In
d
o
n
e
sian
Jou
r
n
al of
Ele
c
tr
i
c
a
l
En
g
in
e
erin
g
a
n
d
C
om
pu
t
er S
c
ien
ce
Vol.
14, No.
1, April 2019,
pp.
258~266
ISSN: 2502-
4752,
DOI
:
10.115
91/ijeecs.
v
14.
i
1
.
pp258-266
258
Jou
rn
a
l
h
o
me
pa
ge
:
ht
tp:
//i
a
e
score
.
com
/
j
o
u
r
na
l
s
/
i
n
d
e
x
.
p
hp/
i
j
eec
s
Edge enhancement of IBP reconstruc
tion
by using sharp
i
n
finite
symmetrical exponential filter
A
.
R.
A
N
a
z
ren
1
,
N
gad
i
r
an
.
R
2
,
S.N
. Yaak
o
b
3
1
E
m
bedd
ed Compu
tin
g
Res
earch Cl
u
ster, U
n
i
v
ersity M
alay
s
i
a P
e
rlis
(
Un
iM
AP)
,
M
a
l
a
y
s
i
a
2,
3
S
c
ho
ol
o
f
Com
p
u
t
er
a
n
d
C
ommu
ni
cati
on
En
g
i
neerin
g
o
f
U
niv
e
rsity
M
a
l
a
y
s
ia
P
e
r
l
i
s (Un
i
MA
P),
Ma
la
y
s
ia
Art
i
cl
e In
fo
ABSTRACT
A
r
tic
le hist
o
r
y
:
R
e
c
e
i
v
e
d
Sep
2
7
,
2
018
Re
vise
d N
ov
2
8
,
201
8
A
c
c
e
pte
d
D
ec 2,
201
8
Th
is
s
tud
y
c
ont
rib
u
t
e
s
t
o
p
ro
v
i
de
a
n
enh
a
n
cem
e
n
t
t
echni
qu
e
f
o
r
im
p
r
ovi
ng
th
e
Iterati
v
e
Back
P
roj
ecti
o
n
(IBP
)
S
up
er
R
es
ol
u
t
ion
tech
niq
u
e
by
us
in
g
th
e
S
h
arp
Inf
i
nit
e
S
ym
m
e
tri
cal
F
ilt
e
r
(
SIS
E
F
)
.
Th
eo
ret
i
call
y
,
t
h
e
I
BP
t
echn
i
qu
e
op
erates
a
s
m
i
n
i
m
i
ze
r
the
erro
r
r
econs
tru
c
ti
on
iterat
i
v
e
ly
u
nti
l
it
r
e
f
i
ned
the
Hi
gh
R
es
ol
u
tio
n
(HR)
i
m
a
g
e
.
H
o
wev
e
r,
d
u
e
t
o
i
t
erat
ive
m
a
n
n
er
a
n
d
lack
o
f
edg
e
g
u
i
dan
ce
du
ring
t
h
e
b
ack
p
ro
jecti
o
n
operatio
n,
t
h
i
s
t
echn
i
qu
e
has
su
ff
ered
f
rom
pro
d
u
ced
t
he
r
in
gin
g
a
rtef
act
o
n
the
H
R
i
mag
e
a
p
p
earan
ces
.
Ad
diti
ona
l
l
y,
t
he
I
BP
r
eco
nst
r
uct
i
o
n
a
l
s
o
d
e
m
a
nd
s
f
o
r
l
a
rge
nu
m
b
e
r
i
t
erati
on
f
o
r
accom
p
lis
h
ing
t
h
e
pred
icti
o
n
H
R
i
m
age.
T
h
i
s
p
r
oblem
a
ros
e
w
hen
th
e
IBP
es
tim
ator
t
end
e
d
to
o
sci
l
l
a
t
e
a
t
th
e
s
a
m
e
s
ol
ut
ion
f
r
equen
t
l
y
.
In
o
rder
t
o
ov
ercom
e
t
hes
e
c
on
st
rai
n
ts,
th
e
S
I
S
E
F
is
d
ep
lo
y
e
d
as
r
egul
ato
r
to
i
m
p
rove
th
e
IBP
estimat
o
r
w
ith
pro
v
id
es
a
n
ac
cu
rate
e
dg
e
i
n
f
o
rm
a
tio
n
f
o
r
en
han
c
i
n
g
th
e
edg
e
i
mag
e
a
nd
r
ed
uce
t
h
e
ri
ng
ing
art
e
facts. Fort
unat
e
ly,
h
ig
h
l
y
p
r
ecisi
on
o
f
e
d
g
e
in
fo
r
m
a
t
io
n
pro
v
i
d
e
d
by
S
IS
EF
c
a
p
a
b
le
t
o
re
du
c
e
a
m
ou
n
t
of
est
i
mation process repetition
.
K
eyw
ord
s
:
Iter
ative
ba
c
k
pro
je
ct
ion
IS
E
F
Ri
ngi
ng
a
rt
e
f
a
c
t
s
Ed
ge
e
nha
nc
e
m
ent
S
upe
r r
e
solut
i
o
n
Co
pyri
gh
t © 2
019 In
stit
u
t
e
of Advanced
En
gi
neeri
n
g
an
d
Scien
ce.
All
rights
res
e
rv
ed.
Corres
pon
d
i
n
g
Au
th
or:
Ami
r
N
azren
Ab
du
l
Rahi
m,
S
c
hoo
l
o
f
Com
pute
r
a
nd C
o
mm
unica
t
i
on En
gi
neer
in
g,
Uni
v
ersi
ti
M
al
ays
i
a P
e
rlis (
UniMAP
),
P
a
uh P
u
tra
Cam
pus,
0260
0
A
r
au
P
erlis,
Mal
a
ysia.
Em
ail:
am
i
r
nazr
en@u
n
i
ma
p.edu.
my
1.
I
N
TR
OD
U
C
TI
O
N
T
O
S
U
PER
R
E
S
O
LU
T
I
O
N
E
N
H
A
N
C
EM
EN
T
Th
e
are
a
o
f
thi
s
s
t
u
dy
i
s
g
e
tt
ing
a
lo
t
of
i
n
t
e
r
est
s
fro
m
re
se
a
rc
h
e
r
si
nc
e
th
e
co
nv
e
n
t
i
o
n
a
l
Sup
e
r
R
e
s
o
lu
ti
o
n
(
S
R
)
t
e
ch
n
i
que
s
uffers
f
r
o
m
im
per
f
ec
tio
n
in
t
he
H
ig
h
R
e
s
olu
t
io
n
(H
R
)
i
ma
ge
e
s
t
i
m
ati
o
n.
In
C
hia
n
g
et
a
l
[1]
w
o
rks,
a
n
im
pro
v
em
ent
of
I
BP
i
s
pr
od
u
c
e
d
i
n
t
he
r
esa
m
plin
g
al
g
o
ri
t
h
m
by
inte
gr
at
ing
the
ima
g
e
w
a
rpi
n
g
alg
o
ri
th
m
i
n
si
de
t
he
I
BP
t
ec
hni
que
.
Mor
e
o
v
er
,
t
h
e
y
pr
op
ose
d
t
he
i
m
p
le
me
nt
a
t
i
o
n
of
c
ou
p
l
in
g
the
degra
d
at
io
n
m
ode
l
o
f
i
m
a
gi
n
g
s
ys
tem
i
n
to
i
n
t
e
g
rati
n
g
r
e
sam
p
le
t
o
pr
o
d
uc
e
be
t
t
er
e
stim
at
io
n.
I
n
Cohe
n
e
t
al
[
2]
w
ork,
a
n
impr
ove
me
n
t
i
s
su
g
g
e
s
te
d
thr
o
u
g
h
a
p
p
l
y
i
ng
P
o
l
y
p
hase
f
i
l
ter
t
o
r
e
p
la
ce
t
he
g
e
o
m
e
tric
t
rans-
form
ati
o
n
i
n
si
de
t
he
I
tera
tiv
e
B
ack
P
ro
j
e
ct
ion
(IB
P)
a
nd
i
mp
rov
ed
t
he
i
m
a
ge
r
es
o
l
ut
io
n.
F
urt
h
erm
o
re
,
the
ed
ge
e
n
h
a
n
ce
me
nt
t
e
c
h
n
i
que
p
ro
pose
d
b
y
H
y
u
n
w
o
o
e
t
a
l
[3]
us
i
n
g
an
A
n
i
sot
r
op
i
c
D
i
f
f
u
sion
t
o
supp
re
ss
the n
o
i
se
a
n
d
h
ig
h
lig
hte
d
t
he
e
dge
s of
i
ma
ge
.
On
t
op
o
f
t
h
a
t
,
t
h
e
im
age
en
h
a
ncem
en
t
is
d
o
n
e
i
n
Y
a
n
e
t
al
[
4]
w
o
rks.
T
h
e
c
o
m
b
i
na
t
i
on
of
P
a
p
o
u
lis-
G
e
r
c
hber
g
e
x
t
rapo
lat
i
on
i
n
si
de
t
he
I
BP
r
ec
o
n
struc
tio
n
is
p
r
o
p
o
s
e
d
t
o
e
nha
nce
th
e
sp
i
r
al
C
om
pu
te
d
To
mog
r
a
p
hy
(
C
T
)
sl
i
ce
i
m
age
.
B
esi
d
es
t
h
a
t
,
Q
i
n
s
ug
g
e
st
ed
t
h
e
e
f
ficie
n
t
impr
o
v
e
m
e
n
t
i
n
t
he
I
BP
a
lg
o
r
it
hm
thr
o
u
g
h
i
mpr
ove
me
nt
o
f
the
first
i
n
i
tial
gue
ss
w
h
ic
h
is
e
st
im
a
t
ed
b
y
w
a
ve
l
e
t
l
o
ca
l
l
y
a
d
a
p
ti
ve
a
l
gor
it
hm
com
p
are
d
t
o
n
o
rm
al
i
n
t
e
r
p
o
la
tio
n
me
th
od
[
5].
N
e
xt
L
ian
g
e
t
a
l
[
6]
a
lso
gave
i
mpro
ve
me
nt
i
n
t
h
e
ed
ge
detec
t
i
o
n
d
u
ri
ng
c
on
duc
ti
ng
t
he
i
ni
t
i
a
l
i
nter
po
l
a
t
e
d
i
m
a
g
e
v
i
a
a
ppl
y
Non
-
Lo
c
a
l
IB
P
f
a
st
a
lg
o
r
i
t
h
m.
In
t
he
B
e
n
gt
ss
o
n
e
t
a
l
[
7]
p
ro
pose
d
L
*a
*b
d
o
m
a
in
f
or
i
m
p
rov
i
ng
t
h
e
H
u
m
a
n
V
i
s
u
a
l
S
yst
e
m
in
t
he
H
R
im
age
.
Th
is
t
ech
n
i
que
i
s
capa
b
l
e
to i
m
prove
t
he
ou
t
pu
t o
f
H
R
ima
g
e by
prese
r
v
i
ng
the
e
d
ge i
nform
a
t
i
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
ones
i
a
n
J
E
lec
En
g & Co
mp
S
c
i
IS
S
N
: 2502-
47
52
Ed
ge
e
n
h
a
n
ce
m
e
nt
o
f
IBP re
con
s
tr
uc
tio
n b
y
u
s
i
ng
sh
ar
p
i
n
f
i
ni
te sym
m
e
tric
al
e
x
p
o
n
en
t
i
a
l… (A.R.
A
Nazren)
25
9
I
n
o
rde
r
t
o
r
ecove
r
t
h
e
H
R
i
ma
ge,
Bare
j
a
e
t
al
[
8]
p
r
o
p
o
s
ed
a
c
om
b
i
na
t
i
o
n
o
f
IBP
al
gor
it
hm
w
i
t
h
Ca
nn
y
E
d
ge
D
etect
or
a
nd
e
rror
d
i
ffe
r
enc
e
i
ma
ge
t
o
re
co
v
e
r
loss
of
f
requ
e
n
cy
i
nf
orm
a
tion.
