I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l
E
ng
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
,
p
p
.
1
5
21
~
1
5
31
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
6
.
i
3
.
pp
1
5
21
-
1
5
31
1521
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
No
-
refe
rence
ima
g
e quality
as
ses
sm
ent
ba
sed
on v
isua
l
ex
pla
na
tion ima
g
es a
nd deep
t
ra
ns
fer
lea
rni
ng
B
a
s
m
a
Ahm
ed
1
,
O
s
a
m
a
A.
O
m
er
2
,
Viv
e
k
K
um
a
r
Sin
g
h
3
,
Am
a
l R
a
s
hed
1
,
M
o
ha
m
ed
Abdel
-
Na
s
s
er
2
1
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
e
r
s a
n
d
I
n
f
o
r
ma
t
i
o
n
,
S
o
u
t
h
V
a
l
l
e
y
U
n
i
v
e
r
si
t
y
,
Q
e
n
a
,
E
g
y
p
t
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
,
A
sw
a
n
U
n
i
v
e
r
si
t
y
,
A
sw
a
n
,
E
g
y
p
t
3
B
a
r
t
s
C
a
n
c
e
r
I
n
st
i
t
u
t
e
,
Q
u
e
e
n
M
a
r
y
U
n
i
v
e
r
si
t
y
o
f
L
o
n
d
o
n
,
L
o
n
d
o
n
,
U
n
i
t
e
d
K
i
n
g
d
o
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
19
,
2
0
2
4
R
ev
is
ed
Au
g
23
,
202
4
Acc
ep
ted
A
ug
31
,
2
0
2
4
Qu
a
n
ti
fy
in
g
ima
g
e
q
u
a
li
t
y
i
n
t
h
e
a
b
se
n
c
e
o
f
a
re
fe
re
n
c
e
ima
g
e
c
o
n
ti
n
u
e
s
to
b
e
a
c
h
a
ll
e
n
g
e
d
e
sp
it
e
t
h
e
in
tr
o
d
u
c
ti
o
n
o
f
n
u
m
e
ro
u
s
n
o
-
re
fe
re
n
c
e
ima
g
e
q
u
a
li
t
y
a
ss
e
ss
m
e
n
ts
(NR
-
IQA
)
i
n
re
c
e
n
t
y
e
a
rs.
Un
li
k
e
m
o
st
e
x
i
stin
g
NR
-
IQA
m
e
th
o
d
s,
t
h
is
p
a
p
e
r
p
ro
p
o
s
e
s
a
n
e
fficie
n
t
NR
-
IQA
m
e
th
o
d
b
a
se
d
o
n
d
e
e
p
v
isu
a
l
i
n
terp
re
tati
o
n
s.
S
p
e
c
ifi
c
a
ll
y
,
th
e
m
a
in
c
o
m
p
o
n
e
n
ts
o
f
th
e
p
ro
p
o
se
d
m
e
th
o
d
a
re
:
i
)
g
e
n
e
ra
ti
n
g
a
p
se
u
d
o
-
re
fe
re
n
c
e
ima
g
e
(P
RI)
fo
r
t
h
e
in
p
u
t
d
ist
o
rted
ima
g
e
s,
ii
)
e
m
p
lo
y
in
g
a
p
re
train
e
d
c
o
n
v
o
lu
ti
o
n
a
l
n
e
two
rk
to
e
x
trac
t
fe
a
tu
re
m
a
p
s fro
m
th
e
d
isto
rted
ima
g
e
a
n
d
t
h
e
c
o
rre
sp
o
n
d
i
n
g
P
RI,
iii
)
p
ro
d
u
c
in
g
v
isu
a
l
e
x
p
lan
a
ti
o
n
ima
g
e
s
(VEIs)
b
y
u
si
n
g
th
e
fe
a
tu
re
m
a
p
s
o
f
t
h
e
d
isto
rte
d
ima
g
e
a
n
d
th
e
c
o
r
re
sp
o
n
d
in
g
P
RI,
iv
)
m
e
a
su
rin
g
th
e
sim
il
a
rit
y
b
e
twe
e
n
th
e
two
VEIs
u
sin
g
a
n
i
m
a
g
e
sim
il
a
rit
y
m
e
tri
c
,
a
n
d
v
)
e
m
p
lo
y
i
n
g
a
non
-
li
n
e
a
r
m
a
p
p
in
g
f
u
n
c
ti
o
n
f
o
r
q
u
a
li
ty
sc
o
re
a
li
g
n
m
e
n
t.
In
o
u
r
e
x
p
e
rime
n
ts,
we
e
v
a
lu
a
ted
th
e
e
ffica
c
y
o
f
th
e
p
ro
p
o
se
d
m
e
th
o
d
a
c
ro
ss
v
a
rio
u
s
fo
rm
s
o
f
d
isto
rti
o
n
u
sin
g
fo
u
r
b
e
n
c
h
m
a
rk
d
a
tas
e
ts
(LI
VE,
S
IQA
D,
CS
IQ,
a
n
d
TID2
0
1
3
).
T
h
e
p
ro
p
o
se
d
a
p
p
r
o
a
c
h
d
e
m
o
n
stra
tes
p
a
rit
y
wit
h
t
h
e
late
st
m
e
th
o
d
s,
a
s
e
v
i
d
e
n
c
e
d
b
y
c
o
m
p
a
riso
n
s
wit
h
b
o
th
h
a
n
d
-
c
ra
fted
NR
-
IQA
a
n
d
d
e
e
p
lea
rn
in
g
-
b
a
se
d
a
p
p
r
o
a
c
h
e
s.
K
ey
w
o
r
d
s
:
Dee
p
l
ea
r
n
in
g
Gr
ad
-
C
AM
I
m
ag
e
q
u
ality
ass
ess
m
en
t
Ps
eu
d
o
-
r
ef
er
e
n
ce
Similar
ity
m
ea
s
u
r
es
Vis
u
al
i
n
ter
p
r
etatio
n
s
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Mo
h
am
ed
Ab
d
el
-
Nass
er
Dep
ar
tm
en
t
o
f
E
lectr
ical
E
n
g
i
n
ee
r
in
g
,
Facu
lty
o
f
E
n
g
in
ee
r
in
g
,
Aswan
Un
iv
er
s
ity
Aswan
,
E
g
y
p
t
E
m
ail: m
o
h
am
ed
.
ab
d
eln
ass
er
@
aswu
.
ed
u
.
eg
1.
