
cont
ra
st glob
ally or locally
a
nd
the sp
atial transfo
rmatio
n
to pro
c
e
s
s the
neigh
borhoo
d
pixels of image thro
ugh
template or
masking [8]. Figure 1 de
monst
r
ate
s
two
comm
on tran
sform
a
tion fu
nction
s of sp
atial-do
main i
m
age en
han
cement.
Figure 1. Tra
n
sformation f
uncti
o
n
s of contra
st enha
n
c
eme
n
t
The e
nha
nce
m
ent of fre
q
uen
cy dom
ai
n spac
e i
s
realized th
rou
gh different f
r
equ
en
cy
comp
one
nts i
n
the image.
The image frequ
en
cy sp
ectru
m
gives global ch
ara
c
teri
stics of the
image; the
r
ef
ore, the
fre
q
u
ency-dom
ain
enha
ncement
is
not impl
e
m
ented
pe
r p
i
xel and it i
s
not
as direct a
s
the sp
atial-d
o
m
ain enh
an
cemen
t. The freque
ncy-dom
ain enha
nce
m
ent is reali
z
ed
throug
h the filter and the f
r
eque
nc
y filtered by differe
nt filters
and the
re
serve
d
freque
ncy
diff
er
from ea
ch oth
e
r; therefo
r
e,
it c
an get different en
han
cement effect
s.
3. Artificial Bee Colony
Algorithm
3.1 The Principleof Ar
tifi
cial Bee Colon
y
Algorith
m
The hon
ey-collectin
g pro
c
ess of the bee (nam
ely to
find high-qual
ity honey sou
r
ce
s) is
simila
r to the pro
c
ess to
sea
r
ch the o
p
timal solutio
n
to the pro
b
lem to be o
p
timized in t
he
evolutiona
ry comp
utation.
T
he hon
ey colle
ction i
s
reali
z
ed th
ro
ugh the
com
m
unication, the
transfo
rmatio
n
and
the col
l
aboration a
m
ong differe
nt
bee
s. Th
e
pro
c
e
s
s for the be
e
colo
n
y
to
colle
ct hon
ey includ
es th
ree ba
sic
pa
rts an
d tw
o b
a
s
ic
behavio
rs. The thre
e p
a
rts a
r
e: foo
d
s,
employed
be
es a
nd
unem
ployed b
e
e
s
and the
two
b
ehaviors
are
to re
cruit an
d
aban
don
ce
rt
ain
foods [9].
The e
s
sen
c
e
of artifici
al b
ee
colonyal
g
o
rith
m i
s
to
search
optimal
solutio
n
th
ro
ugh the
rand
om b
u
t
targeted
evol
ution on
the
gro
up
fo
rm
ed by the
candid
a
te sol
u
tions. In
every
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Fuzzy
Enhan
cem
e
n
t
Based on S
e
lf-Adapt
i
v
e
Bee Colo
ny A
l
gorithm
(Men
g Lei)
877
circulatio
n, the numb
e
rs o
f
leaders a
n
d
follo
we
rs
are the same
and the
r
e i
s
only one o
r
no
scouter.
The
solutio
n
evol
ution i
s
com
p
leted
by
the
above-mentio
ned th
re
e
kin
d
s
of b
e
e
s
: (1)
Employed be
e co
ndu
cts l
o
cal
se
arch i
n
the nei
g
h
b
o
rho
od d
o
ma
in of its corresp
ondi
ng fo
ods
and u
pdate
s
i
t
s food
s
whe
n
finding
ne
w foods optima
l
to the curre
n
t foods;
(2
)
Acco
rdi
ng to
the
food info
rmat
ion p
r
ovide
d
by the
emplo
y
ed be
e,
the
followe
r ch
oo
se
s
the
foo
d
throug
h ce
rta
i
n
sele
ction
met
hod; m
a
kes l
o
cal
sea
r
ch
near the
sele
cted fo
od
so
urces;
inform
s
co
rre
sp
ondi
ng
employed
be
e to the current food
s an
d upd
ates
th
e food
s whe
n
finding m
o
re excell
ent n
e
w
foods. Wh
en
the
follo
we
r cho
o
ses
the foods, excell
ent food
s
(so
l
utions with
h
i
gh fitne
s
s)
can
attract mo
re followers. With several se
a
r
ch
es in the
n
e
ighb
orh
ood
domain a
nd these food
s h
a
ve
more
evolutio
n opp
ortu
nities. (3
) In the
stagnati
on
of the termi
natio
n sol
u
tion of
scouter,
nam
ely
whe
n
the sol
u
tion evolutio
n stagn
ates,
the
un-e
m
plo
y
ed bee aba
ndon
s the cu
rre
nt foods a
n
d
become
s
a
scoute
r
. Th
en
it rand
omly
searche
s
and
gene
rate
s a
f
easi
b
le
sol
u
tion a
s
a n
e
w food
and conveys the relevant
information t
o
the
employ
ed bee. Th
ro
ugh the colla
boratio
n of the
above-mentio
ned th
ree
ki
nds
of be
es,
ABC algo
rit
h
m gradu
ally conve
r
g
e
s
and o
b
tain
s
the
optimal sol
u
tion or ap
proximate optimal
solu
tio
n
in the feasibl
e
sol
u
tion sp
ace [10].