T
he
i
m
p
ro
ve
me
nt
is
c
o
n
ti
n
u
ed
i
n
N
a
yak
e
t
a
l
[9],
by
pr
op
os
in
g
a
hy
bri
d
t
ec
hn
iqu
e
bet
w
e
e
n
I
BP
t
ech
ni
q
u
e
a
n
d
Cu
c
koo
op
tim
izer
t
o
e
nha
nce
the
i
n
i
t
ial
g
u
ess
a
nd
H
R
o
u
t
p
u
t
i
m
a
ge
.
In
P
e
jm
an
e
t
a
l
w
o
r
k
[1
0],
t
h
e
impr
ove
me
nt
i
n
t
h
e
IBP
a
l
go
rit
h
m
i
s
p
ro
po
sed
t
h
ro
ug
h
app
l
y
i
ng
t
h
e
i
nt
e
r
po
l
a
t
i
o
n
of
L
R
i
m
age
s
b
efor
e
re
giste
r
ed
i
ns
i
d
e
t
h
e
IBP
alg
o
ri
thm.
A
no
t
h
e
r
w
ork
in
C
here
f
e
t
a
l
[1
1]
i
nt
r
o
d
u
c
ed
e
d
g
e
e
nha
ncem
ent
b
y
u
sin
g
t
h
e
A
n
i
so
t
r
op
ic
Di
f
f
u
s
io
n
t
e
c
h
n
i
qu
e
t
o
p
rese
rv
e
an
y po
s
s
ib
le
e
dg
e
i
m
a
g
e. In
M
e
i
et
a
l
[12
] w
o
rks, imple
m
e
ntat
i
o
n o
f
a
da
pt
i
v
e
P
e
rona
M
a
l
i
k
a
s e
dge pr
e
ser
v
a
tio
n
i
s
d
one i
n
S
R
re
c
ons
t
r
uc
ti
on.
The
e
n
ha
ncem
ent
w
o
rks
in
t
he
I
BP
i
s
c
o
ntin
ue
d
w
i
t
h
r
ed
uc
in
g
t
h
e
r
i
n
g
i
ng
ar
t
e
fac
t
s
b
y
ap
pl
y
i
ng
a
varia
n
t
a
n
d
lo
ca
l
var
i
ance
v
a
r
ian
t
b
a
s
ed
o
n
ada
p
t
i
ve
a
na
l
y
sis
[
13].
M
o
re
ov
e
r
,
the
q
u
a
l
i
t
y
o
f
visua
l
o
u
t
put
re
co
n
s
t
r
u
c
t
i
on
i
m
a
g
e
i
s
i
mp
rov
e
d
t
h
rough
a
pp
lyi
n
g
the
sp
arse
l
i
ne
a
r
r
egre
ssion
a
nd
IBP
.
A
ddit
i
o
n
a
l
l
y
,
modi
fic
a
t
i
on
i
n
G
a
ussia
n
h
i
g
h
pa
ss
f
i
l
t
er
i
s
do
ne
t
o
refine
t
he
i
n
itia
l
gues
s
[
1
4
].
F
ur
th
e
r
m
o
re,
the
a
ccu
r
acy
o
f
rec
ons
t
r
uc
tio
n
c
a
n
be
i
nc
re
ase
d
w
hen
t
h
e
re
sidua
l
error
is
e
nha
nce
d
b
y
i
m
plem
e
n
t
i
ng
t
he
A
n
i
so
trop
ic
D
i
ffus
i
on
w
i
t
h
t
he
h
e
l
p
of
I
BP
e
st
ima
t
o
r
t
o
pro
d
u
c
e
f
ine
H
R
i
ma
ge
w
it
h
e
d
ge
p
r
e
serva
tio
n
t
e
c
h
ni
q
u
e
[15]
.
In
M
ak
wan
a
e
t
al
[
16
]
p
r
opo
se
d
a
co
mb
i
n
a
t
io
n
of
I
B
P
a
lgori
t
h
m
w
it
h
Ca
nny
e
d
ge
d
e
t
e
c
t
or
a
nd
G
a
b
or
f
i
l
t
e
r
for
im
prov
in
g
t
h
e
e
rror
m
i
ni
m
i
za
t
i
on
d
u
rin
g
b
a
c
k
p
roj
e
c
t
i
o
n
ope
r
a
t
i
o
n
.
B
e
s
i
d
e
s
t
h
a
t
,
N
a
y
a
k
e
t
a
l
[
1
7
]
pro
v
i
d
e
d
t
he
e
dg
e
im
age
i
m
pr
ovem
e
n
t
i
n
the
I
B
P
tec
h
n
i
que
b
y
usi
ng
the
P
-
S
p
l
i
ne
a
n
d
M
u
C
S
O
-QP
S
O
alg
o
ri
t
h
m.
T
he
P
-S
pli
n
e
pro
v
i
d
e
s
s
h
ar
p
e
dge
i
nform
a
ti
on
a
nd
M
u
C
SO-QPSO
o
p
t
i
m
ized
t
h
e
est
i
ma
t
i
o
n
m
etho
d.
The
Itera
t
i
v
e
B
a
c
k
P
rojec
t
i
o
n
tec
h
n
i
q
u
e
ha
s
a
b
il
i
t
y
to
m
i
n
imiz
e
t
he
r
ec
o
n
s
t
ruc
tion
error
ef
f
i
c
i
ently
in
iter
a
t
i
v
e
m
a
nner.
T
h
i
s
erro
r
rec
onstruc
ti
o
n
i
s
back
p
r
o
je
c
t
i
n
g
t
o
H
R
g
r
i
d
t
o
c
or
rec
tin
g
the
ne
x
t
e
st
i
m
a
t
i
o
n
ima
g
e.
A
lth
o
u
gh
th
is
t
e
c
h
n
i
que
i
m
p
ro
ve
s
the
im
ag
e
i
n
ten
s
it
y
w
i
t
h
r
e
m
ovin
g
t
he
b
l
u
r
effec
t
.
H
o
w
e
ver
,
t
h
e
i
t
e
rat
i
v
e
r
o
u
t
i
n
e
t
e
nd
s
to
p
rod
u
ce
th
e
ri
ngi
ng
a
rt
ef
act
s
a
l
on
g
t
h
e
edge
s
bec
a
u
s
e
o
f
b
ack
p
r
o
j
e
cti
n
g
is
w
itho
u
t
a
n
y
e
d
g
e
g
u
i
d
a
n
ce
.
In
t
he
p
r
e
sen
t
w
ork,
t
h
i
s
st
ud
y
pr
op
ose
s
t
he
e
dge
e
nha
nce
m
ent
tec
h
niq
u
e
i
ns
i
d
e
t
h
e
It
era
t
i
v
e
Ba
c
k
P
ro
j
e
c
t
i
on
t
e
chni
qu
e.
T
hi
s
ed
g
e
e
nh
an
ce
me
n
t
tec
h
ni
que
i
s
c
o
mb
in
ed
b
e
t
w
een
t
h
e
I
BP
rec
ons
t
r
uc
tio
n
an
d
S
h
ar
p
In
fi
n
i
t
e
S
ymm
e
trica
l
E
x
p
one
n
t
ial
F
i
lte
r
(S
IS
EF
)
.
T
he
S
ISEF
f
unction
a
s
s
har
p
edge
s
de
tec
t
i
o
n
a
n
d
the
n
p
r
e
serve
s
t
he
p
o
s
s
i
b
l
e
e
dge
s
de
t
e
c
t
e
d
by
re-
h
i
g
hl
i
g
h
t
s
the
e
dge
s
i
n
side
t
he
e
s
ti
mat
i
on
p
roc
e
ss
.
C
o
in
c
i
d
e
,
t
h
e
i
n
fo
rmat
i
o
n
suppl
i
e
s
w
o
rk
s
as
e
dge
g
ui
dance
dur
ing
t
h
e
bac
k
p
ro
jec
t
i
o
n
t
o
avo
i
d
a
c
ross
e
dge
p
ro
jec
t
i
o
n
th
us
t
he
r
in
g
i
ng
e
f
fec
t
s
ca
n
be
r
e
m
ove
d.
T
hi
s
sha
r
p
e
d
ge
d
e
t
ec
te
d
ca
p
a
ble
t
o
enha
nc
e
t
h
e
re
con
s
truc
t
i
o
n
H
R
est
i
ma
tio
n
with
r
ed
uce
th
e
err
o
r
rec
ons
t
r
uct
i
on
a
n
d
pr
eservin
g
t
h
e
l
ost
fre
que
nc
y
inf
o
rma
t
i
o
n
w
ith
r
ob
us
t
t
o
w
a
rd
r
ing
i
ng
effect
s a
nd ra
pi
d
e
s
ti
ma
t
i
on
p
ro
ce
ss.
2.
RESEARCH
M
ETH
O
D
2.1.
I
t
erat
i
ve
B
ac
k
Pr
oj
ect
ion
The
m
a
in
i
dea
of
t
he
I
ter
a
ti
ve
B
ac
k
P
r
ojec
ti
on
(IBP
)
t
ec
h
n
i
q
u
e
is
t
o
rec
o
ns
t
r
uct
t
h
e
H
i
g
h
R
es
o
l
u
t
ion
(HR
)
i
mag
e
f
ro
m
ob
se
rvin
g
th
e
mult
ip
l
e
d
e
g
ra
d
e
d
i
npu
t
i
m
a
g
es.
T
he
e
st
ima
tio
n
d
one
b
y
pro
j
ec
ti
ng
b
a
c
k
t
h
e
di
ffe
re
nce
be
tw
een
s
imu
l
a
t
e
d
L
ow
R
es
ol
u
t
i
o
n
(LR) i
m
a
g
e
s
and t
h
e
o
bse
r
ved
L
R
i
ma
ge
s
[1
8].
The
est
i
ma
ti
o
n
kee
p
s
co
nt
i
nue
d
u
n
t
i
l
t
he
H
R
im
age
ge
ne
rat
e
d
w
h
ich
ha
s
sam
e
L
R
ima
g
es
w
it
h
o
b
se
rv
ed
L
R
i
m
age
s
a
fter
pass
in
g
t
h
e
sa
m
e
b
l
u
r
r
i
ng
a
n
d
d
o
w
n
s
am
p
l
i
ng
proce
s
s.
B
a
s
ica
l
l
y
the
IBP
m
e
tho
d
c
a
n
b
e
di
v
i
de
d
i
n
to
t
w
o
pha
ses.
I
n
t
h
e
firs
t
p
h
ase,
i
t
ge
nera
t
e
s
the
si
m
u
late
d
ima
g
e
a
n
d
t
h
en
p
ro
jecte
d
b
ac
k
the
err
o
r
fro
m
th
e
di
ffe
re
nce
be
t
w
ee
n
sim
u
l
a
te
d
an
d
o
b
se
rved
L
R
im
a
g
e
s
t
o
re
co
ns
tr
uct
t
h
e
H
R
e
stim
at
ion
[
19].
The
co
nstr
uct
i
o
n
o
f
s
im
u
l
a
t
e
d
the
L
R
ima
g
es
can
b
e
form
ul
a
t
e
d
in 1.
(1)
Th
is
s
im
ula
t
e
d
L
R
im
age
ha
s
n
u
m
b
er
o
f
im
age
a
nd
the
pi
xel
i
n
te
nsit
y
of
L
R
i
m
age
s
d
e
r
ive
d
from
t
he
H
R
im
age
a
fte
r
g
et
i
n
f
lue
n
ced
by
t
h
e
de
gra
d
a
t
ion
fu
nc
ti
o
n
.