I
NT
RO
D
UCT
I
O
N
I
n
d
ee
d
,
2
-
d
im
e
n
s
io
n
al
(
2
D)
m
u
ltime
d
ia
tech
n
o
lo
g
ies
with
th
eir
ap
p
licatio
n
s
,
n
o
tab
ly
co
m
m
u
n
icatio
n
s
,
ed
u
ca
tio
n
,
an
d
in
d
u
s
tr
y
,
ar
e
a
k
ey
ele
m
en
t
f
o
r
h
u
m
an
v
is
u
aliza
tio
n
.
Ho
wev
e
r
,
d
i
g
ital
im
ag
es
s
h
o
w
lo
w
p
e
r
ce
iv
ed
v
is
u
al
q
u
ality
f
o
r
s
o
m
e
ca
p
tu
r
in
g
s
ettin
g
s
,
n
o
tab
ly
n
o
is
e
s
en
s
itiv
ity
,
ap
er
tu
r
e
,
lig
h
tin
g
,
ex
p
o
s
u
r
e,
a
n
d
len
s
lim
itatio
n
s
.
T
h
ese
s
ettin
g
s
ca
n
cr
ea
te
d
is
tr
ac
tin
g
im
ag
e
ar
tifa
cts
th
at
n
eg
ativ
ely
im
p
ac
t
th
e
r
elate
d
h
u
m
an
v
i
s
u
al
q
u
ality
.
T
h
e
im
p
lem
e
n
ta
tio
n
o
f
an
a
u
to
m
atic
im
ag
e
q
u
ality
ass
ess
m
en
t
(
I
QA)
alg
o
r
ith
m
to
f
o
r
ec
ast
th
e
d
ig
ital
im
ag
es’
q
u
ality
ca
n
f
ac
ilit
ate
th
e
m
an
ag
em
e
n
t
o
f
s
u
b
s
tan
d
ar
d
im
ag
es
o
r
th
e
co
r
r
ec
tio
n
o
f
th
eir
q
u
a
lity
wh
ile
th
ey
ar
e
b
ein
g
ca
p
tu
r
ed
.
R
ec
en
tly
,
n
u
m
e
r
o
u
s
I
Q
A
alg
o
r
ith
m
s
h
av
e
b
ee
n
s
u
g
g
ested
.
Su
b
jectiv
e
an
d
o
b
jectiv
e
I
QA
alg
o
r
ith
m
s
ca
n
b
e
d
is
tin
g
u
is
h
ed
b
y
th
e
d
eg
r
ee
to
wh
ich
h
u
m
an
ey
es
ev
alu
ate
th
e
r
an
k
in
g
s
.
B
ased
o
n
th
e
ex
is
ten
ce
o
r
lack
th
er
eo
f
o
f
a
r
ef
er
e
n
ce
im
ag
e,
o
b
jectiv
e
ass
es
s
m
en
t
ca
n
b
e
ess
en
tially
ca
teg
o
r
ize
d
in
to
t
h
r
ee
g
r
o
u
p
s
:
i
)
f
u
ll
r
ef
er
en
ce
I
QA
(
FR
-
I
QA)
,
ii
)
n
o
r
ef
er
en
ce
o
r
r
ef
er
en
ce
-
f
r
ee
I
QA
(
NR
-
I
QA)
,
an
d
iii
)
r
e
d
u
ce
d
r
ef
er
e
n
ce
I
Q
A
(
R
R
-
I
QA)
[
1
]
–
[
4
]
.
I
n
d
ee
d
,
r
e
g
ar
d
less
o
f
th
e
ab
s
en
ce
o
r
p
r
esen
ce
o
f
a
r
e
f
er
en
ce
im
ag
e
d
u
r
in
g
th
e
q
u
ality
a
s
s
es
s
m
en
t
p
r
o
ce
d
u
r
e,
th
e
f
u
n
d
a
m
en
tal
co
n
ce
r
n
o
f
I
QA
is
to
q
u
an
tify
th
e
“d
is
tan
ce
”
b
etwe
en
th
e
r
ef
er
en
ce
an
d
d
is
to
r
ted
im
ag
es.
I
t
is
wo
r
th
n
o
tin
g
th
at
th
e
ter
m
“d
is
tan
ce
”
r
e
f
er
s
to
t
h
e
in
d
e
x
th
at
is
b
ein
g
ass
ess
e
d
(
e.
g
.
m
ea
n
s
q
u
a
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
:
1
5
2
1
-
1
5
3
1
1522
er
r
o
r
an
d
s
ig
n
al
-
to
-
n
o
is
e
r
atio
)
.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
NR
-
I
QA
alg
o
r
ith
m
s
o
p
er
ate
u
n
d
er
t
h
e
ass
u
m
p
tio
n
th
at
ass
es
s
in
g
im
ag
e
q
u
ality
d
o
es
n
o
t
r
e
q
u
ir
e
a
d
ir
ec
t
c
o
m
p
ar
is
o
n
b
etwe
en
th
e
o
r
i
g
in
al
a
n
d
te
s
t
im
ag
es.
T
h
u
s
,
i
n
s
itu
atio
n
s
wh
er
e
th
e
o
r
ig
in
al
im
ag
e
is
u
n
av
ailab
le,
th
e
y
m
ay
b
e
u
tili
ze
d
[
1
]
.
Desp
ite
th
e
ex
is
ten
ce
o
f
n
u
m
er
o
u
s
NR
-
I
QA
ap
p
r
o
ac
h
e
s
,
o
n
ly
a
lim
ited
n
u
m
b
er
o
f
s
tu
d
ies h
av
e
f
o
c
u
s
ed
o
n
a
n
ad
d
itio
n
al
s
ev
er
e
q
u
ality
ev
alu
atio
n
,
n
am
el
y
th
e
d
eg
r
e
e
o
f
d
eg
r
ad
atio
n
in
an
im
ag
e'
s
q
u
ality
.
An
NR
-
I
QA
m
eth
o
d
was
p
r
o
p
o
s
ed
b
y
L
i
et
a
l.
[
5
]
t
o
h
an
d
le
f
u
zz
y
im
ag
es.
T
h
is
m
eth
o
d
u
s
ed
th
e
o
r
i
g
in
al
b
lu
r
r
e
d
im
ag
e
as a
p
s
eu
d
o
-
r
ef
er
e
n
ce
im
ag
e
(
PR
I
)
an
d
im
p
o
s
ed
ad
d
itio
n
al
b
lu
r
d
eg
r
a
d
atio
n
o
n
it.
T
h
e
f
in
al
ev
alu
atio
n
s
co
r
e
was
ca
lcu
lated
as
a
f
u
zz
y
in
d
ex
b
etwe
e
n
th
e
two
im
ag
e
s
.
On
o
f
t
h
e
m
ain
lim
itatio
n
s
o
f
th
is
ap
p
r
o
ac
h
is
th
at
it
s
o
le
ly
ac
co
u
n
ted
f
o
r
a
s
o
litar
y
h
az
y
d
is
to
r
tio
n
.
