Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
Vol
.
5
,
No
. 3,
J
une
2
0
1
5
,
pp
. 42
1~
42
8
I
S
SN
: 208
8-8
7
0
8
4
21
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Evalu
a
ti
on of Noise E
x
clusion of Medical Images
using
Hyb
r
idization of
Partical Swarm
Optimization an
d Bivari
at
e
Shrinkage Methods
Shruti B
h
ar
gava
*
, A
j
ay
S
o
m
kuw
ar
**
* Department of
Electronics an
d
Communication Engineering,
Dr.
K.
N.
Modi University
, India
** Departmen
t
o
f
Electron
i
cs
and
Communi
cation
Engin
eering
,
M
ANIT Bhopal, I
ndia
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Dec 17, 2014
Rev
i
sed
Ap
r
21
, 20
15
Accepte
d
May 3, 2015
Denoising of im
ages got
corrupt
ed b
y
add
ition
o
f
noise signa
ls (
g
enera
t
ed
b
y
no single reason) has alway
s
a subject of
interes
t
for resear
chers. This paper
proposes and classifies the efficien
cy
of an algorithm based on bivariate
shrinkage furth
e
r optim
iz
ed b
y
Part
icl
e
Swar
m
Optim
ization
(PSO).The
estimator for undecimatedfilterb
ank whic
h incor
porate th
e adap
tive subbands
thresholding fur
t
her r
e
presented
with
singal thr
e
shold based o
n
denosin
g
perform
s
.
The p
a
per ev
alu
a
tes
p
e
rform
ance of
m
e
dical
im
age
denois
i
ng
b
y
calculation of PSNR, MSE, WPSNR and
SSI
M.
The
simula
tion re
sults ba
se
d
on testing
the model at MA
TLAB 2010A
platform shows significant
enhanc
em
ent in
m
itigation of
Gaussian
noise,
speckle no
ise, p
o
isson noise
and salt & pepp
er noises
from ex
perimental d
a
ta.
Keyword:
Discrete Wavelet
Tran
sform
Mean
Squ
a
r
e
Er
ro
r
Peak
Sign
al
to
No
ise
Ratio
Wav
e
let De-noisin
g
Copyright ©
201
5 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Sh
ru
ti Bh
arg
a
va,
Depa
rt
m
e
nt
of
El
ect
roni
cs
an
d C
o
m
m
uni
cati
on
En
gi
nee
r
i
n
g,
Dr.K.N.M
od
i Un
i
v
ersity,
INS-
1,
RII
C
O Ind
u
s
tr
ial Ar
ea Ph
-I
I,
New
a
i,
D
i
stt. Tonk
, R
a
j
a
sth
a
n
–
30
40
21
, Ind
i
a
Em
a
il:b
h
a
rg
ava.shru
ti198
7@g
m
ai
l.co
m
1.
INTRODUCTION
M
e
di
cal
i
n
f
o
r
m
at
i
on, com
p
o
s
ed
of i
m
ages,
phy
si
ol
o
g
i
cal
si
gnal
s
a
n
d
ot
he
r cl
i
n
i
cal
dat
a
,
has
becom
e
an esse
ntial part of a
patient’s
care,
during
screen
i
n
g, in
the d
i
agn
o
s
tic st
age and in t
h
e
treatm
e
nt phas
e
. Over
the pa
st fe
w
decade
s
, t
h
ere
has
bee
n
a
rapid
de
ve
lopment in information t
echnology (IT
) & Medical
In
st
ru
m
e
n
t
atio
n
wh
ich
h
a
s lead
& facilitate
d
th
e g
r
owth
of d
i
g
ital
m
e
d
i
cal i
m
ag
in
g
.
This g
r
owth
h
a
s main
ly
foc
u
se
d o
n
C
o
m
put
ed Tom
ogra
p
hy
(C
T), t
h
e di
f
f
ere
n
t
di
gi
t
a
l
radi
ol
o
g
i
cal
processe
s f
o
r va
scul
ar
, n
u
cl
ear
m
e
di
cal
im
agi
ng
wi
t
h
Si
n
g
l
e
Ph
ot
o
n
Em
i
ssion C
o
m
put
ed
Tom
ogra
phy
(
SPEC
T
)
,
car
di
ova
scul
ar a
n
d cont
rast
im
agi
ng, m
a
m
m
ograp
hy
, M
a
gnet
i
c
R
e
s
ona
nce Im
agi
n
g (
M
R
I),
di
ag
n
o
s
t
i
c
ul
t
r
aso
u
n
d
im
agi
ng, a
n
d
Posi
t
r
o
n
Emissio
n
To
m
o
grap
h
y
(PET). All th
ese pro
cesses
are produ
cing
in
cr
easin
g
qu
an
tities o
f
im
ag
es.
Th
ese
im
ages are t
y
pi
cal
l
y
di
fferent
fr
om
ot
her p
hot
og
ra
phi
c
i
m
ages beca
use
they reveal internal fram
e
work a
s
oppose
d to an image
of e
x
ternal surfaces.
In
v
i
ew of th
i
s
, surv
ey of literatu
re
h
a
s been
do
n
e
i
n
th
e area of tom
o
g
r
aph
y
, wav
e
lets, m
u
lt
i
wavel
e
t
s
a
n
d
vari
ous
de
n
o
i
s
i
ng t
e
c
hni
que
s
..
A n
u
m
b
er
of
resea
r
che
r
s
ha
ve
pu
bl
i
s
h
e
d i
m
age den
o
i
s
i
n
g
literatu
re [9-2
5] th
ere is tre
m
en
dou
s research
th
at is g
o
i
ng
o
n
, fo
r b
e
tter i
m
ag
e q
u
a
lity t
h
rou
gho
u
t
th
e
g
l
ob
e.
The t
h
res
h
ol
di
ng
i
s
un
dert
a
k
en
o
n
t
h
e
pi
xel
by
pi
xel
basi
s
[2
6
–
2
8
]
o
r
by
co
nsi
d
eri
n
g t
h
ei
nfl
u
e
n
ce
of
n
e
igh
borhoo
d
wav
e
let co
efficien
ts on
th
e
wav
e
let co
effi
ci
ent
s
t
o
be t
h
resh
ol
de
d.