3.2 Math
ema
t
ical Des
c
rip
t
ionof
Ar
tific
i
al Bee Colo
n
y
Algorithm
Con
s
id
er opti
m
ization p
r
o
b
l
em(P):
is the obje
c
tive optimizati
on functi
o
n
;
is the variabl
e
to be optimized
;
is the solutio
n
spa
c
e a
nd
.
The
set form
ed by the fe
a
s
ible
sol
u
tion
s to
Problem (P) can be
a
b
s
tra
c
ted as
th
e
food
s
of a bee
co
lony. The p
o
sition
(fea
si
ble soluti
on
) of every e
m
ployed b
e
e
in the col
o
n
y
c
o
rres
ponds
to a food, whic
h is
determined
by t
he func
tion value
of the objec
t
ive func
tion and
the num
be
r o
f
employed
b
ees and
follo
wers i
s
th
e
same a
s
th
at
of food
s (sol
u
t
ions).
The
r
ef
ore,
the positio
n o
f
a certain foo
d
c
an b
e
expressed
with the vector
.
Firstly, initialize
with ABC
algorith
m
. Ra
ndomly ge
ne
rate an i
n
itial
populatio
n with
solutio
n
s
according to
Formul
a (3);
every solut
i
on is
and
is a d-
d
i
me
ns
io
na
l ve
c
t
o
r
.
(3)
Then, the be
es be
gin to
con
d
u
c
t cycli
c
se
arch fo
r the foods a
nd the cy
cle
time is
until it rea
c
he
s to the
spe
c
i
f
ied pre
c
i
s
ion
or the m
a
ximum iteratio
n
s
MAX _Gen. The empl
oye
d
bee sea
r
ches
co
rre
sp
o
nding foo
d
s,
namely ran
domly cho
o
se a
different b
ee
as
a nei
ghb
o
r
an
d rand
o
m
ly cho
o
ses
a dime
nsi
on
as it
s sea
r
ch
guide
directi
on.
The se
arch p
r
ocess is
con
ducte
d acco
rding to Form
ula (4
) and (5
).
(4)
(5)
Among the two formula
s
,
is the se
a
r
ch di
rectio
n and
step length.
and
are rand
o
m
ly selecte
d
and
is
a
rand
om num
ber am
ong [-1,1]. If
exce
eds the
solut
i
on spa
c
e
ra
nge, tran
sform it into the
boun
dary val
ue acco
rdin
g to Formula
(6
):
mi
n
{
(
)
:
}
d
f
xx
S
R
f
12
(,
,
)
d
X
xx
x
S
mi
n
m
a
x
{(
,
)
|
1
,
2
,
,
}
ii
Sx
x
j
d
12
(,
,
)
ii
i
i
d
X
xx
x
SN
12
(,
,
)
ii
i
i
d
X
xx
x
SN
mi
n
m
ax
mi
n
(0
,
1
)
(
)
1,
2
,
,
,
1,
2
,
,
ij
j
j
j
x
x
rand
x
x
iS
N
j
d
(
1
,
2
,
...,
_
)
Gen
G
en
MA
X
G
en
(.