A
d
d
it
io
n
ally
,
Min
et
a
l.
[
6
]
p
r
o
p
o
s
ed
th
r
ee
d
is
to
r
tio
n
-
s
p
ec
if
ic
m
etr
ics
b
ased
o
n
PR
I
to
esti
m
ate
b
lo
ck
in
ess
,
s
h
ar
p
n
ess
,
an
d
n
o
is
in
ess
,
r
esp
ec
tiv
ely
.
I
n
o
r
d
e
r
to
g
en
er
ate
th
e
PR
I
,
a
d
eg
r
a
d
atio
n
was
ap
p
lied
to
th
e
d
is
to
r
ted
im
a
g
e
f
o
r
ev
er
y
f
o
r
m
o
f
d
is
to
r
tio
n
.
Nex
t,
th
e
PR
I
was
u
s
ed
as
a
n
ew
b
en
c
h
m
ar
k
to
d
eter
m
in
e
t
h
e
“
d
is
tan
ce
”
b
et
wee
n
th
e
PR
I
an
d
th
e
d
is
to
r
t
ed
im
ag
e,
an
d
ea
c
h
esti
m
atin
g
m
o
d
el
u
s
ed
th
is
d
is
tan
ce
as
th
e
q
u
ality
s
co
r
e
p
r
ed
ictio
n
.
Z
h
a
n
an
d
Z
h
a
n
g
[
7
]
p
r
o
p
o
s
ed
a
n
ef
f
ec
tiv
e
NR
-
I
QA
m
eth
o
d
f
o
r
ass
es
s
in
g
th
e
q
u
ality
o
f
J
PEG
im
ag
es
b
ased
o
n
co
m
p
u
tin
g
t
h
e
lu
m
in
an
ce
a
n
d
b
l
o
ck
in
ess
in
8
×
8
b
lo
c
k
s
in
a
s
in
g
le
J
PEG
im
ag
e.
On
o
f
th
e
m
ain
lim
itatio
n
s
o
f
th
is
m
eth
o
d
is
th
at
it
h
as
b
ee
n
cu
s
to
m
ize
d
f
o
r
J
PEG
im
ag
es
an
d
it
m
ay
g
iv
e
lim
ited
p
e
r
f
o
r
m
an
ce
with
o
th
er
k
i
n
d
s
o
f
im
ag
es
.
Oszu
s
t
[
8
]
,
in
tr
o
d
u
ce
d
a
n
o
v
el
NR
-
I
QA
ap
p
r
o
ac
h
b
y
ch
ar
ac
te
r
izin
g
th
e
d
is
to
r
ted
im
ag
e
b
y
th
e
d
is
tr
ib
u
tio
n
o
f
th
e
o
r
ien
tatio
n
s
o
f
lo
ca
l
g
r
ad
ien
ts
in
r
eg
io
n
s
o
f
d
if
f
e
r
en
t
s
izes.
W
it
h
th
e
aim
o
f
im
p
r
o
v
in
g
th
e
p
e
r
f
o
r
m
a
n
ce
o
f
NR
-
I
QA
m
eth
o
d
s
,
r
esear
ch
er
s
h
av
e
r
ec
en
tly
em
p
lo
y
e
d
d
ee
p
lear
n
in
g
as th
e
co
r
e
tech
n
o
lo
g
y
to
a
d
d
r
ess
th
e
ch
allen
g
es o
f
q
u
alit
y
ass
es
s
m
en
t.
Mo
s
t
d
ee
p
lear
n
i
n
g
-
b
ased
m
eth
o
d
s
h
av
e
em
p
l
o
y
ed
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
f
r
o
m
i
n
p
u
t
im
a
g
es
to
ass
ess
th
eir
q
u
ality
[
9
]
–
[
1
2
]
.
R
av
ela
et
a
l.
[
1
3
]
,
in
itially
d
eter
m
in
ed
th
e
n
atu
r
e
o
f
th
e
im
ag
e
d
is
to
r
tio
n
.
T
h
en
,
f
o
r
ea
c
h
d
is
tin
ct
f
o
r
m
o
f
d
is
to
r
tio
n
,
th
ey
p
r
ed
ic
ted
th
e
p
er
ce
iv
ed
d
eg
r
ad
atio
n
o
f
im
ag
e
q
u
ality
u
tili
zin
g
C
NN.
Pan
et
a
l.
[
1
4
]
in
tr
o
d
u
ce
d
a
d
is
to
r
tio
n
-
awa
r
e
C
NN
m
o
d
el
ca
lled
DACNN
f
o
r
b
lin
d
I
QA.
T
h
e
DACNN
m
eth
o
d
d
em
o
n
s
tr
ated
ef
f
icac
y
in
h
an
d
lin
g
g
e
n
u
in
e
im
ag
e
d
is
to
r
tio
n
s
as
well
as
th
o
s
e
in
d
u
ce
d
ar
tific
ially
.
Z
h
o
u
et
a
l.
[
1
5
]
,
in
t
r
o
d
u
ce
d
a
n
o
v
el
NR
-
I
QA
ap
p
r
o
ac
h
u
tili
zin
g
a
n
atten
tio
n
al
f
ea
tu
r
e
f
u
s
io
n
m
ec
h
an
is
m
.
I
n
th
is
ap
p
r
o
ac
h
,
th
ick
n
ess
,
lo
ca
l,
an
d
g
lo
b
a
l
in
f
o
r
m
atio
n
wer
e
ex
tr
ac
ted
f
r
o
m
th
e
im
ag
es,
c
o
r
r
esp
o
n
d
in
g
ly
,
u
s
in
g
th
r
ee
C
NNs
n
etwo
r
k
s
,
n
am
ely
VG
G1
9
,
VGG1
6
,
an
d
R
esNet
-
5
0
.
T
h
en
,
th
ese
f
ea
tu
r
es
wer
e
f
u
s
ed
to
f
ac
ilit
ate
th
e
p
er
ce
p
tio
n
o
f
v
ar
io
u
s
ty
p
es
o
f
d
is
to
r
tio
n
s
th
r
o
u
g
h
th
e
in
co
r
p
o
r
atio
n
o
f
atten
tio
n
m
ec
h
an
is
m
s
.
L
in
an
d
W
an
g
[
1
6
]
u
s
ed
g
en
er
ativ
e
ad
v
er
s
ar
ia
l
n
etwo
r
k
s
(
GANs)
f
o
r
NR
-
I
QA.