C
a
i
an
d Si
l
v
e
r
m
a
n [
2
9
]
p
r
op
o
s
ed
a thresho
l
d
i
ng
m
e
t
h
od
wh
ich
takes th
e immed
i
a
t
e n
e
ig
hbo
ri
n
g
co
efficien
ts in
t
o
accoun
t to
form
th
e
th
resh
o
l
d
.
Th
e id
ea
of
n
e
ighb
oring
wav
e
let th
resho
l
d
i
n
g
was ex
tend
ed
b
y
Ch
en and
Bu
i [30
]
in
t
o
th
e m
u
lti
wav
e
letsch
em
e
.
It was p
r
o
v
e
d th
at n
e
ig
hb
or m
u
l
ti wav
e
let
d
e
no
ising
ou
tperfo
rm
s th
e n
e
ig
hbo
r sing
le wav
e
let
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
42
1 – 4
2
8
42
2
den
o
i
s
i
n
g [
3
1]
fo
r s
o
m
e
t
e
st im
ages and
r
eal
t
i
m
e
si
gnal
s
. C
h
e
n
et
al
.
[3
2]
p
r
o
p
o
sed
a noi
se s
u
pp
re
ssi
on
m
e
thod whic
h considers
asquare
nei
g
hborhood window
t
o
custom
ize the wavelet filter thres
h
old for im
age
d
e
no
ising
.
These m
e
th
o
d
s re
m
o
v
e
th
e
no
ises fro
m
th
e im
ages effecti
v
ely. Crouse et
al [33] de
vel
ope
d a
fram
e
wor
k
f
o
r
st
at
i
s
t
i
cal
si
gnal
p
r
ocessi
n
g
base
d
o
n
wa
v
e
l
e
t
dom
ai
n h
i
dde
n m
a
rkk
o
v
m
odel
s
(H
M
M
)
.
K
i
ng
sbur
y [
34] p
r
op
osed
th
e 2
D
du
al tr
ee co
m
p
lex
w
a
v
e
let wh
ich
satisfies th
ese req
u
irem
en
ts effectiv
ely.
B
u
t
t
h
i
s
m
e
t
hod i
s
l
e
ss ef
fi
ci
ent
fo
r m
o
t
i
on est
i
m
at
i
on si
nce t
h
e m
o
t
i
on i
n
fo
rm
ation i
s
r
e
l
a
t
e
d
t
o
t
h
e
coef
fi
ci
ent
p
has
e
, w
h
i
c
h
i
s
no
nl
i
n
ear
f
u
nct
i
o
n
of
est
i
m
at
i
on.
Ai
m
of t
h
e
pape
r i
s
t
o
em
pha
si
ze t
h
e
p
r
obl
em
s
an
d
so
l
u
tio
ns
in
relatio
n to
to
m
o
g
r
aph
i
c imag
es wh
ich
arise in m
e
dical field in
view of its inc
r
e
a
sing
i
m
p
o
r
tan
c
e
in
t
h
e p
r
esen
t d
a
y requ
irem
en
ts.
2.
R
E
SEARC
H M
ETHOD
The
pr
o
p
o
s
ed
m
e
t
hod
ol
o
g
y
i
s
basi
cal
l
y
co
n
t
ai
ns t
w
o f
u
nct
i
onal
st
e
p
s
1.
g
e
n
e
ration
o
f
i
n
itial ele
m
en
t
2.
d
e
term
in
atio
n
o
f
fitn
ess fun
c
tio
n
A.
Generation
o
f
Initia
l Element
In
t
h
e first pr
o
cess, n
at
initial atom
, each pa
rt of elem
ent n
E
are g
e
n
e
rated
.
Th
e set rep
r
esen
tatio
n of in
itial
ele
m
ents are
given as
R
il
{
r
0,
r
1,
r
2… r
nE – 1
}
il
;0
≤
l
≤
n
at
−
1
Whe
r
e {R}
il
is th
e
l
th
ele
m
e
n
t g
e
n
e
r
a
ted
to o
b
t
ain
w
i
ndow
s th
at ar
e closer
to
th
e
i
th
wi
n
dow of th
e
o
r
ig
in
al
noisy im
age. Each at
om
of
the ge
ne
rated
ele
m
ent r
il
k
{R}
il
; 0
≤
k
≤
n
E
-1
, is an
arb
itrary in
teg
e
r
g
e
n
e
rated
with
in
th
e i
n
terv
al [0
,
n
w
-1] p
r
ov
id
ed
th
at all th
e ato
m
s o
f
each
elemen
t h
a
s to
satisfy th
e co
nd
ition
r0
≠
r1
≠
…
≠
r
nE
-1
.
B.
Determination of Fi
tness
F
uncti
on
A fi
t
n
ess
fu
nct
i
on
deci
de
s
wh
et
her t
h
e
gene
r
a
t
e
d el
em
ent
are fi
t
t
o
s
u
rvi
v
e
or
n
o
t
,
t
h
at
can
be
gi
ve
n a
s
Whe
r
,
f
i
(l)
is
th
e fitn
ess
o
f
th
e
l
th
el
em
ent gene
rat
e
d
fo
r
t
h
e
i
th
w
i
n
dow
&
L2
ilk
is th
e L2
no
r
m
d
i
s
t
a
n
ce
d
e
term
in
ed
b
e
t
w
een th
e
w
i
& t
h
e
w
i
nd
ow
i
ndex
e
d b
y
th
e
k
th
at
om
of t
h
e
l
th
ele
m
ent.
The
L2
ilk
is
d
e
term
in
ed
as
fo
llows
Whe
r
e,
W’r
ilk
i
s
t
h
e wi
n
d
o
w
i
nde
xe
d
by
r
ilk
th
at is conv
erted
to
m
u
lti-wavelet tran
sfo
r
m
a
tio
n
d
o
m
ain
as d
o
n
e
in
(4
) and
(5
)
.