)
ij
v
r
x
x
ne
ig
hb
ou
r
j
new
ii
ii
x
xv
v
{1
,
2
,
}
m
e
igh
bou
r
S
N
,{
1
,
2
,
}
nei
g
hbo
ur
i
j
d
r
mew
ij
x
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No.4, Decemb
er 201
4: 875
– 882
878
(6)
If the quality (fitness) of the
sea
r
che
d
fo
od (s
ol
ution)
is bette
r than
the cu
rrent fo
od,
repla
c
e
the
n
e
w fo
od
with t
he
cu
rre
nt fo
od; othe
rwise
,
kee
p
the
foo
d
un
ch
ang
ed.
After the
se
arch
of all em
ploy
ed b
e
e
s
, the
y
go b
a
ck to
the da
nci
ng
area
in th
e h
oney
comb
a
nd
sha
r
e th
e
food
informatio
n with the unem
ployed bee
s
in the co
mb
throug
h wa
g
g
le dan
ce an
d the followe
rs
judge
the
ret
u
rn
rate
of e
v
ery food
according
to th
e information
obtain
ed
an
d colle
ct ho
n
e
y
throug
h ro
ul
ette wheel
selectio
n.The
return rate
i
s
expresse
d
with the fitness value
of the
solutio
n
and t
he fitness an
d sele
ction p
o
ssibility are
comp
uted a
c
cording to Fo
rmula (7) a
n
d
(8).
(7)
(8)
Figure 2. ABC Algorithm fl
ow chart
mi
n
m
i
n
ma
x
m
a
x
,
,
ne
w
ij
ij
ij
ne
w
ij
ne
w
i
j
ij
ij
x
xx
x
x
xx
ne
w
i
x
1
,0
1
1|
|
,
0
i
i
i
ii
f
f
fit
ff
1
i
i
SN
j
j
f
it
p
f
it
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Fuzzy
Enhan
cem
e
n
t
Based on S
e
lf-Adapt
i
v
e
Bee Colo
ny A
l
gorithm
(Men
g Lei)
879
In the above
two formula
s
,
is the func
tion valu
e
of the ith solution;
is its
corre
s
p
ondin
g
fitness valu
e and
is the n
u
mbe
r
of solu
tions. Obviou
sly, base
d
on
the roulette
whe
e
l sele
ction, goo
d foo
d
s
can
attra
c
t more
followe
rs
and it
ha
s
highe
r po
ssib
ility for evolution
and it ca
n a
c
cele
rate the
conve
r
ge
nce
rate of al
g
o
rit
h
m. After the followe
rs
ch
oose the foo
d
s,
sea
r
ch the ne
ighbo
rho
od d
o
main of the food
s acco
rdi
ng to Formul
a (4) a
nd (5
); adju
s
t the foods
according to
Form
ula (6); cond
uct g
r
eedy sele
ction to the
n
e
wly-sea
r
che
d
po
sition
s
and
maintain the
better solutions
[11],[12].
If a food stagnate
s
in a certai
n po
sition for more times than th
e pre
-
set time limit, it
demon
strates that this sol
u
tion is trap
p
ed in the local optimal sol
u
tion and the
corre
s
po
ndi
ng
honey-coll
e
cti
ng bee be
co
me a scouter and it
abandon
s the food and ran
d
o
m
ly generate
s
a
new foo
d
(sol
ution) to repl
ace the ab
an
done
d f
ood (solutio
n) in Space
acco
rdi
ng to Formul
a
(3). Th
e limit
here i
s
the
bo
unda
ry pa
ra
meter to
ju
dg
e wh
ether
a certain
solutio
n
jump
s out from
the curre
n
t stagnatio
n stat
us [13].
The flow
cha
r
t of ABC Algorithm is sho
w
n as Figu
re 2.
4. Conduc
t Image Fuzz
y
Enhanceme
n
t
w
i
th
Artifi
cial Bee Colon
y
Algorith
m
4.1 Defini
tion of Fuzz
y
Entropy
Fuzzy ent
rop
y
quantificati
onally reflect
s
the
fuzzy
degree
of a
n
imag
e a
n
d
is the
averag
e difficulty level to
determi
ne
whether a
pix
e
l ca
n b
e
se
en a
s
a
n
ele
m
ent of a fu
zzy
s
u
bs
et.