T
h
e
g
en
er
ativ
e
n
etwo
r
k
was e
m
p
lo
y
ed
to
cr
ea
t
e
a
r
ef
er
en
ce
im
ag
e
th
at
was f
r
ee
o
f
d
is
to
r
tio
n
f
o
r
in
p
u
t
im
ag
e
th
at
was
d
is
to
r
te
d
.
T
h
e
f
ea
tu
r
es
d
er
iv
ed
f
r
o
m
th
e
d
is
to
r
ted
im
ag
e
wer
e
s
u
b
s
eq
u
en
tly
co
m
b
in
ed
with
th
e
in
f
o
r
m
atio
n
e
x
tr
ac
te
d
f
r
o
m
t
h
e
h
allu
cin
ated
r
ef
er
e
n
ce
im
ag
e
in
o
r
d
er
t
o
f
o
r
ec
as
t
th
e
q
u
ality
o
f
t
h
e
p
er
ce
p
tu
al
im
ag
e
.
Oth
er
s
tu
d
i
es
in
liter
atu
r
e
h
av
e
em
p
lo
y
e
d
p
r
e
-
tr
ain
e
d
C
NNs
lik
e
Ale
x
Net
o
r
VGG1
6
as
f
ea
tu
r
e
ex
tr
ac
t
o
r
s
.
B
ian
co
et
a
l.
[
2
]
d
em
o
n
s
tr
ated
th
e
ef
f
ic
ac
y
o
f
a
f
in
e
-
tu
n
in
g
p
r
e
-
tr
ai
n
ed
C
NN
m
o
d
el
to
ex
tr
ac
t
f
ea
tu
r
e
v
ec
to
r
s
f
r
o
m
r
an
d
o
m
r
e
g
io
n
s
o
f
an
in
p
u
t
i
m
ag
e.
Gao
et
a
l.
[
1
7
]
,
o
n
t
h
e
o
th
er
h
an
d
,
u
tili
ze
d
g
lo
b
al
m
in
im
u
m
an
d
m
a
x
im
u
m
p
o
o
lin
g
to
ex
tr
ac
t
f
ea
t
u
r
es
th
at
wer
e
n
o
t
d
ep
en
d
e
n
t
o
n
r
eso
lu
tio
n
f
r
o
m
m
u
ltip
le
lay
er
s
o
f
an
Alex
Net
m
o
d
el.
L
i
k
e
[
2
]
,
th
e
p
er
ce
p
t
u
al
q
u
ality
was
ascer
tain
ed
th
r
o
u
g
h
th
e
ar
ith
m
etic
m
ea
n
o
f
th
e
s
u
b
s
co
r
es
o
b
tain
ed
f
r
o
m
th
e
lay
er
-
wis
e
f
ea
tu
r
e
v
ec
to
r
s
a
n
d
a
tr
ain
ed
s
u
p
p
o
r
t
v
ec
to
r
r
eg
r
ess
o
r
.
Als
o
,
Var
g
a
[
1
8
]
p
r
o
p
o
s
ed
a
NR
-
I
QA
m
eth
o
d
b
ased
o
n
attac
h
in
g
g
lo
b
al
av
er
ag
e
p
o
o
lin
g
lay
er
s
to
I
n
ce
p
tio
n
m
o
d
u
les o
f
C
NN
n
etwo
r
k
p
r
e
-
tr
ain
ed
o
n
I
m
ag
eNe
t
d
ataset.
I
n
g
en
er
al,
m
o
s
t
r
elate
d
wo
r
k
s
ex
h
ib
it
s
o
m
e
lim
itatio
n
s
,
s
u
ch
as
a
lack
o
f
co
m
p
r
e
h
en
s
iv
e
ev
alu
atio
n
,
d
ep
en
d
e
n
ce
o
n
s
p
ec
if
ic
ass
u
m
p
tio
n
s
,
an
d
lim
ited
in
ter
p
r
etab
ilit
y
in
d
e
ep
lea
r
n
in
g
-
b
ased
m
eth
o
d
s
.
I
n
c
o
n
tr
ast
to
all
cu
r
r
en
t
NR
-
I
QA
m
eth
o
d
s
,
we
p
r
esen
t
an
e
f
f
ec
tiv
e
N
R
-
I
QA
m
eth
o
d
in
th
is
p
a
p
er
u
tili
zin
g
d
ee
p
v
is
u
al
in
ter
p
r
etatio
n
s
.
I
n
p
a
r
ticu
lar
,
we
p
r
o
p
o
s
e
ex
tr
ac
tin
g
d
ee
p
v
is
u
al
ex
p
lan
atio
n
s
f
r
o
m
t
h
e
d
eg
r
ad
ed
im
ag
e
a
n
d
th
e
co
r
r
esp
o
n
d
in
g
PR
I
an
d
th
en
m
ea
s
u
r
in
g
th
e
d
is
tan
ce
b
et
wee
n
th
e
v
is
u
al
in
ter
p
r
etatio
n
s
o
f
th
e
in
p
u
t
im
ag
e
an
d
its
PR
I
u
s
in
g
a
f
u
ll
r
e
f
er
e
n
ce
im
ag
e
q
u
ality
ass
ess
m
en
t
m
etr
ic
(
e.
g
.
,
s
ig
n
al
-
to
-
n
o
is
e
r
a
tio
)
.
T
h
e
s
im
ilar
ity
s
co
r
e
th
at
is
o
b
tain
ed
s
er
v
es
as
a
m
etr
ic
to
ass
ess
th
e
q
u
ality
o
f
th
e
i
n
p
u
t
im
ag
es.
T
h
e
Gr
ad
-
C
AM
[
1
9
]
m
eth
o
d
is
u
tili
ze
d
to
ex
tr
ac
t
d
ee
p
v
is
u
al
in
ter
p
r
etatio
n
s
.
I
n
th
is
p
ap
er
,
we
ass
u
m
e
t
h
at
r
o
b
u
s
t
m
eth
o
d
s
,
lik
e
Gr
ad
-
C
AM
,
f
o
r
in
ter
p
r
etin
g
d
ee
p
lear
n
in
g
m
o
d
els
will
p
r
o
d
u
ce
th
e
s
am
e
v
is
u
al
in
ter
p
r
etatio
n
f
o
r
two
id
en
tical
im
ag
es,
an
d
th
e
s
im
ilar
ity
b
etwe
en
th
e
two
v
is
u
al
in
ter
p
r
etatio
n
s
o
f
two
id
en
ti
ca
l
im
ag
es
will
b
e
m
ax
im
u
m
.
T
h
er
ef
o
r
e,
m
ea
s
u
r
i
n
g
th
e
s
im
ilar
ity
b
etwe
en
th
e
v
is
u
al
in
ter
p
r
etatio
n
s
o
f
th
e
d
eg
r
ad
ed
im
a
g
e
an
d
its
P
R
I
ca
n
b
e
a
s
u
itab
le
way
o
f
ass
ess
in
g
NR
im
ag
e
q
u
alit
y
.