Th
e
p
r
o
c
ed
ure
for
wav
e
let m
u
lti reso
lu
tion
tran
sform
in
is describ
e
d
b
e
low
Step 1:
At
lev
e
l
j
2
-
D
real im
a
g
e is con
v
o
l
v
e
d
with
scale and
wav
e
let
filters alon
g th
e
rows of
2
-
D im
ag
e.
Step 2:
Th
e resu
lts o
b
t
ain
e
d
after step
1
are co
nv
o
l
v
e
ag
ai
n
with
scale & wav
e
let filters alo
n
g
th
e co
lum
n
s o
f
the 2-D im
age.
Step 3:
T
h
e
n
fi
l
t
e
r i
s
su
b-
sam
p
l
e
d
by
a
fact
o
r
of t
w
o
.
Step 4:
At lev
e
l
j
th
e ap
p
r
o
x
i
m
a
tio
n
is c
o
n
s
i
d
ered
as th
e in
pu
t to
the n
e
x
t
lev
e
l
j+1
. Th
is pr
o
c
ed
ur
e is
fo
llowed
for al
l th
e lev
e
ls.
C.
B
i
vari
a
te S
h
ri
nka
ge Fun
c
ti
on
M
o
del
(
B
FM)
:
Biv
a
riate shrink
ag
e fun
c
tio
n m
o
d
e
l (BFM
) is a
n
e
w
mo
d
e
st
n
o
n
-
Gaussian
b
i
v
a
riate
p
r
ob
ab
ility
d
i
stribu
tio
n
fun
c
tio
n
t
o
p
e
rfect th
e statistics
o
f
wav
e
let coefficien
ts
o
f
n
a
tu
ral im
ag
es. Th
e m
o
d
e
l arrests th
e
depe
n
d
ence a
m
ongst
a wave
l
e
t
coeffi
ci
ent
& i
t
s
parent
.
Usi
n
g B
a
y
e
si
an est
i
m
a
t
i
on t
h
eory
we
devel
op f
r
o
m
t
h
i
s
m
odel
a m
odest
no
n
-
l
i
n
ear
s
h
ri
nka
ge
fu
nct
i
o
n fo
r w
a
vel
e
t
de
n
o
i
s
i
n
g, w
h
i
c
h
t
a
ke
a
b
r
oa
d vi
e
w
of
so
ft
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Eval
u
a
t
i
o
n of
N
o
i
s
e
Excl
usi
o
n of
Me
di
cal
I
m
a
g
es usi
n
g
H
y
bri
d
i
z
at
i
o
n
of
…
(
Shrut
i
B
h
ar
gav
a)
42
3
t
h
res
hol
di
n
g
a
p
p
r
oach
. The
new s
h
ri
n
k
age
fu
nct
i
o
n, w
h
i
c
h ha
n
g
s o
n
bot
h t
h
e c
o
ef
f
i
ci
ent
& i
t
s
parent
,
pr
o
duces
i
m
prove
d
res
u
l
t
s
f
o
r w
a
vel
e
t
-
base
d i
m
age den
o
i
s
i
ng.
Let
us c
onsi
d
e
r
t
h
at
w
2
sy
m
bol
i
ze t
h
e
pa
r
e
nt
o
f
w
1
(w2 is th
e wav
e
let co
efficien
t at th
e id
en
tical sp
atial
p
o
s
ition
as w1
, bu
t th
en
ag
ain at th
e
n
e
x
t
co
arser scale). Th
en
y = w
+ n
(4
)
W
h
er
e
w
=
(w1
,
w2
)
,
y = (
y
1, y2
) &
n
= (n
1, n2
). Th
e
noise standards
n1, n2 are
iid
ze
ro-m
ean Gaussia
n
with
vari
a
n
ce\
si
gm
a n
2
.
The st
a
nda
r
d
M
A
P est
i
m
at
or f
o
r
w
gi
ve
n t
h
e n
o
i
s
y
o
b
se
rv
at
i
on y
i
s
:
Th
e equ
a
tio
n fo
r wav
e
let co
efficien
t
w
1
i
s
gi
ven
as
Let con
s
id
er that
(5
)
The
n
we
den
o
t
e t
h
e
bi
va
ri
at
e sh
ri
n
k
a
g
e f
u
nct
i
o
n
m
odel
(B
FM
) B
F
M
=
(
Y
c
, Y
p
,
n
,
, T
),
w
h
er
e
Y
c
is the
coefficient of
each sub-ba
nd Y
p
is its p
a
ren
t
o
f
th
e co
efficien
t,
n
is th
e v
a
rian
ce
o
f
no
isy sig
n
a
l,
is
t
h
e
marginal varia
n
ce
&
T
i
s
t
h
e t
h
res
h
ol
d
val
u
e
.
D. Metho
d
o
l
og
y
fo
r Window
Selectio
n
in Imag
e Deno
ising
:
Let, f
i
rst d
i
scuss th
e alr
e
ad
y
p
r
esen
t
w
i
ndow
selection
m
e
th
od
o
l
o
g
y
u
s
ed
i
n
th
e techn
i
qu
e, h
e
r
e
, is a
b
r
ief
descri
pt
i
o
n
.
Let
us
assum
e
t
h
at
,
I
(x
, y
)
be t
h
e
uni
que
C
T
im
age an
d
I
AW
GN
(x
, y)
b
e
th
e im
ag
e co
rrup
ted b
y
Ad
d
itiv
e
Wh
ite
Gau
s
ian No
ise, wh
ere
0
≤
x
≤
M
−
1, 0
≤
y
≤
N
−
1. T
h
e
I
AW
GN
is applied to t
h
e fi
rst stage
of t
h
e propos
ed
t
echni
q
u
e,
wi
n
d
o
w
-
b
ase
d
t
h
r
e
sh
ol
di
n
g
. T
h
e
wi
n
d
o
w
sel
e
c
t
i
on p
r
oced
ure
descri
bed
he
r
e
i
s
one
of t
h
e key
mechanism
s
of the
first stage
of proce
ssing
o
f
t
h
e
C
T
i
m
age de
noi
si
ng
t
ech
ni
q
u
e.