The
de
grad
a
t
i
o
n
fu
nc
ti
o
n
is
a
com
b
i
n
a
t
i
o
n
of
b
l
u
r
ke
rnel
f
unc
t
i
on
a
nd
tr
ans
l
at
ion
w
r
appi
n
g
f
u
n
c
tio
n
inf
l
ue
nce
d
b
efo
r
e
ge
t
t
in
g
t
h
e
dow
n sam
p
l
i
ng
by the
dec
i
m
a
t
i
o
n
o
pera
t
o
r
.
In
t
he
r
e
c
onstr
uct
i
on
o
f
I
BP
p
r
o
cess,
t
h
i
s
techn
i
que
s
tar
t
s
w
i
t
h
com
p
u
t
i
n
g
the
in
i
t
ia
l
gu
ess
H
R
ima
g
e
via
ap
pl
ies
the
ave
r
age
me
th
o
d
.
Then
,
the
i
m
agi
ng
proc
ess
d
e
t
ermine
s
the
set
of
s
imu
l
ate
d
L
R
i
m
a
g
e
.
H
o
w
e
v
e
r,
t
h
e
i
n
i
tia
l
g
u
ess
i
n
g
p
roc
e
ss
m
a
y
pr
ov
ide
m
a
ny
so
l
u
t
i
on
s
tha
n
o
ne
a
c
c
u
rate
s
olu
t
i
o
n.
The
se
le
c
t
i
o
n
of
a
ccur
a
te
t
he
i
nit
i
a
l
gue
ss
me
tho
d
c
o
u
l
d
dri
v
i
n
g
t
h
e
a
l
g
o
ri
t
h
m
to
d
e
t
er
m
i
ne
a
n
u
n
i
q
ue
,
faster
and
sm
o
o
th
s
o
l
u
t
i
o
n.
A
goo
d
selec
t
i
on
o
f
a
n
ini
t
ial
gues
s
me
tho
d
w
a
s
sug
g
es
te
d
b
y
Ira
ni
i
s
t
h
e
ave
r
a
g
e
me
tho
d
o
f
t
h
e
attr
ib
u
t
es of i
n
form
ati
o
n
from
r
egistere
d L
R
ima
g
es o
bse
r
v
e
d [1
8] suc
h i
n
the
2
.
0
1
∑
1
1
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2502-
4752
I
n
do
n
e
si
an
J
E
l
e
c
E
n
g
&
C
o
m
p
S
ci
, V
o
l
.
1
4
,
No. 1, April 2019 :
258 –
2
66
26
0
I
f
t
h
e
i
n
i
t
i
a
l
g
u
e
s
s
H
R
i
m
a
g
e
0
i
s
ma
t
c
hing
e
n
o
u
gh
w
ith
t
h
e
r
e
f
e
r
ence
i
ma
ge,
t
h
e
n
t
he
s
i
m
ula
t
e
d
ima
g
es
sho
u
l
d
be
i
den
t
ica
l
w
it
h
the
obse
r
ved
ima
g
e
s
a
nd
no
mor
e
c
orr
ecti
on
is
requ
ire
d
.
I
f
t
he
d
i
f
fere
nce
im
ages
a
rise
b
e
t
we
en
a
nd
,
then
t
he
i
m
p
ro
ve
m
e
nt
t
he
i
n
itia
l
g
u
e
s
s
by
bac
k
pro
j
ec
ti
ng
t
h
e
e
rror
w
i
th
c
o
m
bi
n
i
n
g
t
he
c
urr
e
nt
i
nform
a
tio
n
in
the
H
R
g
r
i
d.
T
h
i
s
ste
p
i
s
yie
l
di
ng
a
n
i
m
p
r
ov
emen
t
to
t
h
e
n
ext
HR
ima
g
e
1
s
uc
h e
xpr
essed
in
t
he
3.
1
1
∑
1
(
3
)
Whe
r
e
(4)
Th
is
p
roc
e
dur
e
ke
e
p
s
repea
tin
g
i
n
iter
a
tive
ma
nne
r
un
t
i
l
it
m
e
e
ts
t
he
l
oc
al
m
i
n
i
m
um
e
rror.
The
err
o
r fu
nc
t
i
o
n
can
b
e
desc
ribe
d
as in
5.
∑
(
5
)
2.2.
Th
e
In
finite
S
ymmet
r
ical
E
xp
on
en
tia
l
F
i
l
t
er
(IS
E
F)
The
Inf
i
ni
te
S
y
m
m
e
trica
l
E
x
p
o
n
e
n
t
i
a
l
F
il
te
r
or
S
hen
Cas
t
a
n
E
d
g
e
D
e
t
e
c
t
o
r
i
s
a
n
o
p
t
i
m
a
l
s
m
o
o
t
h
i
n
g
fi
l
t
e
r
w
ith
a
v
ery
sim
p
l
e
r
e
c
ursi
ve
a
lg
or
i
t
hm
.
The
co
re
c
onc
ep
t
o
f
th
is
f
il
t
e
r
is
i
mple
me
n
tin
g
t
h
e
di
ffe
re
nt
ia
tio
n
of
o
p
tim
al
e
x
p
one
n
t
i
a
l.
I
t
gi
v
e
s
a
good impa
ct
d
u
r
i
ng
e
dge dete
c
t
ion
w
h
e
n
i
t
ha
s
l
e
ss
r
e
s
pon
ds
to
n
o
i
se
a
n
d
g
i
v
e
prec
i
s
io
n
i
n
e
dge
l
oca
l
iza
tio
n.
G
ene
r
al
l
y
,
t
h
e
p
r
ob
l
e
m
fac
i
ng
o
f
La
pl
a
c
i
a
n
Ga
u
s
si
a
n
f
ilt
e
r
dur
in
g
t
h
e
e
d
ge
d
e
t
e
c
tio
n
w
h
ic
h
i
t
h
as
c
on
tra
d
ic
t
i
o
n
b
etw
e
e
n
t
h
e
n
o
ise
inse
ns
ib
i
lit
y
a
nd
t
h
e
pr
ec
isio
n
loca
l
i
za
ti
o
n
.
Thus
t
o
o
v
erc
o
me
t
his
pro
b
le
m
,
t
he
I
S
E
F
prop
ose
d
the
l
i
ne
a
r
f
ilter
ba
se
d
on
one
s
te
p
mode
l
an
d
mult
i e
dge
de
t
e
c
tio
n.
Th
is
o
pt
ima
l
s
m
oot
h
i
n
g
f
ilte
r
is
a
s
ym
m
e
tric
e
x
p
o
n
e
n
tial
in
i
n
f
ini
t
e
l
a
rg
e
wi
ndo
w
si
ze
a
n
d
ve
ry
si
m
p
le
r
ec
ursi
ve
a
l
gori
t
hm
.
The
d
i
ffe
r
enc
e
betw
ee
n
i
n
pu
t
a
nd
o
u
t
p
ut
o
f
th
is
r
e
c
ursi
ve
f
il
t
e
r
has
the
c
a
pab
i
lit
y
to
d
e
t
ec
t
the
e
dge
s
w
i
th
l
es
s
sens
i
tiv
it
y
t
o
n
o
i
se
a
n
d
h
ig
h
pre
c
i
si
on
i
n
l
oca
l
i
z
at
i
o
n.
T
he
I
S
E
F
fi
l
t
er
h
a
s
pro
pose
d
t
w
o
m
etho
ds
f
o
r
e
dge
d
e
t
e
c
t
i
o
n
eithe
r
c
a
n
b
e
u
s
ed
t
he
ma
xi
ma
o
f
t
h
e
g
r
ad
i
e
n
t
(
GEF)
o
r
t
h
e
ze
ro
crossi
n
g
o
f
sec
o
n
d
d
irec
tio
na
l
D
e
r
i
va
tive
a
l
ong
t
he
G
r
a
die
n
t
(S
D
E
F
)
.
T
h
e
IS
EF
f
i
l
t
e
r
pro
duce
d
t
he
G
EF
a
n
d
SDE
F
f
ro
m
t
h
e
fi
rs
t
and
se
co
nd
d
e
r
i
v
a
t
iv
es
o
f
t
h
e
o
p
erato
r
f
o
r
sy
mmet
ri
c
e
xpo
n
e
nt
i
a
l
fi
lt
e
r
s
resp
e
c
ti
v
e
ly
.
Ini
t
i
a
lly,
the
no
r
m
a
l
iz
ed sym
me
tric e
xp
one
nt
i
a
l
fi
l
t
e
r
in
o
n
e d
im
en
si
ona
l
(1D
)
is wr
i
tte
n
i
n
6
.
(
6
)
W
h
ere
(7)
The
exp
one
n
t
i
a
l
i
s
a
s
ym
me
t
r
ical
f
u
n
ct
i
on
w
h
i
c
h
is
can
b
e
div
i
d
e
d
i
n
t
o
the
righ
t
side
f
u
n
ct
i
on
and
t
h
e
l
e
ft
s
id
e
func
t
i
o
n
.
T
hus
t
he
s
e
c
o
n
d
d
eri
v
at
i
v
e
of
e
xp
o
n
en
t
i
al
f
un
cti
on
for
the
SDE
F
alon
g
the
x
axi
s
c
an de
s
cri
b
e
i
n
8
.
,
,
∗
,
(
8
)
Whe
r
e
in
t
he
d
iscre
t
e f
o
rm
of SD
EF
is wr
i
tte
n
in
9
.
,
,
∗
1
∗
2
∗
1
2
,
∗
1
∗
2
(
9
)
The
e
qua
t
i
o
n
ab
o
v
e
a
l
so
c
an
b
e
use
d
t
o
ca
l
c
u
l
a
t
e
t
h
e
grad
ie
nt
a
lo
n
g
t
he
y
a
xi
s
di
r
e
c
t
i
o
n.
F
o
r
inst
a
n
ce,
thi
s equa
tio
n sc
hem
e
c
an
b
e il
l
u
stra
ted i
n
t
he
p
r
oce
ss flow
F
ig
ure
1.
2.3.
Pr
e
vi
ou
s Wor
k
s
in
E
d
g
e
En
h
a
n
cem
e
n
t Tec
h
n
i
q
u
e
The
ed
ge
e
n
h
anc
e
m
e
n
t
t
e
c
h
n
i
que
p
ro
vi
des
ad
d
iti
ona
l
c
o
rre
ct
ion
i
n
f
orm
a
tio
n
d
u
ri
ng
th
e
rec
ons
t
r
uc
tio
n
com
p
ut
a
t
ion
base
d
on
t
he
g
uide
d
e
d
g
e
i
nfor
ma
t
i
o
n
d
etec
te
d.
T
he
i
nfor
ma
t
i
o
n
o
f
e
d
ge
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
ones
i
a
n
J
E
lec
En
g & Co
mp
S
c
i
IS
S
N
: 2502-
47
52
Ed
ge
e
n
h
a
n
ce
m
e
nt
o
f
IBP re
con
s
tr
uc
tio
n b
y
u
s
i
ng
sh
ar
p
i
n
f
i
ni
te sym
m
e
tric
al
e
x
p
o
n
en
t
i
a
l… (A.R.
A
Nazren)
26
1
detec
t
e
d
,
t
h
e
n
i
s
use
d
t
o
stre
ng
t
h
en
b
ac
k
t
h
e
l
o
st
e
dge
i
nform
a
ti
on.
A
d
d
iti
ona
l
l
y
,
t
h
i
s
p
roce
ss
pro
v
i
d
e
s
an
edge
g
u
i
de
d
t
o
t
he
I
BP
t
ec
hn
i
que
d
urin
g
t
h
e
ba
ck
p
r
o
je
cti
ng
to
HR
g
rid
si
ze
t
o
av
oi
d
a
ri
ngi
ng
e
f
fe
ct
di
st
urba
nce.