(i) Acco
rdin
g
to different enha
ncement
pur
p
o
ses a
nd image
s,
set the mem
bership
para
m
eters
in Form
ula (9
); the plan
e forme
d
by all
is the fuzzy
cha
r
a
c
teri
stic
plane;
is th
e
maximum pix
e
l value;
and
a
r
e
expon
e
n
tial an
d reci
pro
c
al fu
zzy f
a
ctors
and th
eir val
ues will
di
re
ctly affect the
fuzzi
ne
ss
of t
he fu
zzy
characteri
stic pl
a
ne. Th
erefo
r
e
,
in
fuzzy
enh
an
cement
pro
c
e
ssi
ng, to
ch
o
o
se
go
od fu
zzy pa
ram
e
ters
and
is an
i
m
porta
nt
step to get a satisfa
c
tory enha
nced im
age.
A parti
cu
lar gray
scale
meeting
is
calle
d cro
sso
ver point. Th
e sele
ction of
fuzzy pa
ram
e
ter is
relate
d to the sele
ction of cro
s
so
ver
point
and the
cro
s
sover p
o
i
nt meets the
followin
g
req
u
irem
ents:
Therefore,
after dete
r
mini
ng the
cro
ssover p
o
int
,
can be
dete
r
mined th
rou
g
h
Formul
a (9
) whe
n
is determined.
(ii) Tran
sform the imag
e from the spat
ial d
o
ma
in to the fuzzy dom
ain
through
Tran
sfo
r
mati
on ;
(9)
(iii) Modify the membe
r
shi
p
: through the followin
g
tran
sform
a
tion
, namely
the reg
r
e
ssi
o
n
of the fuzzy
enhan
cem
e
n
t
operato
r
(INT);
(10)
The
key of fu
zzy
enh
an
ce
ment is to u
s
e t
he fu
zzy e
nhan
cem
ent
operator to re
duce the
membe
r
ship
value sm
aller than 0.5 by increa
si
ng th
e
membe
r
ship
value
bigg
er than 0.5 so
as to reduce
the fuzziness of
. The fuzzy enha
ncement opeartor generat
es another fuzzy set
in the fuz
z
y
set
.
i
f
i
f
it
SN
S
ma
x
(,
,
)
ed
FF
g
mn
ma
x
g
e
F
d
F
e
F
d
F
()
0
.
5
mn
C
Gg
C
g
0.
5
0.
5
0.
5
mn
C
mn
m
n
C
mn
C
g
g
Gg
g
g
g
C
g
d
F
e
F
G
ma
x
()
[
1
]
e
F
mn
mn
mn
d
gg
Gg
F
()
mn
m
n
2
2
00
.
5
2[
]
()
0.
5
1
12
[
1
]
mn
mn
mn
mn
mn
T
mn
G
G
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No.4, Decemb
er 201
4: 875
– 882
880
(11)
In the form
ul
a,
is d
e
fined
as the
multip
le calli
ng of
. In the extre
m
e case, wh
en
,
gene
rate
s a two-g
r
ay
scale (two
-valu
e
) imag
e. In orde
r to avoi
d the loss of detail
informatio
n a
nd the
defi
c
ie
ncy of fu
zzy i
m
age
enh
an
cement,
cho
o
s
e
s
1,
2 a
nd
3 a
c
cordi
ng t
o
different enh
a
n
cem
ent pu
rp
ose
s
an
d ima
ges.
(iv) New gra
y
scal
e
is ge
nerate
d
thro
ugh inve
rse
tran
sform
a
tion
so
as t
o
transfo
rm the
data from fuzzy domain to
the spatial d
o
m
ain of the image:
(12)
4.2 Choo
se
The Optio
n
al
Fuzzy
Parameter By
Using Bee
Colo
n
y
Algorithm
Con
s
id
erin
g the expone
ntial prop
ertie
s
of
information increa
se,
we have p
r
o
posed a
new
definitio
n of fuzzy ent
ropy to
cond
u
c
t self
-ada
ptive fuzzy e
n
h
ancem
ent of t
he imag
e ba
sed
on the above
-
mentione
d an
alysis a
nd its
definition is a
s
follows:
(13)
and,
(14)
In these two formula
s
,
is
the fuzzy
set
,
is the numb
e
r of the sub
s
et
o
f
fuzzy set
.
is the
partition method
of
fu
zzy
d
o
main
and unifo
rm
partition o
r
n
on-u
n
iform
partition
ca
n
be
cho
s
e
n
according
to
different i
m
age
s.