In the procedure, a
carbon copy of the I
AW
GN
, lab
e
led
as I’
AWGN
, i
s
gener
a
t
e
d. I
AW
GN
an
d I’
AW
GN
, a w
i
n
dow
of
p
i
x
e
ls are con
s
id
ered
&set to
a
m
u
lti-wav
e
let tran
sfo
r
m
a
tio
n
.
The p
r
o
cess
o
f
ex
t
r
actin
g
the win
dows from th
e
im
age I
AW
GN
i
s
gi
ve
n i
n
t
h
e Fi
gu
re
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
42
1 – 4
2
8
42
4
Fi
gu
re
1.
Pr
oce
ss o
f
e
x
t
r
act
i
n
g
t
h
e
wi
n
d
o
w
s
f
r
om
t
h
e gi
ven
i
m
age I
AW
GN
In th
e
Figu
re 1, w ind
i
cates the windo
w
o
f
p
i
x
e
ls
p
u
lled ou
t
fro
m
th
e im
ag
e I
AW
GN
& S size is the
step
size
of
th
e wi
n
dow. Th
is is carried
ou
t all ov
er th
e
i
m
ag
e and
so
w
j
w
i
n
dow
s
ar
e a
c
h
iev
e
d,
w
h
er
e
,
0
≤
i
≤
n
w
-
1. By
th
e
sam
e
way, it is also e
x
ecute
d in t
h
e im
age I
AW
GN
and r
e
c
e
i
v
es w
j
,
0
≤
j
≤
n
w
-1
,
w
h
er
e,
n
w
si
gni
fi
es num
ber o
f
w
i
nd
ow
s. Th
en
, th
e
o
b
t
ain
e
d w
i
n
dow
of
p
i
x
e
ls ar
e tr
ansfo
r
m
e
d
to
m
u
lt
i wav
e
let tran
sform
a
t
i
o
n
do
main
as
fo
llows
W
i
(a
,b
)
=
F
GHM
(a
,
b
)
.
w
i
(a
,
b
)
.
F
T
GHM
(
a
,
b
) [3
0
]
(6
)
W
j
(a
,b
)
=
F
GHM
(a
,
b
)
.
w
j
(a
,
b
)
.
F
T
GHM
(a
,
b
)[5
0
]
(
7
)
Whe
r
e, 0
≤
a
≤
m
−
1, 0
≤
b
≤
n
−
1 a
nd
m
X
n
i
n
dicates the
window size.
In (6) and
(7)
F
GHM
is the concatenated
filter co
efficien
t of GHM mu
lti-wav
e
let tran
sform
a
t
i
o
n
,
W
i
and
W
j
are
w
i
and
w
j
in
m
u
lti-wav
e
let domain
,
respectively.
For eac
h
W
i
, W
j
that are nea
r
er to
W
i
are selec
t
ed founde
d
on L2 norm
distance (
L
2
ij
), whi
c
h can
be calc
u
lated
usi
n
g (9
),
(8
)
Using
the
L
2
ij
, th
e
W
j
windows that are
nea
r
er to the
W
i
, W
‘
L2ij
can be
de
marcated as
W
‘
L2ij
= W
L2ij
-
,
w
h
er
e,
W
L2ij
i
s
gi
ven
as
(9
)
Every
i
th
w
i
ndow
sets in
W
‘
L2
ij
are or
gani
ze
d
i
n
ascendi
ng
or
der
base
d o
n
t
h
ei
r cor
r
esp
o
ndi
ng
L
2
ij
. F
r
om
the
sorte
d
window set,
n
c
nu
m
b
er
o
f
w
i
ndo
ws ar
e chose
n
(for every
W
i
) and
the rem
a
in
in
g
are o
m
itted
o
u
t
, wh
i
c
h
leads to recei
ve
W
‘
L
2
ik,
wh
e
r
e, 0
≤
k
≤
n
c
−
1.
3.
R
E
SU
LTS AN
D ANA
LY
SIS
W
i
t
h
t
h
is algo
rith
m
su
bj
ected to
v
a
ri
o
u
s type of im
ag
e co
rrup
ted b
y
well kn
own
typ
e
of
n
o
i
ses it i
s
bei
n
g f
o
un
d
fr
om
t
h
e t
a
bl
e 2
t
h
at
al
l
t
h
e i
m
ages
has s
h
ow
n
a q
u
i
t
e
im
pro
v
e
m
e
nt
i
n
w
h
e
n
cor
r
upt
e
d
by
e
i
t
h
er
t
y
pe of
ab
o
v
e
m
e
nt
i
oned
n
o
i
s
es.Ta
b
l
e
1 s
h
ows
t
h
e c
o
m
p
ari
s
o
n
of
o
u
r
obt
ai
ne
d
dat
a
wi
t
h
t
h
e
res
u
l
t
s
fr
om
WT-
T
N
N
a
p
pr
oach
wi
t
h
db
8
and
bi
or
6.
8
[
4
4]
i
t
can
be
see
n
t
h
at
o
u
r
p
r
op
ose
d
m
e
t
hod
h
a
s pe
rf
orm
e
d
wel
l
i
n
rem
ovi
ng
t
h
e
d
i
ffere
nt
t
y
pe
o
f
n
o
i
s
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Eval
u
a
t
i
o
n of
N
o
i
s
e
Excl
usi
o
n of
Me
di
cal
I
m
a
g
es usi
n
g
H
y
bri
d
i
z
at
i
o
n
of
…
(
Shrut
i
B
h
ar
gav
a)
42
5
(
a
)
(
b
)
(
c
)
Figure
2. Im
age (a) origi
n
al image, (b) Im
age
with Ga
ussia
n
noise
(c
) denoised
im
age
Im
age denoi
si
ng
usi
n
g
bi
va
ri
at
e shri
n
k
a
g
e
fu
nct
i
on a
n
d
whi
c
h are i
n
t
u
rn s
u
bject
e
d
t
o
Pa
rt
i
c
l
e
Swarm
O
p
t
i
m
i
zat
i
on
(PS
O
)
h
a
s sh
o
w
n
a m
u
ch si
gni
fi
cant
i
m
provem
e
nt
i
n
sal
t
an
d
pe
ppe
r
noi
se.