F
urt
h
er
mor
e
,
the
e
ffic
i
e
n
c
y
o
f
rec
onst
r
uct
i
on
p
roc
ess
is
i
n
f
lue
n
c
e
d
b
y
the
pr
ecisi
o
n
t
he
e
dg
e
detec
t
i
on
pr
o
v
ide
d
.
Th
is
h
i
g
h
pr
eci
si
o
n
o
f
l
o
c
a
l
i
z
at
i
on
a
nd
h
i
g
h
rate
d
etec
ti
o
n
a
re
i
mpor
t
a
nt
f
e
a
t
u
res
to
incre
a
se
t
he
r
ate
o
f
e
d
g
e
de
t
e
cti
on
an
d
at
t
h
e
s
am
e
t
i
m
e
,
i
t
h
e
lpe
d
t
o
i
n
cr
ease
the
ra
t
e
o
f
cor
r
ec
t
i
o
n
p
r
o
c
e
ss.
Th
is
s
tud
y
m
ee
ts
t
w
o
g
o
o
d
t
ec
hn
iq
ue
h
a
v
e
i
m
plem
en
ted
w
i
th
I
B
P
i
n
edge
e
nha
nce
m
ent
te
c
h
n
i
qu
e
suc
h
A
n
iso
t
rop
i
c
D
i
ffus
io
n a
nd
ISEF fil
te
r.
F
i
gure
1.
IS
E
F
proce
s
s
fl
ow
2
.
4
.
Pro
p
o
se
I
BP t
ech
nique
w
ith
Sha
rp ISE
F
reg
u
l
a
t
o
r
Thi
s
s
tu
dy
h
a
s
c
on
ce
rn
e
d
o
n
t
h
e
e
d
g
e
d
e
t
ecti
o
n
i
s
p
rovi
d
e
d
by
t
he
I
S
E
F
me
t
h
o
d
.
T
h
i
s
S
hen-Cas
t
a
n
me
tho
d
h
a
s
s
o
m
e
adva
n
t
a
g
es
s
uc
h
as
y
i
e
ld
s
ed
ge
d
e
t
a
i
l
i
n
form
ati
o
n
w
i
t
h
l
e
ss
re
spon
ds
t
o
noi
s
e
a
nd
hi
gh
prec
isio
n
e
d
ge
l
oca
liza
tio
n.
T
hi
s
s
t
u
d
y
a
ls
o
i
n
ter
e
s
t
s
t
o
u
se
e
dge
d
e
t
ec
tio
n
t
h
r
o
u
gh
z
e
ro
c
r
o
ssi
ng
p
r
o
v
i
d
e
d
b
y
the
Sec
o
n
d
D
i
r
e
c
tiona
l
D
e
r
i
v
a
tive
F
i
l
lter
(SD
E
F).
This
f
ilt
er
i
s
ba
se
d
on
e
xp
o
n
en
t
i
a
l
f
unc
t
i
o
n
w
hic
h
g
i
v
e
s
bet
t
er
p
e
rform
anc
e
t
ha
n
t
h
e
Ga
ussia
n
b
a
s
e
d
f
il
ter.
H
ence
,
this
t
e
ch
n
i
q
u
e
c
ou
l
d
p
ro
vi
d
e
m
uch
be
tter
e
dge
p
r
e
s
erv
i
ng
[
20].
Th
e
hi
gh
a
c
c
u
r
a
c
y
in
S
DEF
d
e
t
e
ct
ion
sup
p
l
i
e
s
m
u
c
h
pos
sib
l
e
a
d
di
tio
na
l
i
n
f
o
r
m
a
t
i
o
n
espec
i
al
ly
a
b
o
u
t
hi
g
h
f
re
que
ncy
inform
at
i
o
n.
T
h
i
s
pre
c
i
sion
o
n
i
n
form
ati
o
n
he
lps
the
est
i
m
a
t
o
r
to
m
a
k
e
pred
ic
ti
ons a
nd
p
rese
rves the
w
e
a
k a
nd l
o
ss
o
f
ed
ges m
o
re
e
ffic
ie
n
t
.
As
a
forem
e
ntioned
earlier,
t
he
I
B
P
m
et
hod
p
erfo
rmed
w
e
l
l
in
m
in
imiz
ing
t
h
e
resi
dua
l
er
ror
dur
i
n
g
the H
R
est
i
m
at
io
n. The es
t
i
m
ati
o
n
is d
one
b
y bac
k
pr
o
jec
t
i
ng t
h
e
di
ff
ere
n
c
e
o
f
ob
se
rv
ed
i
ma
g
e
a
nd
s
i
m
u
l
at
ed
ima
g
e
t
o
t
he
H
R
c
o
rre
cti
o
n
proce
s
s.
T
h
i
s
iter
a
t
i
ve
p
r
o
c
e
ss
ke
e
p
s
on
r
e
p
ea
ti
ng
unt
i
l
n
o
mo
re
e
rro
r
co
s
t
fu
nc
ti
o
n
a
rises
w
itho
u
t
a
n
y
e
dge
g
ui
da
nc
e.
C
onse
q
uen
t
ly,
est
i
ma
t
e
d
t
h
e
H
R
b
y
the
I
B
P
alg
o
ri
thm
is
n
ot
a
b
l
e
to
f
ull
y
r
ec
o
v
e
r
t
he
l
ost
ed
ge
i
nform
a
ti
on.
T
h
u
s,
t
he
I
S
E
F
me
th
o
d
i
s
the
be
st
s
o
l
uti
o
n
t
o
i
mpr
o
ve
t
he
d
e
-
bl
urrin
g
proc
e
s
s
and
sharpe
n t
h
e e
dge
d
e
t
a
i
l
s
.
This IS
E
F
imple
m
e
n
ta
tio
n c
a
n
b
e
il
l
us
tra
t
e
d
a
s
in t
he F
ig
ure
2.
A
c
c
o
rdi
n
g
t
o
t
he
p
r
o
p
o
se
d
al
gor
ithm
i
n
F
i
g
ur
e
3,
t
he
i
n
i
t
i
a
l
g
u
ess
ge
ts
t
he
f
irst
e
n
h
anc
e
m
e
nt
f
r
o
m
the
im
p
l
e
m
e
n
te
d
I
S
EF.
This
p
r
o
c
e
ss
is
i
m
p
orta
nt
s
i
n
ce
th
e
o
b
ser
ve
d
L
R
w
il
l
be
m
a
g
n
i
fi
e
d
t
o
t
h
e
H
R
g
ri
d,
th
us
s
om
e
in
fo
rm
atio
n
w
i
ll
b
e
l
ost
an
d
b
l
ur
ry
i
m
a
ge
i
s
ge
nera
t
e
d.
A
lso,
t
h
i
s
sta
g
e
g
i
ve
s
impro
v
e
m
e
n
t
i
n
t
h
e
qua
l
ity
o
f
i
n
i
t
i
a
l
g
u
ess
ima
g
e
to
g
e
t
c
lose
t
o
the
p
o
ss
i
b
le
t
o
o
p
t
i
m
um
e
s
t
i
m
atio
n.
A
s
the
H
R
i
n
itia
l
g
u
e
ss
has
go
o
d
qua
l
i
t
y
,
the
IBP
re
co
ns
t
r
uct
i
o
n
onl
y
re
qu
i
r
es
f
e
w
e
ff
orts
for
en
ha
nc
in
g
proce
ss.
B
e
s
ide
s
i
nc
rea
s
i
n
g
t
h
e
qua
l
ity
o
f
H
R
e
stim
ati
o
n,
t
h
i
s
stage
a
l
s
o
c
o
n
tr
ib
u
t
e
s
t
o
sa
ve
a
l
o
t
i
f
r
e
s
o
u
r
c
e
s
i
n
a
f
a
s
t
e
r
w
a
y
.
T
h
e
f
i
r
s
t
r
e
f
i
n
e
proce
s
s
i
n
t
he
i
ni
t
i
a
l
g
uess c
a
n
be
e
xpresse
d
i
n
1
0.
∗
(
1
0
)
Whe
r
e
i
s
con
s
ta
nt
f
or
t
he
s
ta
bi
lit
y
e
q
ua
tio
n.
i
s
the
first
i
n
i
tia
l
g
u
ess
pr
ojec
te
d
fr
o
m
t
he
o
b
s
er
ved
L
R
ima
g
es
b
y
us
i
n
g
t
h
e
a
v
era
g
e
me
t
h
o
d
.
In
o
rd
er
t
o
en
hanc
e
the
,
the
S
D
EF
m
etho
d
i
n
t
he
I
S
E
F
obse
r
ves
the
pri
o
ri
i
nf
orma
tio
n
i
n
it
ial
gu
ess
to
p
r
o
d
u
c
e
t
he
g
ra
die
n
t
inform
ati
o
n.
T
his
gr
adie
n
t
c
o
n
ta
i
n
s
a
n
e
dge
in
form
ation
w
h
ic
h
is
t
hen
fuse
d
i
n
s
i
de
t
h
e
c
urr
e
nt
e
st
im
atio
n
i
m
a
ge
.
Th
is
a
d
d
i
tio
n
a
l
e
dge
i
nf
or
ma
tio
n
pro
v
i
d
es r
ecov
e
rin
g
o
f
the
w
e
aker
fr
e
que
nc
y
inform
ati
o
n and
s
t
r
eng
t
h
e
n
s
bac
k t
h
e
st
r
u
c
t
ure
i
m
age
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
250
2-
4
7
5
2
In
do
n
e
sia
n
J
Elec Eng
&
C
o
mp
S
ci,
Vo
l. 1
4
,
No
. 1
,
Ap
r
i
l
2
0
1
9
:
258
–
2
6
6
26
2
B
a
si
c
a
lly
,
t
h
e
resi
du
al
i
nf
o
r
ma
t
i
o
n
i
s
g
en
era
t
e
d
f
ro
m
th
e
di
ff
e
r
e
n
c
e
of
s
i
m
ul
a
t
ed
a
nd
o
b
s
e
r
v
e
d
LR
im
ages
w
h
i
c
h
a
r
e
u
sed
to
c
or
r
e
ct
t
he
c
ur
r
e
nt
H
R
es
ti
m
a
ti
o
n
.
M
e
a
nw
h
i
le,
the
c
o
r
r
e
ct
io
n
pr
oc
ess
c
a
n
b
e
op
t
i
mize
d
w
h
e
n
a
dd
i
n
g
the
e
x
tr
a
e
d
ge
i
n
f
or
m
a
tio
n
.
I
n
P
at
el
[
2
1
]
w
or
k,
t
he
a
dd
it
io
na
l
gr
adie
nt
i
n
f
or
m
a
tio
n
is
dir
e
c
tly
f
use
d
i
ns
ide
the
r
e
si
d
u
al
i
n
f
or
m
a
ti
o
n
.
H
o
w
e
ve
r
,
t
his
a
pp
ro
ach
h
as
s
o
m
e
d
i
ff
ere
n
ce
s
sin
c
e
add
i
tio
nal
in
form
atio
n
ne
e
d
s
a
s
t
ep
s
iz
e
or
w
eig
h
ti
n
g
f
act
or
t
o
en
sur
e
t
h
i
s
c
o
m
b
ina
tio
n
co
n
v
erge
s
to
t
h
e
o
p
tima
l
l
oc
a
l
mi
n
i
ma
f
a
s
t
e
r
.
T
h
e
s
uit
a
bl
e
v
a
l
u
e
of
s
t
e
p
si
z
e
i
s
c
r
u
ci
al
c
r
i
t
e
r
i
a
t
o
e
ns
ur
e
the
e
s
t
i
m
a
tio
n
ac
hie
v
e
s
o
p
t
i
m
um
HR
o
u
t
pu
t.