is the
memb
ership
of imag
e
grayscal
e val
ue an
d
is th
e freq
uen
cy
of image
gra
y
scal
e
value.
is the
sum o
f
th
e
freque
nci
e
s
of the
spatial
-
dom
ain
gray
scale val
u
e
s
whe
n
the
sp
atial-do
main
pixel poi
nt
is
mappe
d onto
the fuzzy su
bset
throu
ght
he mem
bership fun
c
tion
. It can
sho
w
th
at whe
n
, the fuzzy e
n
tropy
can
a
m
ount to th
e
maximum v
a
lue; the
r
efore, in the
part
i
tion
method
s of fuzzy do
main,
should b
e
mad
e
in
.
Based
on
the
above
resea
r
ch,
the
self-adapt
ive im
a
ge e
nha
ncem
ent of ABC al
gorithm
can b
e
reali
z
ed throu
gh th
e followin
g
st
eps.
(a)
Tran
sfo
r
m the image from
the grayscal
e domain to t
he fuzzy dom
ain within the
value rang
e
of the pa
ram
e
ters;
com
put
e the fu
zzy e
n
tropie
s
of
different parameters res
p
ec
tively to mak
e
the param
eter sel
e
ctio
n method to maximize
fu
zzy entropy is the optimal param
ete
r
sele
ction met
hod an
d re
co
rd the pa
ram
e
ter and fu
zzy entropy.
(b)
Use the dete
r
mine
d para
m
et
ers to tra
n
sform the image from th
e grayscal
e domain to th
e
fuzzy do
main
and co
ndu
ct fuzzy en
han
cement.
(c)
Tran
sfo
r
m th
e data f
r
om
the fuzzy do
main to th
e
spatial
dom
ai
n of the i
m
a
ge
so a
s
to
reali
z
e the
sele
ction of
self-a
daptive par
ameters, namely the self
-ada
p
t
ive fuzzy
enha
ncement
.
5. Simulation Experimen
t
and Result
Analy
s
is
The
self-ada
p
t
ive fuzzy
en
h
ancement
alg
o
rith
m
propo
sed in
thi
s
p
a
p
e
r i
s
re
alized
unde
r
the maximum
fuzzy e
n
trop
y criteri
on; th
erefo
r
e, t
he
selectio
n of op
timal fuzzy p
a
ram
e
ters is t
he
para
m
eter o
p
t
imization un
der the maxi
mum fuzzy
entropy in e
s
sence and
it can dire
ctly u
s
e
fuzzy e
n
tropy
as fitne
ss fu
nction
and it
adopt
s the
n
e
w d
e
finition
of fuzzy e
n
tropy pro
p
o
s
e
d
in
this pa
per, a
s
dem
on
strat
ed in Fo
rmul
a (12
)
a
nd (13). In o
r
de
r to redu
ce
a
s
mu
ch p
r
o
g
r
am
()
(
1
)
()
(
(
)
)
,
1
,
2
,
,
rr
m
n
mn
mn
TT
T
r
()
r
T
T
r
()
r
T
r
mn
g
1
G
1
1
ma
x
()
()
1
e
F
mn
m
n
d
m
n
gG
g
F
{1
(
)
}
(
)
1
1
(,
,
,
)
[
(
)
{
1
(
)
}
]
Pi
P
i
N
PA
PA
AP
i
P
i
i
KA
N
M
P
A
e
P
A
e
N
()
()
(
)
Ai
Pi
xA
P
AP
x
A
N
1
,,
N
A
A
A
M
()
A
x
()
P
x
()
P
i
P
A
x
i
A
()
A
()
0
.
5
Pi
PA
()
0
.
5
Pi
PA
M
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Im
age Fuzzy
Enhan
cem
e
n
t
Based on S
e
lf-Adapt
i
v
e
Bee Colo
ny A
l
gorithm
(Men
g Lei)
881
runni
ng time as po
ssible, it stabilizes wh
en the
popul
a
t
ion size is 30
and the term
ination alge
bra
is 10
0. The i
n
itial value of
fuzzy p
a
ra
m
e
ter
in bee
colony algo
rith
m is ge
ne
rat
ed ra
ndo
mly.
Therefore, we cho
o
se the terminatio
n al
gebraof 100.