Nearl
y
5
0
%
i
m
p
r
ov
em
en
t i
s
seen
in
PSNR, wh
ereas M
S
E h
a
s
d
ecr
eased
to
n
early 9
7
% after ad
ap
tiv
e filters.
Th
ese
im
pro
v
em
ent
s
has hel
p
ed t
o
achi
e
ve
bet
t
e
r WPS
N
R
bet
w
een 2
9
% t
o
4
0
% i
m
provem
e
nt
i
n
va
ri
o
u
s im
age
t
y
pe.A
part
fr
o
m
sal
t
and
p
e
ppe
r noi
se
, gau
ssi
o
n
noi
s
e
,
spec
kl
e no
i
s
e
&
p
o
i
s
s
o
n noi
se has sho
w
n
im
pro
v
em
ent
to
s
o
m
e
ext
e
nt
but
sal
t
an
d pe
ppe
r has st
o
o
d
apart
fr
om
al
l
the noi
ses
.
Tab
l
e 1
.
C
o
m
p
arativ
e p
e
rforman
ce
of p
r
o
p
o
s
ed
ap
pro
a
ch
with
d
b
8
and
bio
r
6
.
8
wav
e
let
filter
Set of r
e
sults fr
o
m
W
T
-
T
NN appr
oach with db8
wavelet filter[44]
Set of results fro
m
WT
-TN
N
appr
oach with bior
6.
8[4
4
]
Pr
oposed with PSO
Im
a
g
e
Noise Std.
dev.
PSNR
Noise Std.
dev.
PSNR
Noise Std.
dev.
PSNR
L
eena 10
34.
29
10
34.
34
10
35.
32
Bar
b
ar
a 10
31.
76
10
31.
81
10
34.
3
Ultr
a Sound
10
33.
86
10
34.
64
10
33.
85
Tab
l
e
2
.
Perf
orman
ce co
m
p
ar
isio
n of
p
r
op
o
s
ed
m
e
th
o
d
o
l
ogy an
d its eff
ect on
v
a
r
i
o
u
s
no
i
s
e and
im
ag
e typ
e
I
m
age-
B
arbara
PSNR
MSE
WPSNR
SSIM
TIME
POI
SSON NOI
S
E
36.
924
8
13.
200
8
42.
127
1
0.
9472
44
0.
2255
09
GAUSS
I
AN NOIS
E
34.307
8
24.11
5
5
37.742
7
0.8834
2
0.2327
28
SAL
T
& PE
PPE
R
NOI
SE
46.
209
5
17.
236
3
48.
801
6
0.
8600
89
0.
2181
33
SPE
CKLE
NOI
S
E
34.
164
24.
927
9
39.
123
0.
9181
54
0.
2486
14
I
m
age-
L
eena
PSNR
MSE
WPSNR
SSIM
TIME
POI
SSON NOI
S
E
37.
839
4
10.
694
43.
088
0.
9539
33
0.
2066
91
GAUSS
I
AN NOIS
E
35.320
5
19.1
38.527
1
0.8907
29
0.2125
08
SAL
T
& PE
PPE
R
NOI
SE
47.
080
4
16.
273
64
48.
924
9
0.
8349
26
0.
2131
86
SPE
CKLE
NOI
S
E
34.
765
3
21.
704
7
39.
669
3
0.
9231
08
0.
1845
77
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
42
1 – 4
2
8
42
6
I
m
age-
C
T Scan
PSNR
MSE
WPSNR
SSIM
TIME
POI
SSON NOI
S
E
35.
061
3
20.
274
3
40.
795
8
0.
9546
22
0.
2289
42
GAUSS
I
AN NOIS
E
34.514
9
22.992
6
38.152
8
0.7898
65
0.2088
39
SAL
T
& PE
PPE
R
NOI
SE
48.
800
2
18.
857
1
43.
869
8
0.
8284
32
0.
2223
14
SPE
CKLE
NOI
S
E
33.
767
5
27.
310
3
38.
714
8
0.
9493
95
0.
2205
01
I
m
age-Ul
t
rasound
PSNR
MSE
WPSNR
SSIM
TIME
POI
SSON NOI
S
E
34.
879
5
21.
141
5
40.
543
9
0.
9610
04
0.
2311
37
GAUSS
I
AN NOIS
E
33.851
8
26.785
3
37.431
4
0.7675
68
0.2116
41
SAL
T
& PE
PPE
R
NOI
SE
48.
543
4
20.
909
3
42.
813
3
0.
8407
76
0.
1965
96
SPE
CKLE
NOI
S
E
33.
341
8
30.
123
5
37.
856
7
0.
9506
67
0.
2255
87
3.
CO
NCL
USI
O
N
In t
h
i
s
pa
per
,
new t
ech
ni
q
u
e
has been p
r
es
ent
e
d. T
h
e p
r
o
pos
ed B
i
vari
at
e and PS
O bas
e
d t
echni
qu
e
app
r
oach
not
onl
y
com
put
at
i
onal
l
y
effi
ci
ent
but
al
so
gi
ves bet
t
e
r
per
f
o
rm
ance i
ndi
c
a
t
e
d by
pe
rf
or
m
a
nce
i
ndi
ces P
S
NR
,
M
S
E,
WPS
N
R
,
SSIM
a
n
d t
i
m
e
. Fi
nal
l
y
, i
t
i
s
concl
ude
d t
h
at
t
h
e
pr
op
os
ed ap
p
r
oac
h
i
n
t
e
rm
s
o
f
PSNR,
WPSNR im
p
r
o
v
e
men
t
is o
u
t
perform
e
d
.