Up
to
th
e
au
th
o
r
s
'
k
n
o
wled
g
e,
v
er
y
f
ew
wo
r
k
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
in
th
e
lit
er
atu
r
e
th
at
co
n
s
id
er
v
is
u
al
in
ter
p
r
etatio
n
s
f
o
r
NR
-
I
QA.
Prim
ar
y
r
esu
lts
h
av
e
b
ee
n
p
r
esen
ted
i
n
[
2
0
]
;
h
o
we
v
er
,
in
th
is
p
ap
er
,
we
p
r
esen
t
a
d
etailed
s
tu
d
y
o
f
v
ar
io
u
s
d
is
to
r
tio
n
ty
p
es
an
d
m
an
y
d
atasets
.
A
co
m
p
a
r
ativ
e
s
tu
d
y
with
b
o
t
h
tr
ad
itio
n
al
a
n
d
d
ee
p
lear
n
in
g
NR
-
I
QA
m
eth
o
d
s
is
p
r
o
v
id
e
d
.
Sectio
n
2
ex
p
lain
s
th
e
m
ain
s
tep
s
o
f
th
e
p
r
o
p
o
s
ed
NR
-
I
QA
alg
o
r
ith
m
.
Sectio
n
3
p
r
o
v
i
d
es
th
e
r
esu
lts
o
f
o
u
r
m
eth
o
d
o
n
d
if
f
er
en
t
b
e
n
ch
m
ar
k
d
atasets
co
n
s
id
er
in
g
d
if
f
er
en
t
d
is
to
r
tio
n
ty
p
es,
an
d
also
it
p
r
esen
ts
a
d
etailed
ab
latio
n
s
tu
d
y
an
d
c
o
m
p
ar
is
o
n
with
ex
is
tin
g
NR
-
I
QA
m
eth
o
d
s
.
Sectio
n
4
co
n
clu
d
es
o
u
r
s
tu
d
y
an
d
p
r
o
v
id
es so
m
e
s
u
g
g
esti
o
n
s
f
o
r
f
u
tu
r
e
r
esear
ch
o
n
NR
-
I
QA.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
No
-
r
eferen
ce
ima
g
e
q
u
a
lity a
s
s
es
s
men
t b
a
s
ed
o
n
visu
a
l
…
(
B
a
s
ma
A
h
med
)
1523
2.
M
E
T
H
O
D
Fig
u
r
e
1
d
e
p
icts
th
e
wo
r
k
f
l
o
w
o
f
th
e
s
u
g
g
ested
m
eth
o
d
.
T
h
e
k
ey
s
tep
s
o
f
th
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
ar
e:
−
Simu
latin
g
a
PR
I
,
−
ex
tr
ac
tin
g
d
ee
p
lear
n
in
g
f
ea
tu
r
es th
r
o
u
g
h
d
ee
p
t
r
an
s
f
er
lear
n
in
g
,
−
g
en
er
atin
g
v
is
u
al
ex
p
lan
atio
n
im
ag
e
(
VE
I
)
u
s
in
g
Gr
ad
-
C
AM
,
−
ca
lcu
latin
g
th
e
s
im
ilar
ity
b
etwe
en
th
e
VE
I
o
f
th
e
d
is
to
r
te
d
in
p
u
t
im
ag
e
a
n
d
t
h
e
VE
I
o
f
PR
I
u
s
in
g
a
s
im
ilar
ity
m
ea
s
u
r
e
lik
e
th
e
p
ea
k
s
ig
n
al
-
to
-
n
o
is
e
-
r
atio
(
PS
NR
)
,
−
em
p
lo
y
in
g
a
n
o
n
lin
ea
r
m
ap
p
in
g
f
u
n
ctio
n
to
tr
an
s
f
o
r
m
q
u
ality
s
co
r
es
i
n
to
a
f
in
al
NR
-
I
QA
s
co
r
e,
q
u
an
tify
in
g
th
e
o
v
er
all
q
u
ality
o
f
th
e
d
is
to
r
ted
in
p
u
t im
ag
e.
I
n
th
e
s
u
b
s
eq
u
en
t su
b
s
ec
tio
n
s
,
we
ex
p
lain
th
e
k
ey
s
tep
s
o
f
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
in
d
etail.
Fig
u
r
e
1
.
T
h
e
wo
r
k
f
lo
w
o
f
th
e
p
r
o
p
o
s
ed
NR
-
I
QA
ap
p
r
o
ac
h
2
.
1
.
Ste
p 1
:
s
im
ula
t
ing
P
RI
Min
et
a
l.
[
6
]
,
e
m
p
lo
y
ed
th
e
s
am
e
d
eg
r
a
d
atio
n
p
r
o
ce
s
s
th
at
g
en
er
ates
th
e
d
is
to
r
ted
im
ag
e
to
f
u
r
th
e
r
d
eg
r
ad
e
it
is
u
s
ed
t
o
p
r
o
d
u
ce
a
PR
I
im
ag
e.
I
n
th
e
ca
s
e
o
f
v
ar
y
in
g
d
is
to
r
tio
n
ty
p
es,
a
d
eg
r
ad
atio
n
m
o
d
el
f
o
r
ea
ch
ty
p
e
is
d
ef
in
ed
.
I
n
th
is
s
tu
d
y
,
Gau
s
s
ian
b
lu
r
(
GB
)
,
wh
ite
n
o
is
e
(
W
N)
d
is
to
r
tio
n
,
J
PEG
an
d
J
PEG2
0
0
0
co
m
p
r
ess
io
n
,
ar
e
co
n
s
id
er
e
d
.
We
b
r
ief
ly
in
tr
o
d
u
ce
ea
ch
d
eg
r
ad
atio
n
m
o
d
el
:
−
Gau
s
s
ian
b
lu
r
d
eg
r
a
d
atio
n
m
o
d
el
[
2
1
]
:
f
o
llo
win
g
th
e
f
ac
t
th
at
ex
ce
s
s
iv
e
b
lu
r
in
tr
o
d
u
ce
s
s
p
u
r
io
u
s
s
tr
u
ctu
r
es
th
at
ca
n
b
e
u
s
ed
a
s
a
r
ef
e
r
en
ce
im
ag
e
f
o
r
th
e
b
lu
r
r
ed
im
ag
e
.
T
o
d
eter
m
in
e
th
e
PR
I
,
a
3
×3
av
er
ag
in
g
f
ilter
is
u
s
ed
.