Figure 3 is t
he co
ntra
st
cha
r
t betwee
n
the
ori
g
ina
l
image an
d
the algo
rithm
of this
pape
r. Table
1 is the avera
ge va
lue, sta
ndard deviati
on and e
n
tro
p
y of original
image, PSO and
algorith
m
of this pa
per. Fi
g.4 is the hi
stogram
of the avera
ge
value, stand
a
r
d deviation
and
entropy of ori
g
inal imag
e a
nd algo
rithm
of this pape
r.
Figure 3. Orig
inal image a
n
d
result of algorithm of this
pape
r
Table 1 Average Value, Standa
rd Devia
t
ion and En
tropy of Origin
al Image, PSO and Algo
rithm
of This Pape
r
Average Value
Standard D
e
viation
Entrop
y
Original Image
50.2849
43.7312
7.3471
PSO 151.4758
82.8231
6.8263
Algorithm of This Algorithm
121.2631
51.8472
8.3489
Figure 4. Hist
ogra
m
of average,
st
anda
rd deviation a
nd entro
py
The avera
g
e
value
in
cre
a
se
s after hi
stogram
equ
alizatio
n, de
monst
r
ating
that th
e
brightn
e
ss is
high and the
stand
ard d
e
viation is
small
and reflectin
g
that the equalization effect
is not g
ood e
noug
h. It can
be seen fro
m
the ex
pe
ri
mental data t
hat avera
ge
value an
d m
ean
squ
a
re d
e
via
t
ion increa
se
after being
proc
esse
d by PSO algorithm; howe
v
er, the entropy
decrea
s
e
s
; th
e bri
ghtne
ss i
n
crea
se
s a
n
d
the def
initio
n
be
come
s
ba
d. It can
be
seen th
at after
d
F
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No.4, Decemb
er 201
4: 875
– 882
882
PSO pro
c
e
s
s, obviou
s
color di
stortio
n
appe
ar
s a
nd definition
decrea
s
e
s
. The artifici
al
bee
colo
ny alg
o
rit
h
m h
a
s bette
r dyn
a
mic ra
nge
co
mpression
an
d d
e
finition e
nha
ncement
as wel
l
as
color fidelity
ability. And we can find that it
can enhance the dynamic
ra
nge compression of the
image, the v
i
sual effe
cts
of the image
and t
he im
age det
ails
and ha
s
ce
rtain color fid
e
lity
cap
a
cit
y
.
6. Conclusio
n
Image en
han
ceme
nt is th
e ba
sic te
ch
nique
of digi
t
a
l image
pro
c
e
ssi
ng a
nd
it can effe
ctively
improve th
e g
l
obal o
r
lo
cal
cha
r
a
c
teri
stics of t
he ima
g
e
.By using th
e glob
al opti
m
ization
ca
p
a
city
and pa
ralleli
sm of ABC algorithm, this p
aper p
r
o
p
o
s
e
s
a ne
w defin
ition of fuzzy entropy a
s
th
e
fitness fu
ncti
on of bee
colony algo
rit
h
m, whi
c
h
can autom
atically sea
r
ch the optimal f
u
zzy
parameters, i
m
prove the
stability of
the algorithm and the ability to maintain
det
ails, reali
z
e t
h
e
self-a
daptive fuzzy
en
han
cement
of the image, have
better colo
r
fidelity capa
city and impro
v
e
the image qu
ality for appro
x
imate real-ti
m
e appli
c
atio
ns.
Referen
ces
[1]
Kamran B
i
n
a
e
e
, Reza
PR.
Hasa
nzad
eh.
An ultras
ou
nd
imag
e e
n
h
a
n
c
ement meth
o
d
usi
ng
loca
l
grad
ient b
a
sed
fuzz
y
simil
a
rit
y
.
Biomedic
a
l Si
gna
l Processi
n
g
and C
ontro
l
. 201
4; 13(1): 89
-101.
[2]
W
enda
Z
hao,
Z
h
iju
n
X
u, Ji
an
Z
hao, F
a
n
Z
h
a
o
, Xi
zh
en Han.
Variati
o
n
a
l inf
r
ared ima
ge e
nha
nceme
n
t
base
d
on
ad
ap
tive du
al-thres
hol
d
gra
d
ie
nt field
equ
aliz
atio
n.