B
e
s
ide
s
t
ha
t,
it
als
o
g
ive
s
i
mpr
o
ve
me
n
t
i
n
the
n
u
m
b
e
r
o
f
r
epe
tit
ion
s
o
n
of
t
he
s
ystem
.
Th
is
c
alcu
la
t
i
o
n
f
or
s
te
p
size
c
a
n
b
e
r
e
fer
r
ed
i
n
11.
0
‖
‖
(
11)
Wh
ere
i
s
t
h
e
Hessi
an
f
un
ct
io
n
o
f
m
at
ri
x
of
f
a
t
x
and
‖
∙
‖
i
s
the
s
p
ec
tra
l
o
p
e
rator
norm
whic
h
us
ed
t
o
fi
n
d
t
he
l
a
r
ges
t
e
ige
n
val
u
e
o
f
f
unc
t
i
on
f.
T
he
p
ar
am
eter
i
s
kn
ow
n
a
s
t
h
e
s
t
e
p
size
f
or
g
r
a
dien
t
of
i
m
a
ge
ed
g
e
s c
o
mp
ut
ed
.
In
o
r
d
e
r
t
o
d
e
t
e
rmi
n
e
th
e
opti
m
a
l
r
e
s
ul
t
,
t
hi
s
op
ti
mal
re
sidu
a
l
i
nf
or
m
a
ti
on
is
t
he
n
pr
oj
e
c
ted
ba
c
k
t
o
H
R
g
r
i
d.
T
h
i
s
sce
n
ar
i
o
c
an
b
e
de
scr
i
be
d
i
n
m
a
t
hem
a
t
i
c
a
l
a
s
s
uch
1
2
.
∗
∗
(
12)
I
n
P
atel
w
or
k
s
,
t
h
e
e
n
h
a
nce
m
e
n
t
of
t
he
e
dge
s
is
t
ake
n
f
r
o
m
t
h
e
d
i
f
f
e
r
enc
e
o
f
c
om
p
u
te
d
i
n
it
ial
gues
s
a
n
d
c
u
rre
nt
H
R
.
T
h
i
s
d
i
ffer
ence
p
r
o
v
i
des
a
gu
i
d
anc
e
t
o
the
e
dge
i
n
t
e
n
si
ty
a
nd
t
h
e
p
r
ec
i
s
i
on
i
n
t
he
e
dge
loca
liza
tio
n.
H
owe
v
er,
the
p
r
ecis
i
o
n
e
dge
l
oca
liza
tio
n
ca
n
be
i
ncr
ease
d
by
s
h
r
i
nk
i
n
g
the
w
i
d
t
h
e
d
g
e
s
to
b
e
c
o
me
s
h
a
rp
e
r
.
T
h
u
s
,
t
h
e
accu
ra
c
y
i
n
ed
g
e
l
o
c
a
l
i
z
a
tio
n
co
nt
rib
u
t
es
t
o
mo
re
a
cc
u
r
at
e
e
s
ti
mat
i
o
n
,
a
s
wel
l
a
s
a
n
i
ncr
e
me
n
t
i
n
t
h
e
qua
l
i
t
y
of
o
u
t
pu
t
an
d
decr
ease
t
h
e
n
u
mbe
r
o
f
iter
a
tio
ns
f
or
e
st
ima
t
i
o
n
pro
cess.
The
shar
pe
r
e
d
ge
p
r
o
c
e
ss
ca
n
be
d
o
n
e
by
su
btr
a
c
tin
g
the
d
i
f
f
er
e
nce
edge
w
i
t
h
e
dge
d
e
t
e
c
ted
f
r
o
m
the
pr
ev
i
ous
s
t
a
te
H
R
im
age
.
Both
o
f
de
tect
e
d
e
dges
a
r
e
de
t
e
r
m
ine
d
by
t
h
e
S
D
EF
m
ethod
j
u
s
t
b
e
f
o
r
e
t
h
e
su
bt
rac
tio
n o
p
e
ra
ti
on
. Th
is si
t
u
a
ti
o
n
c
a
n
b
e
i
llu
st
ra
t
e
d
su
ch
in
F
i
gu
r
e
2
.
(a)
(b
)
(
c
)
Fi
g
u
re
2
.
Sh
a
r
p
e
r
g
r
a
d
i
e
n
t
i
ma
g
e
(
a
)
d
i
f
f
e
r
e
n
c
e
e
d
g
e
o
f
H
R
(
b
)
edge
o
f
p
r
i
o
r
i
H
R
(
c
)
pr
opose
d
s
har
p
e
dge
3.
RESULTS
A
ND
A
NAL
YS
IS
Thi
s
s
ect
ion
pr
e
s
e
n
t
s
t
h
e
e
xp
e
r
i
m
e
n
t
s
a
nd
re
su
lt
s
obt
ai
ne
d
by
t
he
p
r
o
pos
ed
c
om
b
i
n
a
t
i
o
n
b
e
t
w
e
e
n
I
B
P
w
i
t
h
S
I
S
EF
e
dge
e
nha
nc
em
ent
tec
hni
q
u
e.
T
hes
e
e
xpe
r
i
m
e
nts
u
s
e
the
M
A
TLA
B
platfo
rm
a
s
exp
e
rimen
t
simu
lat
o
r
.
T
he
se
e
x
p
er
im
en
ts
a
ls
o
use
the
s
t
a
ndar
d
i
m
a
ge
f
or
m
I
S
I
a
nd
D
E
S
C
A
I
d
a
t
a
b
a
s
e
w
i
t
h
t
he
s
ta
ndar
d
im
age
si
z
e
i
s
25
6
x25
6
r
e
sol
u
t
i
on.
F
i
r
st
l
y
,
th
i
s
e
xper
i
m
e
nt
c
r
e
a
tes
a
se
q
u
ence
f
o
u
r
LR
i
m
a
ges.
E
a
c
h
of
L
R
im
ages
h
as
s
hif
t
e
d
f
r
o
m
or
ig
ina
l
i
m
a
ge
by
a
pi
xe
l
in
t
he
v
e
r
t
i
c
a
l
d
ir
e
c
t
i
o
n
.
T
hen,
t
he
se
i
m
a
ge
s
a
p
plie
d
e
ffe
c
t
of
c
a
m
e
r
a
PS
F.
T
he
e
f
f
e
c
t
s
i
n
v
o
l
ve
d
i
n
t
h
i
s
pr
oc
e
ss
c
ont
a
i
n
o
f
blur
r
y
a
nd
dec
i
m
a
t
i
o
n
o
p
e
r
at
io
n.
T
h
e
n
th
i
s
sh
if
ted
L
R
i
m
a
ge
c
o
nvo
lu
te
d
w
ith
t
he
G
a
u
ssia
n
l
ow
p
a
ss
fi
lt
e
r
o
f
size
3
x
3
w
it
h
a
st
anda
r
d
d
e
v
i
a
ti
on
of
3
.
The
n
r
esul
t
i
n
g
im
age
w
a
s
dow
n
sa
mpli
n
g
b
y
fa
ct
or
2
i
n
e
ach
d
ir
e
ction
.
T
h
e
ref
o
re,
th
e
same
o
p
e
rat
i
on
w
ith
t
h
e
di
ffe
rent
m
o
t
io
n
v
e
c
t
o
r
s
in
v
e
r
t
i
c
a
l
a
n
d
ho
ri
zo
nt
al
d
i
r
ec
ti
on
s
i
s
u
sed
t
o
p
rodu
c
e
f
ou
r
LR
i
ma
g
e
s.
The
pr
o
pose
d
t
ec
hni
que
m
e
a
sur
e
me
n
t
b
a
s
ed
o
n
t
h
e
h
i
g
h
es
t
P
S
N
R
r
e
c
o
rd
ed
.
B
e
sid
e
s
t
h
at
,
t
h
e
c
o
st
Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesia
n
J
Elec Eng
&
C
o
m
p
S
ci
ISSN:
2502-
4752
Ed
ge
enh
a
n
ce
m
e
nt o
f
I
BP re
c
o
n
s
tr
uc
ti
o
n
b
y
using
sh
arp
inf
i
n
i
t
e
sym
m
et
ri
c
a
l
ex
po
ne
n
t
i
a
l
…
(
A
.
R
.
A
N
a
zren)
26
3
c
o
mpu
t
a
t
io
n
c
o
mple
x
i
t
y
m
e
a
s
ur
em
ent
is
u
s
e
d
ba
se
d
o
n
n
u
m
ber
of
i
ter
a
ti
ons
c
ons
umi
n
g
f
o
r
f
i
n
i
s
h
in
g
t
h
e
H
R
e
s
t
i
m
a
tio
n.
T
h
e
r
e
s
ults
a
n
d
a
nal
y
s
i
s
a
r
e
e
q
u
i
p
p
e
d
t
he
c
om
par
i
s
o
n
wit
h
t
he
C
h
e
r
e
f
w
o
rk
[
11
]
and
Pa
t
e
l
wo
r
k
[2
1]
a
lo
ng
t
h
e
w
o
r
k
.
3
.
1
. Situ
a
tio
n
:
Bl
u
r
ry
Effect
I
n
f
o
r
ma
tio
n
P
erf
ect
l
y
K
no
w
n
The
r
e
sul
t
i
n
T
a
ble 1
s
h
ow
s
tha
t
t
he
c
om
b
i
n
a
t
i
o
n
I
BP
w
ith pr
o
po
se
d t
e
ch
n
i
q
u
e
S
h
ar
p
I
S
EF
r
e
gul
a
t
or
(SIS
E
F)
is
l
ead
i
ng
f
o
r
m
o
s
t
of
PSN
R
resu
lts
.
B
y
look
in
g
at
t
h
e
ave
r
age
re
su
l
t
,
th
i
s
t
ec
h
n
iq
ue
g
i
v
es
a
n
i
m
p
r
ov
e
m
e
n
t
a
b
out
1
.7
5
dB
c
o
m
p
a
re
s
th
a
n
o
r
d
in
a
r
y
I
B
P
t
e
c
hni
qu
e
an
d
th
is
r
esu
lt
i
s
a
h
ig
her
q
u
a
l
i
t
y
im
pro
v
em
ent
i
s
p
r
o
v
i
de
d
c
o
m
p
ar
e
t
o
o
t
h
e
r
t
e
c
h
n
i
q
u
es.
Thi
s
m
eans
,
the
SISE
F
propos
ed
t
echnique
h
as
t
he
c
a
pab
i
l
i
t
y
t
o
pr
o
v
ide
a
be
t
t
e
r
e
dge
e
n
h
a
n
c
e
m
e
nt.
Th
is
e
dge
d
e
t
e
ct
or
e
qu
i
ppe
d
w
i
t
h
h
i
g
her
l
o
c
a
l
iz
ati
o
n
pr
e
c
is
i
o
n
a
n
d
l
e
ss
er
r
o
r
de
tect
io
n,
t
her
e
f
o
r
e
i
n
mos
t
c
a
s
e
s
o
f
i
m
a
g
e
t
e
st
e
d
t
hi
s
p
r
opo
sed
IS
EF
h
a
s
p
roduc
ed
a
be
tter
per
f
or
m
a
nc
e.
Bes
i
de
s
t
h
at,
by
c
o
m
p
ar
ing
w
i
t
h
t
he
P
a
t
el
t
e
c
h
ni
q
u
e.
T
his
pr
o
p
o
sed
t
e
chni
qu
e
su
cc
es
sf
u
lly
t
o
incr
ease
t
h
e PS
NR
u
p t
o
0.
7
5
8
dB
t
h
a
n
Pa
t
e
l
tech
n
i
q
u
e.