Giv
en
a
“Y”
d
is
to
r
ted
im
ag
e,
th
e
PR
I
is
d
eter
m
in
ed
as:
PR
I
=f
*
Y=
1
9
[
1
1
1
1
1
1
1
1
1
]
*Y
(
1
)
wh
er
e
"f
"
is
th
e
b
lu
r
r
in
g
f
ilter
,
we
co
n
f
ig
u
r
ed
it a
s
an
a
v
er
ag
i
n
g
f
ilter
,
a
n
d
*
is
th
e
co
n
v
o
lu
t
io
n
o
p
e
r
ato
r
.
−
W
h
ite
n
o
is
e
d
eg
r
ad
atio
n
m
o
d
el:
i
n
th
is
ca
s
e,
P
R
I
is
f
o
r
m
ed
b
y
in
tr
o
d
u
cin
g
Gau
s
s
ian
n
o
is
e
to
th
e
d
is
to
r
ted
im
ag
e
Y
[
6
]
:
PR
I
=
Y
+
N
(
0
,
σ
w
)
(
2
)
wh
er
e
wh
ite
n
o
is
e
N(
0
,
σ
W
)
with
a
m
ea
n
o
f
ze
r
o
a
n
d
a
σ
w
s
tan
d
ar
d
d
e
v
iatio
n
.
−
J
PEG
an
d
J
PEG2
0
0
0
c
o
m
p
r
es
s
io
n
[
2
2
]
:
h
er
e
,
PR
I
is
o
b
tain
e
d
b
y
co
m
p
r
ess
in
g
th
e
d
is
to
r
ted
im
ag
e.
Giv
en
th
e
d
is
to
r
ted
im
ag
e
Y,
PR
I
is
p
r
o
d
u
ce
d
as:
PR
I
=
J
PEG
(
Y
,
QT
)
(
3
)
I
n
th
is
ex
p
r
ess
io
n
,
J
PEG
r
ef
er
s
to
th
e
J
PEG
en
co
d
er
,
wh
ile
QT
r
ef
er
s
to
th
e
q
u
an
tizatio
n
t
ab
le
u
s
ed
wh
ich
is
f
ix
ed
an
d
r
ep
r
esen
ts
v
er
y
lo
w
co
m
p
r
ess
io
n
q
u
ality
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
:
1
5
2
1
-
1
5
3
1
1524
2
.
2
.
Ste
p 2
:
e
x
t
r
a
ct
ing
f
e
a
t
ure
m
a
ps
utilizing
pret
ra
ined
CNN
m
o
dels
T
r
an
s
f
er
lear
n
in
g
h
as
d
em
o
n
s
tr
ated
its
ab
ilit
y
to
ac
h
iev
e
cu
tt
in
g
-
ed
g
e
p
er
f
o
r
m
an
ce
ac
r
o
s
s
a
r
an
g
e
o
f
ap
p
licatio
n
s
,
p
ar
ticu
lar
ly
wh
e
n
co
n
f
r
o
n
ted
with
lim
ited
d
atasets
[
2
3
]
.
T
h
e
em
p
lo
y
m
en
t
o
f
p
r
e
-
tr
ain
ed
C
NNs
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
allo
ws lev
er
ag
in
g
t
h
e
p
o
wer
f
u
l r
ep
r
es
en
tatio
n
ca
p
ab
ilit
ies o
f
a
p
r
e
-
t
r
ain
ed
C
NN
with
o
u
t
n
ee
d
in
g
to
tr
ain
an
en
tire
n
ew
n
etwo
r
k
f
r
o
m
s
cr
atch
.
W
ith
in
th
e
s
co
p
e
o
f
th
is
s
tu
d
y
,
we
w
ill
em
p
lo
y
tr
an
s
f
er
lear
n
in
g
,
lev
e
r
ag
in
g
b
o
th
p
r
e
-
tr
ain
ed
an
d
f
in
e
-
tu
n
e
d
n
etwo
r
k
s
.
Sp
ec
if
ically
,
we
em
p
lo
y
a
s
et
o
f
r
o
b
u
s
t
p
r
e
-
tr
ain
ed
C
NNs
(
p
r
etr
ain
ed
o
n
I
m
ag
eNe
t
d
ataset)
,
s
u
ch
as
Dar
k
N
et1
9
,
I
n
ce
p
tio
n
R
esNetV2
an
d
VGG1
6
[
2
4
]
to
ex
tr
ac
t
f
ea
tu
r
e
m
ap
s
f
r
o
m
th
e
in
p
u
t
d
is
to
r
ted
im
a
g
e
an
d
th
e
co
r
r
esp
o
n
d
in
g
PR
I
.
E
ac
h
f
e
atu
r
e
m
ap
ca
p
tu
r
es
d
if
f
er
en
t
le
v
els
o
f
a
b
s
tr
ac
tio
n
,
f
r
o
m
b
asic
ed
g
es
an
d
tex
tu
r
e
s
in
th
e
ea
r
lier
lay
er
s
to
co
m
p
lex
o
b
ject
p
ar
ts
an
d
h
ig
h
er
-
le
v
el
co
n
ce
p
ts
in
th
e
d
ee
p
er
lay
er
s
.
T
a
b
le
1
p
r
esen
ts
a
b
r
ief
d
escr
ip
tio
n
o
f
th
e
p
r
etr
ain
ed
C
NN
m
o
d
els
u
s
ed
in
th
is
s
tu
d
y
.
T
ab
le
1
.
Descr
ip
tio
n
o
f
th
e
p
r
e
tr
ain
ed
C
NN
m
o
d
els
P
r
e
t
r
a
i
n
e
d
m
o
d
e
l
I
n
p
u
t
i
m
a
g
e
s
i
z
e
D
e
p
t
h
N
o
.
o
f
p
a
r
a
m
e
t
e
r
s (m
i
l
l
i
o
n
)
D
a
r
k
n
e
t
1
9
2
5
6
×
2
5
6
19
2
0
.
8
I
n
c
e
p
t
i
o
n
R
e
sN
e
t
V
2
2
9
9
×
2
9
9
1
6
4
5
5
.
9
V
G
G
1
6
2
2
4
×
2
2
4
16
1
3
8
T
h
e
p
r
etr
ai
n
ed
C
NN
m
o
d
el
u
s
ed
in
th
is
s
tu
d
y
ca
n
b
e
d
o
wn
lo
a
d
ed
f
r
o
m
h
ttp
s
://www.
m
ath
wo
r
k
s
.
co
m
/.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
NR
-
I
QA
alg
o
r
ith
m
s
th
at
r
ely
o
n
tr
an
s
f
er
lear
n
i
n
g
h
av
e
o
f
ten
in
co
r
p
o
r
ated
d
ata
au
g
m
en
tatio
n
to
en
h
an
ce
t
h
e
d
iv
er
s
ity
o
f
tr
ain
in
g
d
atasets
,
r
esu
ltin
g
in
im
p
r
o
v
e
d
p
er
f
o
r
m
an
ce
[
2
5
]
.