Th
e
p
r
o
p
o
s
ed
techn
i
q
u
e
op
timize t
h
e po
ssi
b
ility
o
f
l
o
w
pass c
o
efficient from
each subba
n
dba
se
d
on am
ount of shri
nka
ge is
re
lated to signal de
pende
n
t nois
e
v
a
rian
ce.In
t
h
is p
a
p
e
r a
n
e
w
tech
n
i
qu
e is
p
r
o
p
o
s
ed
to
m
iti
g
a
te th
e
n
o
i
se
in
im
ag
es. Acco
rd
ing
to
resu
lts th
e
n
o
v
e
l Biv
a
riate tech
n
i
q
u
e
op
ti
m
i
zed
b
y
Particle Swarm
Op
ti
mizatio
n
is co
m
p
u
t
atio
n
a
lly efficien
t and
per
f
o
r
m
s
si
gni
fi
cant
l
y
supe
ri
or i
n
pe
rf
o
r
m
a
nce i
ndi
ces i
n
d
i
cat
ed by
PSN
R
,
M
S
E,
W
P
S
N
R
,
SS
IM
and
t
i
m
e
.
Fi
nal
l
y
, we c
a
n
co
ncl
u
de t
h
at
i
n
t
e
rm
s of
WP
SNR
a
n
d
PS
N
R
t
h
e
pr
op
ose
d
ap
pr
oac
h
i
s
ou
t
p
erf
o
rm
ed.
REFERE
NC
ES
[1]
Arivazhag
an S., Deivalakshmi S., and ka
nnan. K.B.N. Gajbhiy
e
, C.
Muralidhar
,
Sijo N. L
ukose, M.P. Subramanian
,
“
P
erform
ance a
n
al
y
s
is
of wav
e
le
t fil
t
ers
for
im
age deno
is
i
ng”,
Advances in Computational scien
ces
and
Technology
, vo
l. 1, no. 1, pp.
1-1
0
, ISSN 0973-6107, 2007
.
[2]
A.C. Kak and M
.
S
l
ane
y
, “
P
rincipl
e
s
of com
p
uteri
zed tom
ographic im
aging
,
”
IEEE
, Inc.
, New York:
IEEE
Pre
s
s
.
2006.
[3]
Alexander
Wong, Akshay
a Mis
h
ra, Pau
l
Fieguth and David
Clausi, 2008
, “An
adaptive Monte Carlo
approach
to
nonline
a
r im
age
denois
i
ng”
, P
r
oceed
ings
of the
19th IEE
E
In
ternational Con
f
erence on Pat
t
e
rn Recogn
ition
,
Tampa, pp
. 1-4
.
1988.
[4]
Anja Borsdorf,
Rainer R
a
upach
, Thomas Flohr and Joach
im
Hornegger, “
W
a
v
e
le
t ba
sed noise reduction in C
T
im
ages
us
ing cor
r
ela
tion ana
l
y
s
is
,
”
IEEE Transactions on Med
i
ca
l
Imaging
, vol. 27
, no
. 12
, pp
. 168
5-1703. 2008
.
[5]
Aleks
a
ndra P
i
ˇ
z
u
rica
, “
A
Versatile W
a
v
e
le
t Dom
a
in Noise Filtrat
ion Te
chniq
u
e for Medica
l
Im
aging”,
IEEE
Transactions on
Medica
l Imagin
g
, VOL. 22, NO. 3. 2003
[6]
Arivazhag
an S., Deivalakshmi S., and ka
nnan. K.B.N. Gajbhiy
e
, C.
Muralidhar
,
Sijo N. L
ukose, M.P. Subramanian
,
“
P
erform
ance a
n
al
y
s
is
of wav
e
le
t fil
t
ers
for
im
age deno
is
i
ng”,
Advances in Computational scien
ces
and
Technology,
vo
l. 1, no. 1, pp. 1-1
0
, ISSN 0973-6107, 2007
.
[7]
Bart Goossens, Aleksandra Pizu
rica
and W
ilfrie
d
Philips,
“
R
em
oval of corr
ela
t
e
d
noise b
y
m
odeling th
e signal
of
inter
e
st in
th
e w
a
velet domain”,
I
EEE Tr
ansactio
n on Image Processing,
vol. 18
, n
o
. 6
,
pp
. 1153-1
165. 2009
.
[8]
Bhatia
M. Karl
W
.
and W
illsk
y
, A.S.J.P.Morga
n
&
Co., “
T
om
ographic r
econst
r
uction
and esti
m
a
tion based o
n
m
u
lti-scal
e n
a
tur
a
l-pix
e
l b
a
ses”,
I
EEE Transactio
ns on Image
Processing
Vol. 6, I
ssue 3, pp
. 463-4
78. 1997
.
[9]
Bing-gang Ye and Xiao-ming Wu, “Wavele
t d
e
noising arithmetic res
earch b
a
se
d on small hepatocellular
car
cino
ma
CT im
age”
, in
p
r
oceed
ings
of
3rd Internationa
l
Conference on Bio-in
formatics a
nd Biomedical
Engineering
, IS
BN
No: 978-1-4244-
2901-1, pp
. 1-3
.
2009.
[10]
D. Gnanadurai
and V. S
a
das
i
vam
,
“
A
n efficien
t adap
tive thresholding
technique
for wavelet based image
denoising”,
In
ter
national
Journal of S
i
gnal
Processing
, volume 2
,
number 2, pp
. 11
4-119. 2005
.
[11]
Detlev Marp
e Hans L. C
y
con
,
Gunther Zand
er
and Kai-Uwe
Barthel, “Context-based de
no
ising of images using
iter
a
tiv
e wavelet thresholding”,
Pr
oc. SPIE Pr
o
ceed
ings
vol.46
71,
Visual Communications and Image Processing
,
C.C.Jay
Kuo
.
Ed
itors, pp
. 907-91
4. 2002
.
[12]
Dim
itrios
Charal
am
pidis
,
“
S
teera
b
le weight
ed m
e
dian filt
ers
”
,
I
E
EE Transactions on Image Processing
, vol. 19, no.
4, pp
. 882-894
.
2010.
[13]
D. Donoho, M. Raimondo, “A
fast wavelet algorithm for im
age deblurring
”,
IEEE Transactions on image
processing
16 (2
). 2006
[14]
D.R.K. Brownrigg, Th
e weighted median
filter
,
Communications of
the ACM
27
(
8
) 807–818. 198
4
[15]
Erdem
Bal
a
, A
y
sin
Ertuzun
,
“
A
m
u
ltivari
a
te
t
h
resholding t
e
c
hnique for
im
ag
e denoising
usi
ng m
u
ltiwavel
et
s,”
EURASIP Journ
a
l on
Applied
Signal Processing,
Hindawi Pub
lis
hing Corporatio
n
, vol. no
. 2005
, issue 8, pp. 120
5
-
1211. 2005
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Eval
u
a
t
i
o
n of
N
o
i
s
e
Excl
usi
o
n of
Me
di
cal
I
m
a
g
es usi
n
g
H
y
bri
d
i
z
at
i
o
n
of
…
(
Shrut
i
B
h
ar
gav
a)
42
7
[16]
F
.