F
u
r
therm
or
e
,
by
r
e
fer
r
ing
in
F
i
g
ur
e
3,
t
h
i
s
S
I
SEF
ha
s
su
c
c
e
s
sf
ull
y
i
nc
re
as
ed
t
h
e
p
re
ci
si
on
a
nd
accu
ra
cy
o
f
ed
ge
p
ri
o
r
i
i
n
for
m
a
t
io
n
fr
om
t
he
c
o
nve
n
tio
n
a
l
I
S
EF
m
e
tho
d
i
n
P
a
t
e
l
work.
Thus,
th
is
c
o
n
tr
ibu
t
i
on
hel
p
s
t
h
e
I
B
P
e
s
t
i
m
a
t
o
r
p
redi
ct
s
t
h
e
H
R
c
on
st
ru
c
tio
n
mo
re
pr
e
c
ise
l
y
a
nd
ac
cur
a
t
e
l
y
.
A
d
di
t
i
o
n
a
lly,
F
i
g
u
r
e
4
h
a
s
s
ho
w
n
t
he
S
I
S
EF
h
as
s
uc
cess
f
ul
l
y
r
e
d
uce
d
t
he
r
in
gi
n
g
a
r
tefa
c
t
s ar
ou
n
d
t
he
e
d
g
es
o
f
t
h
e
H
R
ima
ge
c
om
par
e
d w
i
th
t
he
o
r
d
i
na
r
y
I
BP
tec
hn
i
q
ue
.
P
e
r
haps,
t
h
i
s
sit
u
at
i
o
n
he
lps t
o
i
nc
rea
s
e the
ap
pe
ara
n
ce
of H
R
i
m
age
.
D
e
spi
t
e
of
t
he
g
oo
d
per
f
o
r
m
a
n
ce,
t
hi
s
t
ech
ni
q
u
e
has
di
ff
i
c
ul
t
y
d
u
r
i
ng
h
a
n
dli
n
g
th
e
crit
i
cal
i
ma
ge
suc
h
t
he
i
m
a
g
e
c
on
ta
in
s
a
l
o
t
of
h
ig
h
fr
eq
ue
nc
y
i
n
f
o
r
m
a
t
i
o
n.
F
o
r
e
x
amp
l
e,
t
h
e
M
and
r
il
l
imag
e
c
o
nt
ai
ns
a
l
ot
of
h
i
g
h
fre
que
ncy
inform
at
ion
suc
h
h
a
i
r
a
r
ou
n
d
t
he
m
a
n
d
r
ill
fac
e.
T
his
d
i
sa
d
v
a
ntag
e
ro
se
w
h
e
n
th
e
SISEF
de
t
e
c
t
or
onl
y
t
r
ac
es
t
he
e
d
g
es
i
n
hor
iz
o
n
ta
l
a
nd
ve
r
t
i
c
al
d
ir
ec
ti
o
n
o
n
l
y.
H
enc
e
,
in
s
o
m
e
di
rec
tion
th
i
s
d
et
ec
t
o
r
fa
i
l
e
d
t
o
pr
o
v
i
de
d
eta
i
ls
h
ig
h
fr
eque
nc
y
in
f
o
r
m
atio
n.
F
i
gur
e
3.
T
he
I
BP
w
i
t
h
lor
e
n
t
zian
I
S
E
F
r
e
gul
a
t
or
p
r
o
c
e
ss
f
l
ow
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2502-
4752
I
n
do
n
e
si
an
J
E
l
e
c
E
n
g
&
C
o
m
p
S
ci
, V
o
l
.
1
4
,
No. 1, April 2019 :
258 –
2
66
26
4
(a)
(b)
F
i
gur
e 4.
Ri
ngi
ng
artefac
t
s (
a
)
IBP
ou
t
p
ut c
o
n
t
a
mi
na
te
d
w
ith
t
h
e r
i
ng
i
ng a
r
tefa
c
t
s (b)
prop
ose
d
t
ech
n
i
q
u
e
h
a
s
r
e
duce
d
the
r
in
gi
n
g
a
rtefa
c
t
s
Ta
bl
e 1
.
P
S
N
R
f
o
r M
u
l
t
i
Ki
nds o
f In
p
u
t
Imag
e
T
he
I
B
P
t
ec
hniqu
e
w
ith
I
B
P
Pa
te
l
C
h
e
r
e
f
C
A
NNY
SISE
F
L
E
N
A
27.
730
28.
651
28.
034
28.
018
28.
952
C
A
M
E
R
A
M
A
N
26.
477
27.
220
27.
319
26.
758
27.
224
M
A
NDR
IL
23.
284
23.
464
23.
486
23.
336
23.
273
PE
PP
ER
25.
941
27.
956
29.
023
26.
391
30.
253
C
L
O
W
N
24.
560
26.
673
27.
932
25.
067
28.
784
JE
T
P
LAN
E
26.
020
27.
177
27.
571
26.
353
27.
648
BR
A
I
N
28.
041
30.
131
31.
319
28.
699
32.
533
SH
IP
28.
923
29.
494
29.
714
29.
188
29.
435
U
S
A
F
22.
670
22.
183
22.
624
22.
656
22.
183
BR
ID
GE
23.
982
24.
648
24.
824
24.
178
24.
890
O
n
t
he
o
t
h
e
r
h
and,
i
n s
o
me
c
a
s
es t
he
tech
n
i
que
p
ro
p
o
se
d b
y
C
he
r
ef a
ls
o p
r
ovide
s
t
h
e hi
g
h
es
t q
u
a
l
i
t
y
.
Th
is
t
e
c
h
ni
qu
e
perform
ed
w
ell
d
u
ri
n
g
h
an
d
l
i
ng
d
i
ff
ic
ult
im
age
s
uc
h
M
a
ndri
l
l
and
U
S
A
F
i
m
a
ge
.
The
A
n
i
s
o
t
r
o
p
i
c
D
i
ffusi
on
in
C
here
f
w
o
rks
has
t
h
e
c
a
p
ab
il
i
t
y
to
s
e
e
k
t
h
e
e
d
g
e
i
n
m
a
n
y
d
i
r
e
c
t
i
o
n
s
(
n
o
r
t
h
,
s
o
u
t
h
,
w
e
s
t
,
a
n
d
e
a
s
t
)
.
T
h
u
s
,
t
h
i
s
m
e
t
h
o
d
h
a
s
t
h
e
c
a
p
a
b
i
l
i
t
y
t
o
p
r
o
v
i
de
t
he
d
e
t
a
il
e
dge
i
nfor
ma
t
i
on
a
n
d
g
i
ve
a
prom
i
s
in
g
r
e
su
l
t
i
n
the
d
i
ffic
ul
t
im
age.
M
o
r
eove
r,
t
hi
s
e
d
ge
d
e
tect
or
h
a
s
g
o
od
r
a
te
d
e
t
ect
i
on
a
nd
mi
nim
u
m
respo
nds
t
ow
a
r
d re
duc
ti
o
n
ed
g
e
s
.
Thus,
th
i
s
t
e
c
h
n
i
q
u
e
i
s
a
m
o
ng
t
h
e be
st
e
dge
e
n
h
a
n
c
e
m
ent tec
h
n
i
que
.
The
G
a
uss
i
a
n
b
ase
d
f
i
l
t
e
r
d
e
s
i
gn
s
u
ch
a
s
Ca
nn
y
te
c
h
niq
u
e
s
o
n
l
y
pro
v
i
d
es
a
lit
t
l
e
im
p
r
o
v
em
ent.
Th
is
c
on
d
i
t
i
on
h
a
ppe
ns
w
h
e
n
the
G
a
uss
i
a
n
b
ased
f
il
ter
has
a
c
o
n
tra
d
ic
tio
n
fu
nc
t
i
o
n
b
etw
e
e
n
p
r
o
vi
din
g
smoo
th
f
u
n
c
t
i
on
a
n
d
pr
ecise
l
oca
liz
a
tio
n.
F
urther
more
,
the
Can
n
y
e
d
g
e
d
e
t
e
c
t
o
r
c
a
n
p
r
o
v
i
d
e
a
v
e
r
y
s
h
a
r
p
edge
i
m
a
ge,
bu
t
the
in
te
nsi
t
y
e
dge
d
e
t
ec
t
e
d
i
s
s
ma
l
l
.
N
o
r
m
a
l
l
y
,
t
h
i
s
t
ype
o
f
fil
t
er
h
as
l
es
s
pr
eci
se
l
oca
l
i
z
a
tio
n
th
an
I
SEF f
i
lter.
3.2.
N
u
m
b
e
r
of
I
ter
a
tion
s
In
t
er
ms
o
f
the
com
p
uta
t
i
o
na
l
com
p
le
x
ity,
the
IBP
w
i
th
e
dge
e
n
h
a
n
c
e
m
e
nt
t
ec
h
n
i
q
ue
m
e
a
surem
e
nt
is
b
ased
o
n
th
e
numbe
r
of
i
tera
t
i
o
n
a
lgor
it
hm
f
or
acc
ompl
is
hin
g
t
he
e
s
t
i
m
at
i
o
n.
T
he
m
a
i
n
i
d
e
a
s
for
th
i
s
impro
v
e
m
e
n
t d
e
pe
n
d
s o
n
t
he ac
c
ura
c
y of
e
d
g
e
i
n
for
m
a
t
i
o
n
prov
i
d
e
d. Th
i
s
con
d
i
tio
n he
l
p
s the o
p
tim
i
z
er to
o
l
s
ma
ke
a
d
e
c
i
s
i
on
d
u
ri
n
g
c
os
t
c
o
m
p
u
t
a
tio
n.
T
he
a
c
c
ura
t
e
ad
di
tio
na
l
inf
o
rma
t
ion
a
voi
ds
t
he
e
st
i
m
a
t
ion
too
l
f
o
r
pred
ic
ti
ng
the
sa
me
s
ol
u
t
i
on
freque
n
t
l
y
.
A
s
a
r
e
s
ult,
t
hi
s
e
dge
e
n
h
a
n
c
e
m
ent
t
r
ansfor
m
s
t
he
I
BP
t
e
c
hn
i
que
mor
e
e
ffi
c
i
e
n
t
i
n pre
d
i
c
tin
g a
nd l
i
gh
t
com
p
u
t
a
t
i
ona
l.
G
r
a
t
e
f
u
lly,
t
h
e
pro
pose
d
te
c
h
n
i
que
c
a
p
ab
le
t
o bo
os
t
up t
h
e G
r
ad
ie
n
t
Desce
n
t
per
form
ance
be
tter
t
h
a
n
prev
io
us.
I
n
a
dd
i
tio
nal
t
h
is
t
e
c
h
n
i
q
u
e
c
a
n
s
pee
d
up
the
cos
t
c
o
mp
ut
a
t
i
o
n
a
l
unt
il
7
0
%
f
a
s
t
e
r
f
r
om
t
h
e
con
v
e
n
t
i
ona
l
G
r
adien
t
D
esc
e
nt
m
e
t
ho
d.
T
hi
s
situa
t
io
n
pr
ove
d
tha
t
t
h
e
edge
e
nha
nce
m
e
n
t
tec
h
n
i
q
u
e
a
ssi
s
t
s
t
o
op
tim
ize the
e
s
t
i
m
a
t
i
o
n
pr
oc
e
ss
d
u
ri
ng
s
e
e
k
i
n
g
t
h
e fi
na
l
H
R
im
a
ge
. The
a
c
c
u
r
a
te es
t
i
m
a
tio
n fu
nc
t
i
o
n
de
p
en
ds
on
t
he
p
re
cis
i
o
n
o
f
ed
ge
d
e
t
e
c
tor
p
r
o
v
ide
s
t
he
h
i
gh
l
o
c
a
l
i
z
a
ti
o
n
an
d
high
ra
t
e
e
d
g
e
d
et
ec
t
i
on
.