I
n
th
is
p
a
p
er
,
m
u
ltip
le
im
ag
e
tr
a
n
s
f
o
r
m
atio
n
s
h
av
e
b
ee
n
u
s
ed
,
wh
i
ch
in
clu
d
e
r
o
tatio
n
,
zo
o
m
,
h
o
r
izo
n
tal/v
er
tical
f
li
p
s
,
an
d
h
o
r
izo
n
tal/v
er
tical
s
h
if
ts
[
2
3
]
,
[
2
6
]
.
I
t
is
wo
r
th
n
o
tin
g
th
at
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
lin
ea
r
f
u
n
ctio
n
s
is
im
p
r
o
v
ed
b
y
in
cr
ea
s
in
g
th
e
q
u
an
tity
o
f
d
is
to
r
ted
in
p
u
t
im
ag
es
u
s
ed
in
th
is
wo
r
k
b
y
in
cr
ea
s
in
g
an
d
d
iv
er
s
if
y
in
g
th
e
s
ize
o
f
th
e
d
atab
ases
in
p
u
t
to
th
e
p
r
e
-
tr
ain
e
d
m
o
d
els.
Alth
o
u
g
h
th
ese
m
o
d
els
ar
e
tr
ain
ed
o
n
v
er
y
lar
g
e
d
ata
b
ases
,
th
ey
m
ay
b
e
in
co
n
s
is
ten
t
with
s
o
m
e
task
s
,
s
o
we
in
cr
ea
s
e
th
e
in
p
u
t
d
ata
to
im
p
r
o
v
e
t
h
eir
p
er
f
o
r
m
an
ce
[
2
7
]
.
2
.
3
.
Ste
p 3
:
g
ener
a
t
ing
VE
I
A
VE
I
is
a
h
ea
t
m
ap
th
at
id
e
n
tifie
s
th
e
im
p
o
r
tan
t
r
e
g
io
n
s
in
th
e
in
p
u
t
im
a
g
e.
I
t
d
ep
icts
th
e
m
o
s
t
im
p
o
r
tan
t
ar
ea
s
o
f
th
e
im
ag
e
th
at
c
o
n
tr
ib
u
te
to
th
e
d
ec
is
io
n
to
class
if
y
th
e
m
o
d
el
f
o
r
a
p
ar
ticu
lar
class
.
Var
io
u
s
m
eth
o
d
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
in
last
y
ea
r
s
to
e
x
p
lai
n
C
NNs,
s
u
ch
as
Gr
ad
-
C
AM
,
g
u
id
ed
B
ac
k
p
r
o
p
,
class
-
ac
tiv
atio
n
m
ap
p
in
g
(
C
AM
)
,
s
alien
cy
m
a
p
,
a
n
d
o
cc
lu
s
io
n
m
a
p
.
I
n
t
h
is
p
ap
er
,
we
em
p
lo
y
th
e
Gr
ad
-
C
AM
g
en
er
ated
f
r
o
m
p
r
e
-
tr
ai
n
ed
C
NNs
to
g
en
er
ate
VE
I
s
o
f
th
e
d
is
to
r
ted
in
p
u
t
im
ag
e
an
d
th
e
PR
I
.
Gr
ad
-
C
AM
is
a
wid
ely
u
s
ed
v
is
u
al
ex
p
la
n
atio
n
m
eth
o
d
an
d
it
p
r
o
v
id
es
a
cc
u
r
ate
in
ter
p
r
etatio
n
w
h
en
c
o
m
p
ar
ed
with
o
th
e
r
m
eth
o
d
s
[
2
8
]
.
Gr
ad
-
C
AM
is
a
v
is
u
aliza
tio
n
tech
n
iq
u
e
u
s
ed
t
o
o
b
tain
a
cla
s
s
-
d
is
cr
im
in
ativ
e
lo
ca
lizatio
n
m
ap
(
i.e
.
,
VE
I
)
,
d
en
o
te
d
as
L
G
r
ad
−
CAM
c
∈
R
u
×
v
in
g
en
er
i
c
C
NN
-
b
ased
ar
ch
itectu
r
es.
T
h
e
Gr
ad
-
C
AM
m
eth
o
d
h
as
s
ev
er
al
s
tep
s
s
tar
tin
g
with
th
e
co
m
p
u
tatio
n
o
f
th
e
g
r
ad
ien
t
o
f
th
e
p
r
ed
icted
class
y
c
with
r
esp
ec
t
to
f
ea
tu
r
e
m
ap
s
A
o
f
a
co
n
v
o
lu
tio
n
al
lay
er
,
r
ep
r
esen
ted
as
∂
y
c
∂
A
ij
k
.
T
o
g
et
weig
h
ts
,
th
ese
g
r
ad
ien
ts
ar
e
g
l
o
b
ally
av
er
a
g
e
-
p
o
o
led
,
d
en
o
ted
as
α
k
c
,
wh
ich
ca
n
b
e
co
m
p
u
ted
as
[
2
8
]
:
α
k
c
=
1
Z
∑
∑
∂
y
c
∂
A
ij
k
j
i
(
4
)
wh
er
e
th
e
weig
h
t
α
k
c
r
ep
r
esen
ts
a
p
ar
tial
lin
ea
r
izatio
n
o
f
th
e
d
ee
p
n
etwo
r
k
d
o
wn
s
tr
ea
m
f
r
o
m
A
,
an
d
th
ese
weig
h
ts
ca
p
tu
r
e
th
e
im
p
o
r
tan
ce
o
r
r
elev
an
ce
o
f
ea
ch
f
ea
tu
r
e
m
ap
ch
an
n
el
k
f
o
r
th
e
tar
g
et
class
c.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
y
c
co
u
ld
b
e
an
y
d
if
f
er
en
tiab
le
ac
tiv
atio
n
.
T
h
e
Gr
ad
-
C
AM
h
ea
t
-
m
ap
is
g
en
er
ated
as
a
weig
h
te
d
c
o
m
b
in
atio
n
o
f
f
ea
t
u
r
e
m
ap
s
u
s
in
g
th
e
o
b
tain
ed
weig
h
ts
.