Natterer
,
Th
e m
a
them
atics
of com
puterized
tom
ograph
y
:
Classics-in applied mathematics
, SIAM 32, ISB
N
0898714931. 20
01.
[17]
G. Chen, T. Bui
,
Multiwavelet d
e
noising using neighbouring co
effici
ents,
IEEE signal processing letters
10 ,211–
214. 2003
.
[18]
G. Chen,
T. Bui, A.Krzy
zak
, Image denoisi
ng
using neighbouring
wa
velet co
efficients,
Integrated
Computer-Aided
Engineering
99–
107. 2005
.
[19]
Hossein Rabbani, “Image denoising
in steerable p
y
r
a
mid domain based on a local Laplace prior”,
Pat
t
e
r
n
Recogn
ition
, vo
l. 42
, no
. 9
,
pp
. 2
181-2193. 2009
.
[20]
Hossein Rabban
i
, “Wavelet-do
main
medical
image denoising
using bi
va
r
i
at
e L
a
pl
a
c
i
a
n mi
x
u
r
e
mo
d
e
l
”
,
IEE
E
Transactions
on Biomedical Engineering
, vol. 56
, no. 12, pp. 2826
-2837. 2009
.
[21]
Joa
o
M.
Sa
nc
he
s,
Ja
c
i
nto C.
Na
sc
ime
n
to,
a
n
d Jorge
S
.
M
a
rques
,
“
M
edical
im
ag
e noise reduction using the
S
y
lvester-Ly
a
pu
nov equation”,
I
EEE Transactio
ns on Image
Processing
, vo
l. 17,
no. 9
,
Sept. pp
. 1
522-1539, 2008
.
[22]
Landi G.
and Lo
li Picco
lomini.
E., “An algorithm for
im
age deno
i
s
ing with auto
ma
t
i
c
noi
se
e
s
ti
ma
t
i
on”,
Journal o
f
Mathematical I
m
aging and Vision
, vol. 34
, no
. 1
,
pp
. 98-106
. 20
09.
[23]
M. Crouse, R
.
Nowak, R. B
a
rani
uk
, Wavelet based
signa
l processing using hidde
n mar
kov models, IEEE
Transactions on
signal pro
cessin
g
, 46
, 886–902
.
1998.
[24]
N.
P.
Anil and S
.
Natarajan,
“A new
techn
i
que f
o
r image denois
i
ng using fourth
order PDE and
Wiener filter
”,
International Jou
r
nal of
Applied
Engineeri
ng R
e
search
, vo
l. 5,
no. 3, pp. 509-516.
2010.
[25]
N.T. B
i
nh,
A.
Khare, Ad
aptiv
e
com
p
lex wav
e
l
e
t t
echn
i
que
for
medical imag
e denoising
, ICD
B
ME in Vietnam,
IFMBE Proceedings, www.springerlink
.
com, 27
, 1
96–199. 2010
,
[26]
P
r
i
y
am
Ch
att
e
rj
ee,
and
P
e
y
m
an
M
ilanfar
, “
I
s
de
nois
i
ng de
ad
?
,
”
IEEE Transactions on Image Processing
, vol
. 1
9
,
no. 4
,
pp. 895-9
11. 2010
.
[27]
Prof. S.A. Ali, Dr. S. Vathsal a
nd Dr. K. Lal
Kishore, “GA-b
a
sed Wi
ndow Selection Method
olog
y
to
Enhan
c
e
Window-based Multi-wavelet tr
ansform
a
tion an
d thre
sholding ai
ded CT im
agedenoising techniqu
e”,
IJCSIS
, V
o
l.
7,
No. 2. 2010.
[28]
R. Vija
y
kum
ar,
P.T. Vanath
i, P. Ka
nagasabapa
t
h
y
and D. Eben
ezer
, “
R
obust st
atisti
cs based al
gorithm
to rem
o
ve
s
a
lt and
pepp
er
nois
e
in
im
ages
,
”
International
Journal of S
i
gn
al Processing
, v
o
l. no
. 5
,
issue
no. 3, pp. 164-1
73.
2009.
[29]
R. Neel
am
ani
,
H. Choi, R
.
B
a
raniuk
, Forwar
d: Four
ier-wav
elet r
e
gularized
deconvolu
tion f
o
r ill-
condit
i
oned
s
y
ste
m
s,
IEEE Transactions
onsignal processing
, 52, 418–433. 20
04.
[30]
R. Silva, R
.
Min
e
tto
, W
.
Schwart
z
, H. Pedrin
i, A
d
aptiv
e edge-pr
e
serving image d
e
noisi
ng using wavelet tr
ansforms,
Pattern
Anal
y
s
is
and Appl
ication
s
1–14doi:10.10
07/
s10044-012-0266-x.URL ht
t
p
://dx.do
i.org
/
10
.1007/s10044-01
2-
0266-x, 2012
.
[31]
S. Arivazh
agan
, S. Deiv
alakshmi, K. Kann
an, B
.
N. Gajbhi
y
e
, C
.
Muralidh
a
r, Sijo N.
Lukose, M.P. Subramanian
,
“
P
erform
ance a
n
al
y
s
is
of im
age denois
i
ng s
y
s
t
em
for different levels of
wav
e
let decomposition”,
Int
e
rnation
a
l
Journal of Imaging Scie
nce and
Engineering
, G
A
, USA, vol. 1,
no. 3
,
July
, pp. 1
04-107. 2007
.
[32]
T. Downie, B.W
.
Silverm
a
n
,
Th
e discret
e
m
u
ltip
l
e
wavel
e
t
transf
orm
and thresho
l
ding m
e
thods,
I
EEE transactions
on signal processing
, 46, 2558–2
561. 1998
.