Infrared Phy
s
ics & T
e
chn
o
l
ogy
. 20
14; 6
6
:
152-
159.
[3]
T
a
malika Cha
i
ra. A rank ord
e
red filter fo
r
medic
a
l ima
g
e
edge
enh
anc
ement an
d det
ection us
in
g
intuitionistic fuzzy
set.
Appl
ied Soft
Computin
g
. 2012; 1
2
(4):
125
9-12
66.
[4
]
Al
ex
F. de
Arau
j
o
, Ch
ri
sto
s
E. C
o
n
s
tan
t
i
n
ou
, Jo
ão
Ma
nu
el
R.S. T
a
vares. Ne
w
artificia
l
l
i
f
e mod
e
l for
imag
e enh
anc
ement.
Expert Systems w
i
th Appl
icatio
ns
. 2
014; 41(
13): 58
92-5
906.
[5] Joseph
Sure
s
h
Pa
ul, J
o
shi
n
Joh
n
Math
e
w
, Cha
ndras
ekh
a
r Kes
a
vad
a
s.
MR im
ag
e e
nha
nceme
n
t
usin
g an e
x
te
n
ded n
e
ig
hb
orh
ood filter.
Jour
nal of Visu
al C
o
mmunic
a
tio
n
and Imag
e Re
prese
n
tatio
n
.
201
4; 25(7): 16
04-1
615.
[6]
Kuld
eep
Si
ngh
, Rajiv
Ka
poo
r. Image
enh
a
n
ceme
nt via
Medi
an-Me
an
Based
Sub-Im
age-
Cli
ppe
d
Histogr
am Eq
ualiz
atio
n.
Optik-Internati
ona
l
Journ
a
l for L
i
ght a
nd El
ectron Optics
. 20
14; 12
5(17)
:
464
6-46
51.
[7]
P. Balasu
bram
ani
am, VP. Ananthi. Imag
e
fusio
n
usi
ng i
n
tuitio
nistic fuzz
y sets.
Information Fusion
.
201
4; 20(1): 21
-30.
[8]
Asmatull
ah
Ch
aud
hr
y
,
Asiful
l
ah K
han,
etc.
Neur
o fuzz
y
a
nd
punct
ual
kr
igin
g
bas
ed fi
lter for im
ag
e
restoratio
n.
Applie
d Soft Co
mputin
g
. 201
3; 13(2): 817-
83
2.
[9]
R.J. Kuo, Y.D
.
Hua
ng, etc.
Automatic k
e
rnel
c
l
usteri
ng
w
i
th be
e
co
l
o
n
y
optimiz
ati
on alg
o
rithm.
Information Sci
ences
. 20
14; 2
83(1): 10
7-1
2
2
.
[10]
Hsing-
Chih T
s
ai. Integr
atin
g the artificial bee colony
and
be
es algorithm to
fac
e
constrained
optimiz
ation pr
obl
ems.
Information Sci
enc
es
. 2014; 2
58(1
0
): 80-93.
[11] Do
ğ
anA
yd
in, SerdarÖz
yö
n, etc. Artificial bee co
lon
y
a
l
gorithm
w
i
th
d
y
nam
ic pop
ul
ation siz
e
to
combi
ne eco
n
o
m
ic
a
nd emiss
i
on disp
atch pr
obl
em.
Intern
ation
a
l J
our
nal
o
f
Electrica
l
P
o
w
e
r & En
ergy
System
s
. 20
14
; 54(6): 144-1
5
3
.
[12]
F
.
GharehMo
h
a
mmadi, M. S
ani
ee A
bad
eh.
Image ste
g
a
n
a
l
y
sis
usin
g a
bee
col
o
n
y
b
a
sed fe
ature
selecti
on al
gor
i
t
hm.
Engine
eri
ng App
licati
ons
of Artificial Intelli
ge
nce
. 20
14
; 31(3): 35-43.
[13]
B. Kala
y
c
i, Suren
d
ra M. G
upta. Artificial
bee colo
n
y
alg
o
rithm for solvin
g seq
u
e
n
ce de
pen
de
n
t
disass
embl
y l
i
n
e
bal
anci
ng pr
obl
em.
Expert Systems w
i
th Appl
icatio
ns
. 2
013; 40(
18): 72
31-7
241.
Evaluation Warning : The document was created with Spire.PDF for Python.