Thi
s
a
ccu
rac
y
edge
detec
t
i
o
n
hel
p
s t
h
e
estim
ator to pred
ic
t furt
her
s
t
e
p
to
g
et
tin
g
cl
ose
t
h
e o
p
t
i
m
al
resu
l
t.
T
ab
le
2
s
ho
ws
t
he
numbe
r of Ite
rati
o
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
ones
i
a
n
J
E
lec
En
g & Co
mp
S
c
i
IS
S
N
: 2502-
47
52
Ed
ge
e
n
h
a
n
ce
m
e
nt
o
f
IBP re
con
s
tr
uc
tio
n b
y
u
s
i
ng
sh
ar
p
i
n
f
i
ni
te sym
m
e
tric
al
e
x
p
o
n
en
t
i
a
l… (A.R.
A
Nazren)
26
5
Tab
l
e
2.
N
umbe
r of Iter
a
tions
T
he
I
B
P
t
e
c
hniqu
e
wit
h
IB
P
P
a
tel
C
h
e
r
e
f
C
A
NNY
S
I
S
E
F
LE
NA
8
9
4
6
4
5
7
2
3
0
C
A
M
E
R
A
MAN
82
4
6
44
6
7
27
M
A
N
D
R
I
L
117
6
5
58
9
9
35
PE
PP
ER
116
6
2
57
9
6
31
C
L
O
W
N
114
6
3
57
9
4
39
J
E
T
P
L
A
NE
9
2
4
7
4
8
7
6
3
5
BR
A
I
N
91
5
1
44
7
3
22
S
H
I
P
7
8
4
1
4
1
6
3
2
5
U
S
AF
4
4
2
7
3
0
3
7
2
5
B
R
I
D
G
E
9
8
5
6
5
0
8
1
3
4
A
n
o
t
her
ke
y
poi
n
t
f
or
s
pe
e
d
up
t
h
e
e
s
t
i
m
a
tio
n
de
pen
d
s
on
the
ac
c
u
r
a
t
e
r
a
t
e
o
f
s
t
e
p
s
i
z
e
f
o
r
t
h
e
g
r
a
d
i
e
n
t
.
In
r
ea
l
i
t
y
,
th
e
g
r
a
d
i
e
n
t
v
alu
e
d
et
ect
ed
v
ari
e
s
l
a
rg
e
in
t
e
n
s
ity
a
mo
u
n
t
,
t
hu
s
the
e
s
ti
ma
t
i
on
n
ev
e
r
b
e
con
v
er
g
i
ng
a
ft
e
r
r
ece
ivin
g th
e
lar
g
e
am
o
unt
o
f d
a
ta.
T
h
is
s
itua
t
i
o
n
c
a
n
be
ha
nd
le
d by
se
t
s
t
he su
ita
b
l
e
r
a
te for
s
t
e
p
s
i
z
e
.
T
h
e
s
e
l
e
c
t
i
o
n
o
f
s
t
e
p
s
i
z
e
i
s
i
m
p
o
r
t
a
n
t
t
o
e
n
s
u
r
e
e
s
t
i
m
a
t
o
r
r
e
c
e
ive
s
a
n
o
p
tim
um
i
nf
orm
a
tio
n
w
i
t
h
less
c
o
m
p
ut
a
t
i
ona
l
c
o
st.
In
t
his
e
xperim
e
nt
a
l
l
t
ec
hn
i
que
e
xam
i
ne
s
prov
id
e
d
s
am
e
o
p
t
i
m
um
s
tep
siz
e
e
x
pect
the
t
e
c
h
n
i
q
u
e
pro
pose
d
by
P
a
tel.
T
hus,
t
h
is
e
stim
ati
on
e
f
fi
cien
c
y
t
o
t
a
l
ly
d
e
p
e
n
ds
on
t
h
e
per
f
orm
a
nce
of
e
d
g
e
d
e
t
e
c
t
o
r
p
e
r
f
o
r
m
a
n
c
e
.
T
h
e
r
e
s
u
l
t
i
n
T
a
b
l
e
4
.
4
s
h
o
w
n
,
t
h
e
S
h
a
r
p
IS
EF
e
dge
d
e
t
ec
tor
is
t
he
b
e
s
t
reg
u
la
t
o
r
dur
in
g
perfo
r
m
ing
t
h
e
es
t
i
m
a
t
io
n
proc
es
s
sinc
e
t
h
i
s
t
e
c
h
n
i
que
p
r
o
v
i
de
s
l
e
sse
r
iter
a
t
i
o
n
c
om
pare
t
he
o
t
h
er
regu
la
tor
tec
h
n
i
q
u
e.
A
dd
iti
o
n
al
ly,
the
pro
p
o
se
d
te
c
h
n
i
q
u
e
an
d
t
he
P
ate
l
t
ec
hn
i
que
u
se
s
am
e
edge
d
e
t
ec
tor
.
H
o
w
e
ve
r,
t
he
s
u
i
ta
b
l
e
ste
p
s
iz
e
ra
te
g
i
v
es
a
n
a
d
van
t
a
g
e
of
p
ro
p
o
s
e
d
t
ec
hn
iq
ues
t
o
b
e
c
ome
lig
ht
c
ost
com
p
le
x
i
t
y
u
p to 4
0%
f
as
ter
.
4.
CONCL
U
S
ION
Th
is
s
tu
dy
prop
ose
d
a
p
rec
i
se
e
d
g
e
e
n
ha
nce
m
e
n
t
r
e
g
u
l
ator
t
o
i
m
p
r
o
v
e
t
h
e
I
B
P
r
e
c
o
n
s
t
r
u
c
t
i
o
n
tech
n
i
q
u
e
pr
o
b
l
e
m
b
y
us
i
n
g
t
h
e
S
h
a
r
p
IS
EF
r
egu
l
ar
i
z
a
t
ion.
N
or
ma
l
l
y,
t
he
I
BP
r
ec
onstru
c
tio
n
te
c
h
n
i
qu
e
ha
s
suffe
r
ed
f
rom
pro
duce
d
t
he
r
in
gi
n
g
a
rtefa
c
ts
on
H
R
e
s
tim
ati
o
n
a
nd
i
t
a
l
so
r
e
q
ui
r
e
d
a
l
a
rge
num
ber
o
f
i
t
erat
ion
s
t
o
acco
mp
li
sh
t
h
e
e
st
i
m
a
t
ion
.
M
o
r
e
o
v
e
r,
t
h
i
s
t
e
ch
niq
ue
l
ea
k
of
c
a
p
ab
i
lit
y
to
p
rese
rve
t
h
e
lo
st
h
ig
h
fre
que
nc
y
in
fo
rm
atio
n
due
t
o
bl
urry
e
ffe
c
t
.
H
o
w
e
ver,
t
he
S
ha
rp
I
S
E
F
h
a
s
a
s
o
l
u
t
i
o
n
t
o
c
u
r
e
t
h
e
I
B
P
rec
ons
t
r
uc
tio
n
pro
b
le
m
w
i
t
h
p
ro
vid
i
n
g
p
re
c
i
se
e
d
g
e
e
nha
nce
m
ent
w
i
t
h
l
i
ght
c
o
s
t
com
p
u
t
a
tio
n.
T
he
S
harp
IS
EF
m
etho
d
ha
s
hig
h
l
y
pre
c
is
i
o
n
l
o
ca
l
i
z
a
t
ion
a
n
d
less
e
r
ror
r
a
t
e
d
e
t
e
c
t
i
o
n
.
T
h
u
s
t
h
e
s
e
a
d
v
a
nt
ag
es
h
el
p
th
e
est
i
ma
t
o
r
t
o
c
om
pute
the
cor
r
ec
tio
n
process
more
a
ccur
a
te and f
as
ter.
ACKNOW
LEDG
E
MEN
T
S
The
au
t
hor
w
o
u
ld
l
i
k
e
t
o
a
c
k
no
w
l
e
d
ge
t
he
s
upp
or
t
from
t
he
F
un
da
me
nta
l
R
e
s
ea
rch
Gran
t
Sch
e
me
(F
RG
S
)
under
a
gra
n
t
n
u
m
b
er
o
f
FRGS
/
1/2
017
/
I
CT05/
UNIM
A
P-/02/
2
from
the
M
i
n
i
str
y
of
High
er
E
d
uc
a
tio
n M
a
la
ysi
a
.
The
aut
h
or
w
ou
ld
like
to
e
x
p
res
s
g
rati
t
ude
t
o
t
h
e
M
i
n
i
str
y
o
f
H
i
gher
E
duca
t
i
o
n
Ma
lays
ia a
n
d
U
ni
ver
s
i
t
y
Mal
a
ysia
P
e
r
lis
f
or
f
ac
ili
tie
s prov
i
d
ed
i
n
t
h
is
w
ork.
REFE
RENCES
[1]
M
.
-C.
Chi
a
ng
a
nd
T
.
E
.
Bo
u
lt,
“
Efficien
t
imag
e
warpi
n
g
and
s
u
p
e
r-resol
u
t
i
on
,”
A
pplica
t
i
o
n
s
of Com
put
er Vision,
19
96
.
WA
CV ’9
6
.
,
Pr
oceedi
n
g
s
3rd
IEEE
W
o
r
k
shop
on
.
pp.
5
6–
61
,
19
96
.
[2]
B.
C
o
h
en
a
n
d
I
.
D
i
ns
tei
n
,
“P
ol
y
p
has
e
b
ack
-pro
jecti
o
n
fi
lt
eri
ng
fo
r
im
age
res
o
l
u
tion
en
han
cemen
t,”
IE
E
P
r
o
c
.
-
Visio
n
,
I
m
ag
e
,
S
i
gn
al
Proc
e
s
s.
,
vol
.
14
7,
n
o
.
4
,
p
.
31
8,
200
0.
[3]
Hy
un
woo
Kim
,
J
eo
ng-H
u
n
Jan
g
,
an
d
Ki
-S
ang
H
o
n
g
,
“
E
d
g
e
-
en
han
c
ing
super-resoluti
on
usi
n
g
anis
o
t
ropic
diff
usion
,
”
in
Proceedi
n
g
s
20
01
In
ter
n
a
t
i
onal Confer
ence on
Ima
g
e Pr
ocess
i
n
g
(Cat
. No
.0
1CH
3
7
2
05
)
,
20
01,
v
o
l
.
2
,
p
p.
1
30
–1
33
.
[4]
Z
.
Y
a
n
,
Y
.
L
u
,
a
n
d
H
.
Y
a
n
,
“
R
e
d
u
c
i
n
g
T
h
e
S
p
i
r
a
l
C
T
S
l
i
c
e
T
h
i
c
k
n
ess
Usin
g
S
up
er
R
esol
ut
io
n
,
”
20
10
IEEE
In
te
rn
at
io
na
l Co
nfe
r
e
n
c
e
o
n
Ima
g
e
Proc
e
ssin
g
.
pp.
5
9
3–5
96
,
2
010.
[5]
F
.
Q
in,
“
A
n
improv
ed
s
u
p
er
r
eso
l
u
t
io
n
recon
s
tru
c
ti
on
m
et
hod
b
as
ed
o
n
ini
tial
value
estima
t
ion,”
20
10 3
r
d In
t
.
Co
ng
r. I
m
ag
e
Si
gn
al
Pro
c
e
s
s
.
,
p
p
.
8
26
–82
9,
2
0
1
0
.
[6]
X
.
L
i
a
n
g
a
n
d
Z
.