[33]
T.
Ca
i, B.
W. Silve
r
ma
n,
Inc
o
rpora
ting informa
t
ion on
neighbouring c
o
e
fficie
n
ts
into wa
ve
let e
s
tima
tion,
Sa
nkhy
a
:
The indian
journ
a
l of statistics
, 6
3
(b), (2)
,
127–1
48. 2001
.
[34]
T.
D.
Bui, G.
Y.
Chen,
Translation inva
r
i
ant deno
ising using multivavelets,
IEEE transactions
on signal
processin
g
46, 3414–3420
,
1998.
[35]
V. Strela
, P. H
e
ll
er, G. Str
a
ng
, P. Topiwa
la
,
C. Heil
, Th
e a
pplic
ation of m
u
ltiwav
ele
t
fil
t
e
r
banks to im
a
g
e
processing,
IEEE transactions o
n
image processing
8, 548–563. 1
999.
[36]
W.
N,
W.
GT
,
Wa
ve
l
e
t
shri
nk
age and gener
a
lized cross valid
ation for image
denoising,
IEEE Transactions o
n
image processin
g
,
7(1)
, 82–90
. 1
998.
[37]
Yaniv Gal, “Denoising of
Dy
n
a
mic Contrast-Enhanced MR Imag
es Using
Dy
namic Nonlocal Means”,
IEEE
Transactions on
Medica
l Imagin
g
, Vol. 29
, No. 2
.
2010
.
[38]
Akshay
a. K. M
i
shra, Alexander Wong,
David. A. Clausi
and
Paul. W.
Fieg
uth, “Adaptiv
e
nonlinear image
denoising and
restoration usin
g a co-opera
tiv
e
Bay
e
sian estimation approach”,
Sixth Ind
i
a
n
Conference o
n
Computer Vision
. 2011
.
[39]
Francisco Estr
ad
a, Dav
i
d Flee
t a
nd Allan Jepson,
“
S
toch
astic
im
age denoising
”, B
r
itish Mach
ine V
i
sion Conferen
c
e
,
London 2009.
[40]
G. Eason, B
.
No
ble,
and I
.
N. Sn
e
ddon. On certain integr
als of
Lipschitz-H
ank
e
l
ty
pe involv
i
ng pr
oducts of Bessel
functions [J]
.
Ph
il. Tr
ans. Ro
y
.
S
o
c.
London. 199
5, A247: 529-55
1.
[41]
G.
G.
Bhutada,
R.
S.
Anand,
S.C.
Saxena, “PSO b
a
sed learn
i
ng of sub-ba
nd adaptive thresholding
function for image
denoising”, SIViP (2012) 6: 1-7.
[42]
H. Chipman, E.
Kolaczy
k
, R
.
M
c
Cullo
ch, Adap
tive ba
y
e
sian wavelet shrinkag
e
,
J Am Stat Assoc, 440(92)
, 1413–
1421, 1997
.
[43]
J. K. Mandal, S. Mukhopadh
y
a
y
,
A novel techniq
u
e for removal o
f
random
valued
impulse noise u
s
ing all n
e
ighbo
r
directional weig
htedpix
e
ls, Proceedings
of
the I
n
terna
tiona
l Co
nferenc
e
on Co
mmmunications in Computer
and
Information Science(CCIS), Springer, 203
, 1
02–1
11. 2011
.
[44]
J. K. Mandal,
S. Mukhopadh
y
a
y
,
Edge pr
eserving restoratio
n of ra
ndom v
a
lued impulse
noises (eprrvin)
,
Proceedings of
Interna
tiona
l Co
nferenc
e
on R
e
c
e
nt
Trends in
I
n
form
ation S
y
st
em
s-RETIS 2011, IE
EE 309–31
4.
2011.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
42
1 – 4
2
8
42
8
BIOGRAP
HI
ES OF
AUTH
ORS
S
h
ruti Bhargava
receiv
e
d her Ba
chelors
degre
e
i
n
Elec
tronics
an
d Comm
unicatio
n Engineer
ing
in the
y
e
ar 2007 and Masters degree in Digita
l
Communication in the
y
e
ar 2010 from R.G.P.V,
Bhopal (M.P) India. Curr
ently
she is pursing
her PhD from t
h
e Dr. K.N. Modi University
,
Newai.
Her r
e
s
e
arch
inte
res
t
s
a
r
e
Im
age pro
ces
s
i
n
g
.
D
r. Ajay
Somkuwar receiv
ed BE with honors fro
m Jabalpur Eng
i
neering
College
and M.
Tech
.
degree in Dig
ital Communication
Engineering
fro
m MACT Bhopal, subseque
ntly
he carried
ou
t
his resear
ch for
m
Indian Institute of
Tec
hnolog
y
,
New Delh
i and
awarded
Ph.D. in 2003
.
Presentl
y he
is
working as Professor in the
D
e
partment of
Electroni
cs and Co
mmunication at
Maulana Azad
National Institute of
Technolog
y
,
Bhopal. He has
published more
than 100 p
a
pers
of national and
International r
e
pute. He has
been M
e
m
b
er/ Chairm
an of man
y
s
e
lec
tion
com
m
ittees
for
r
ecrui
tm
ent of
s
t
a
ff and
facu
lt
y.
H
i
s
res
ear
ch
are
a
s
includ
e s
i
gna
l pr
oces
s
i
ng
Im
age proces
s
i
n
g
and Biom
edic
al Engin
eer
ing.
He has produced
5 Ph.D. degrees
. He is membe
r
of IET
E
,
New
Delhi
and Int
e
r
n
ation
a
l As
s
o
ci
ation o
f
Eng
i
ne
ers
(IAENG).
Recen
tl
y h
e
is
awarded b
y
"In
d
ira Gandhi Shiromani award-2
011"
for his co
ntribution to n
a
tion. He also
awarded
b
y
Best Citizen
of Ind
i
a-2012. and
“Siksha Rattan
Award
- 2012 ”
Evaluation Warning : The document was created with Spire.PDF for Python.