G
a
n
,
“
I
m
p
r
o
v
e
d
N
o
n
-
l
o
c
a
l
I
t
e
r
a
t
i
v
e
B
a
c
k
-
P
r
o
j
e
c
t
io
n
M
e
th
od
f
o
r
I
m
a
ge
S
u
p
er-
R
eso
l
utio
n,
”
20
11
Si
xth
In
t.
Conf.
Ima
g
e G
r
ap
h
.
,
pp.
1
76
–1
81,
2
0
1
1
.
[7]
T.
B
e
n
gt
sson
,
I.
Y
.
H.
G
u,
M
.
Vi
be
rg
,
a
n
d
K.
L
in
dströ
m
,
“Re
g
u
l
a
rized
o
pt
i
m
izati
o
n
f
o
r
jo
int
supe
r-res
o
lut
i
on
a
nd
hi
gh
d
ynam
i
c
ra
n
g
e
im
age
recon
s
tru
c
ti
on
in
a
p
ercep
tu
a
l
ly
u
nif
o
rm
d
om
ai
n
,
”
2012 IEEE
Int
e
rnat
ional
Confer
ence
on
Aco
u
stics, S
p
eech
an
d S
i
g
nal
P
r
oces
s
i
ng
(
I
CAS
S
P)
.
p
p
.
1
097
–
1
10
0,
2
01
2.
[8]
M
.
N
. Bareja, “An
E
ff
ecti
v
e It
erat
i
v
e Back
P
roject
io
n b
a
sed Sin
g
l
e
Imag
e S
u
p
e
r Reso
lu
tio
n
App
roach
,” 20
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2502-
4752
I
n
do
n
e
si
an
J
E
l
e
c
E
n
g
&
C
o
m
p
S
ci
, V
o
l
.
1
4
,
No. 1, April 2019 :
258 –
2
66
26
6
[9]
R.
N
ay
ak,
S
.
M
o
n
al
is
a,
a
nd
D
.
P
a
t
r
a,
“
S
p
atial
sup
e
r
resol
u
t
i
o
n
b
a
sed
i
m
age
rec
o
n
s
t
r
ucti
on
u
sing
HIBP
,
”
i
n
2
013
An
nu
al
IEEE
India
Confer
ence
(
I
NDICO
N
)
,
20
1
3
,
p
p.
1–6
.
[10]
P
.
R
asti,
H.
D
em
ire
l
,
and
G.
A
nb
arjaf
a
ri,
“
I
m
a
ge
r
eso
l
u
t
i
on
enh
anc
e
m
e
nt
b
y
us
ing
interp
olatio
n
fo
ll
ow
ed
b
y
it
erativ
e
b
ack
p
roj
ecti
o
n
,
”
20
1
3
21
st Sign
al Pr
oc
e
s
sing
an
d Co
mmun
ic
a
t
io
n
s
App
l
ic
a
tion
s
Co
nfe
r
e
n
c
e
(
S
IU
)
.
p
p
. 1–
4
, 2
01
3.
[11]
Y. Cheref
and
D. Yo
u
sfi, “
Imag
e
recon
s
t
r
uct
i
on
w
ith
b
e
tter
e
dge
e
n
h
a
nce
m
e
n
t u
s
in
g
su
p
e
r re
so
lu
ti
on
a
lg
ori
t
hm
,”
i
n
2014 Int
e
rnational
Con
f
ere
n
ce on
M
u
ltimed
i
a
C
o
m
p
ut
ing and Syst
ems (ICMCS
)
,
2
0
1
4
,
pp
.
2
83–
28
8.
[12]
J.
M
e
i
,
H.
G
ao,
B
.
M
aiseli,
a
nd
O
.
Elisha,
“Edg
e
preservat
i
on
i
m
a
ge
e
nl
argem
e
nt
a
n
d
e
nh
ance
m
e
nt
m
et
ho
d
ba
s
e
d
on
t
h
e
a
daptiv
e
P
e
ron
a
–M
ali
k
n
on-l
i
n
ear
d
iff
u
si
on
m
od
el,
”
IE
T Imag
e Pr
ocess.
,
v
o
l
.
8
,
no
.
12
,
pp
.
7
53–
76
0,
2014.
[13]
X.
Y
an
g,
Y
.
Zh
an
g,
D
.
Zh
ou,
a
nd
R
.
Yan
g
,
“An
im
pro
v
ed
i
t
e
rati
ve
b
a
c
k
p
ro
jec
t
io
n
a
l
go
rith
m
b
a
se
d
on
r
in
ging
artif
acts
sup
p
ressi
on,”
Neur
ocom
pu
ti
ng
,
vol.
1
6
2
,
pp.
1
7
1–1
79,
2
0
1
5.
[14]
M
.
N
awaz,
R
.
Xie,
L
.
Zh
ang,
M
.
A
s
f
a
n
dyar,
a
n
d
M
.
Hu
s
s
ain
,
“
Imag
e
sup
e
r
resol
u
t
i
o
n
b
y
spars
e
l
inear
r
egress
io
n
and
i
t
erat
ive
ba
ck
p
roj
ecti
on,
”
2
0
1
6
IEEE In
t
e
rna
t
i
o
n
a
l S
y
mp
o
s
i
u
m
o
n
Br
oadba
nd
M
u
l
t
i
m
e
d
i
a
Sys
t
ems and
Br
oa
dcasti
ng (
B
M
S
B)
. p
p.
1–
6
, 2
01
6.
[15]
A
.
R
.
A
.
N
a
z
r
e
n
,
S
.
N
.
Y
a
a
k
o
b
,
R
.
N
g
a
d
i
r
a
n
,
M
.
B
.
H
i
s
h
a
m
,
a
n
d
N
.
M
.
W
a
f
i
,
“
I
m
provi
ng
iterative
back
p
roject
ion
su
per
res
o
luti
on
m
o
d
e
l
v
i
a
anis
otro
p
i
c
diff
u
s
ion
edg
e
e
nh
an
cem
e
nt
,
”
i
n
2
0
1
6
In
tern
a
t
io
nal Confer
ence o
n
Ro
bo
tics
,
A
u
t
o
mati
on an
d S
c
ien
ces
(
I
CO
RAS)
,
20
16
, pp
. 1
–4
.
[16]
R.
R
.
Mak
w
an
a
and
N.
D
.
M
e
hta,
Em
erging
Trends
i
n
E
l
ectric
a
l,
El
ectro
n
i
c an
d
Comm
unica
tio
ns
En
g
i
neer
ing
,
vo
l.
416
,
n
o
.
1.
C
h
a
m
:
S
p
r
in
ger In
ternat
io
nal
P
u
b
lishi
ng,
2
017.
[17]
R
.
N
a
y
a
k
a
n
d
D
.
P
a
t
r
a
,
“
A
n
e
d
g
e
p
r
e
s
e
r
v
i
n
g
I
B
P
b
a
s
e
d
s
u
p
e
r
r
e
s
o
luti
on
im
age
reco
nst
r
uct
i
o
n
u
sin
g
P
-s
p
l
ine
an
d
MuCSO-QPSO al
g
orithm,”
M
i
cr
o
s
ys
t. Tech
no
l.
,
v
o
l.
2
3,
n
o.
3,
pp
. 55
3
–
5
6
9
, Mar.
2
0
1
7
.
[18]
M
.
Irani
,
M
.
Ira
n
i
,
S.
P
e
l
eg
,
and
S
.
P
e
l
eg
,
“M
otio
n
A
n
al
ysis
f
o
r
Im
age
E
n
h
a
nce
m
ent:Reso
l
uti
on,
O
ccl
usio
n,
a
n
d
Tran
sparen
cy,”
J
.
Vi
s
.
Co
mm
un
. Ima
g
e R
e
pr
esent.
,
vol.
4
,
p
p.
3
2
4
–3
35
,
1
993
.
[19]
S
.
R
at
ho
re
a
n
d
Y
.
Hars
halat
h
a
,
“
Iterativ
e
Bac
k
-P
ro
ject
ion
Alg
o
r
i
t
hm
B
ased
S
i
g
n
a
l
P
r
o
cessin
g
A
p
p
roach
T
o
Enh
a
nc
e
An
I
m
a
ge
Re
s
olut
io
n
,
”
I
n
t
.
J. En
g.
R
e
s
.
Ap
pl
.
,
vo
l.
2
,
no.
6
,
pp.
1
449
–1
454,
2
01
2.
[20]
S
.
C
as
tan,
J
.
Z
h
ao
,
an
d
J.
S
hen
,
“
N
e
w
ed
ge
d
et
ect
io
n
met
h
o
d
s
b
a
s
e
d
o
n
e
xp
one
n
t
i
a
l
fil
t
e
r
,
”
i
n
[19
9
0
]
Pro
c
eedi
ngs
.
10
th
In
t
e
rna
t
i
onal
Conferen
ce on
Patt
ern
Reco
gn
iti
o
n
,
1
99
0,
v
o
l
.
i
,
n
o.
X
,
p
p
.
709–7
11
.
[21]
P
.
S
.
A.
P
a
t
el
S
hrey
as
A
.,
“
N
ovel
Iterat
ive
Back
P
ro
jecti
o
n
App
roach
,”
IOSR
J.
Comp
ut.
En
g.
,
vol
.
11
,
no.
1
,
pp
.
65
–6
9,
2
01
3
B
I
OGRAPHIES
O
F AUTHO
RS
F
i
rs
t
Auth
or
r
eceiv
e
d
the
B.
E
n
g
i
n
eerin
g
(Hon
s)
i
n
Co
m
puter
E
ng
ineeri
n
g
f
r
om
U
ni
versity
M
a
l
a
ysi
a
P
erlis
(
Un
iM
AP),
M
alay
si
a
and
P
h
.D.
deg
r
ee
i
n
t
h
e
C
om
p
u
t
er
E
ngi
neerin
g
f
r
o
m
Un
iv
e
r
s
i
t
y
M
a
l
a
y
s
i
a
P
e
r
l
i
s
,
M
a
la
y
s
ia
.
S
eco
nd
Au
th
or
r
ec
ei
ved
t
h
e
B.
E
n
g
i
n
eerin
g
(Ho
n
s
)
i
n
Co
m
m
u
n
i
cati
on
a
nd
C
o
m
put
e
r
E
n
g
i
neeri
n
g
f
r
o
m
U
n
i
vers
it
y
Keban
g
saan
M
ala
y
si
a
(UKM),
M
alay
si
a
.
and
t
h
e
M
.
S.
d
eg
ree
in
Co
mp
ut
er
f
rom
Un
iv
ersity
K
eb
ang
s
aan
M
ala
y
s
i
a
(UKM
),
i
n
2
0
0
6
,
an
d
P
h
.D.
d
e
gree
in
Co
mm
u
n
i
cati
o
n
s
, S
e
n
s
o
r
s an
d Sign
al P
roces
s
i
ng
f
ro
m
N
e
wcastle U
n
i
v
ersity,
UK.
T
h
i
r
d
Aut
h
o
r
r
eceiv
e
d
th
e
B.E
n
g
.
i
n
El
ectri
c
a
l
-
El
ectronics
(
H
o
n
s)
f
rom
Uni
v
er
sity
T
eknologi
M
a
l
a
ysi
a
(
UTM)
a
nd
M
.S
.
d
e
g
r
ees
b
y
re
s
e
arc
h
i
n
Co
m
p
u
t
er
E
ng
in
ee
ri
ng
f
ro
m
U
n
i
v
er
s
i
t
y
M
a
l
a
ysi
a
P
erlis
(
Un
iM
AP
)
th
e
P
h
.
D
.
degree
in
t
he
C
o
m
puter
S
y
s
t
e
m
E
n
gi
n
e
e
r
i
n
g
f
r
om
U
n
i
v
ersi
ty
S
outh
Au
st
ralia,
Au
s
t
r
alia.
Evaluation Warning : The document was created with Spire.PDF for Python.