Sp
ec
if
ically
,
it
is
co
m
p
u
ted
as
th
e
n
o
n
-
lin
ea
r
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
ac
tiv
atio
n
f
u
n
ctio
n
a
p
p
lied
to
th
e
s
u
m
o
f
α
k
c
m
u
ltip
lied
b
y
A
k
ac
r
o
s
s
all
ch
an
n
els,
as:
L
G
r
ad
−
CAM
c
=
R
e
L
U
(
∑
α
k
c
A
k
k
)
(
5
)
T
h
e
r
esu
ltin
g
Gr
ad
-
C
AM
h
ea
t
-
m
ap
r
e
p
r
esen
ts
a
v
is
u
aliza
tio
n
o
f
th
e
im
p
o
r
tan
t
r
eg
io
n
s
in
t
h
e
f
ea
tu
r
e
m
ap
s
f
o
r
th
e
tar
g
et
class
.
T
h
is
h
ea
t
-
m
ap
is
u
s
ed
as
a
VE
I
.
No
tab
ly
,
th
e
VE
I
s
o
f
t
h
e
d
is
to
r
ted
im
ag
e
an
d
its
co
r
r
esp
o
n
d
in
g
PR
I
ar
e
th
e
s
am
e
s
ize.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
No
-
r
eferen
ce
ima
g
e
q
u
a
lity a
s
s
es
s
men
t b
a
s
ed
o
n
visu
a
l
…
(
B
a
s
ma
A
h
med
)
1525
2
.
4
.
Ste
p 4
:
s
im
ila
rit
y
c
a
lcula
t
io
n
I
n
th
is
p
ap
er
,
v
ar
io
u
s
s
im
ilar
ity
m
ea
s
u
r
es
ar
e
ass
ess
ed
in
clu
d
in
g
,
g
r
a
d
ien
t
m
ag
n
itu
d
e
s
im
ilar
ity
d
ev
iatio
n
(
GM
SD
)
[
2
9
]
,
p
ea
k
s
ig
n
al
to
n
o
is
e
r
atio
(
PS
NR
)
[
3
0
]
,
NC
C
[
3
0
]
,
f
ea
tu
r
e
s
im
i
lar
ity
(
FS
I
M
)
[
3
1
]
,
n
o
r
m
alis
ed
ab
s
o
lu
te
e
r
r
o
r
(
N
AE
)
[
3
0
]
,
v
is
u
al
s
alien
cy
-
b
ased
in
d
ex
(
VSI
)
[
3
2
]
a
n
d
s
tr
u
ct
u
r
al
s
im
ilar
ity
in
d
ex
(
SS
I
M
)
[
3
3
]
.
T
h
e
ab
latio
n
s
tu
d
y
(
s
ec
tio
n
3
.
5
)
d
em
o
n
s
tr
ated
th
at
th
e
PS
NR
y
ield
s
to
th
e
b
e
s
t
NR
-
I
QA
r
esu
lts
,
an
d
th
u
s
it is
em
p
lo
y
ed
in
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
.
2
.
5
.
Ste
p 5
:
m
a
pp
ing
f
un
ct
io
n f
o
r
s
co
re
a
lig
nm
ent
Oszu
s
t
[
8
]
,
em
p
lo
y
a
f
iv
e
-
p
a
r
am
eter
lo
g
is
tic
f
u
n
ctio
n
as
a
n
o
n
lin
ea
r
m
ap
p
i
n
g
f
u
n
ctio
n
to
ass
ig
n
q
u
ality
s
co
r
es:
q
=
̀
β
1
(
1
2
-
1
1+
e
xp
(
β
2
(q
-
β
3
))
)
+
β
4
q+
β
5
(
6
)
T
h
e
p
a
r
am
eter
s
{
β
j
|
j
=
1
,
2
,
…
.
5
}
ar
e
th
o
s
e
f
iv
e
p
ar
am
eter
s
d
er
iv
e
d
th
r
o
u
g
h
cu
r
v
e
f
itti
n
g
.
T
h
e
o
r
i
g
in
al
a
n
d
m
ap
p
ed
q
u
ality
s
co
r
es,
d
e
n
o
te
d
as
q
̀
an
d
q
,
r
esp
ec
tiv
ely
.
2
.
6
.
E
v
a
lua
t
i
o
n m
et
rics
I
n
th
is
s
tu
d
y
,
th
r
ee
p
er
f
o
r
m
an
ce
m
etr
ics
wer
e
u
s
ed
to
ev
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
o
f
NR
-
I
QA
m
eth
o
d
s
:
i
)
p
ea
r
s
o
n
lin
ea
r
c
o
r
r
elatio
n
co
e
f
f
icien
t
(
PLCC
)
,
ii
)
Sp
ea
r
m
a
n
'
s
r
an
k
o
r
d
e
r
ed
co
r
r
elatio
n
co
e
f
f
icien
t
(
SR
OC
C
)
[
3
4
]
,
an
d
iii
)
r
o
o
t
-
m
ea
n
-
s
q
u
ar
e
e
r
r
o
r
(
R
MSE
)
[
6
]
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
B
enchm
a
r
k
da
t
a
ba
s
es
Fo
u
r
s
tan
d
ar
d
I
QA
d
atab
ases
,
in
clu
d
in
g
L
I
VE
[
3
5
]
,
C
SIQ
[
3
6
]
,
SIQ
AD
[
3
7
]
,
a
n
d
T
I
D
2
0
1
3
[
3
8
]
,
ar
e
u
s
ed
in
th
is
s
tu
d
y
t
o
ev
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
Mo
r
e
o
v
er
,
f
o
u
r
co
m
m
o
n
d
is
to
r
tio
n
ty
p
es:
GB
,
W
N,
J
PEG,
an
d
J
P2
K,
ar
e
co
n
s
id
er
ed
in
o
u
r
ex
p
er
im
e
n
ts
.
Deta
iled
in
f
o
r
m
atio
n
ab
o
u
t
th
e
d
atab
ases
is
s
u
m
m
ar
ized
in
T
a
b
le
2
.
T
ab
le
2
.
Deta
iled
in
f
o
r
m
atio
n
o
n
th
e
b
en
ch
m
a
r
k
d
ata
b
ases
B
e
n
c
h
mar
k
d
a
t
a
b
a
se
s
N
o
.
o
f
r
e
f
e
r
e
n
c
e
im
a
g
e
s
N
o
.
o
f
G
B
d
i
s
t
o
r
t
e
d
im
a
g
e
s
N
o
.
o
f
W
N
d
i
s
t
o
r
t
e
d
i
ma
g
es
N
o
.
o
f
JP
EG
d
i
s
t
o
r
t
e
d
i
ma
g
e
s
N
o
.
o
f
JP
2
K
d
i
s
t
o
r
t
e
d
i
ma
g
e
s
S
c
o
r
e
t
y
p
e
LI
V
E
[
3
5
]
29
1
4
5
1
4
5
1
6
9
1
7
5
D
M
O
S
[
0
,
1
0
0
]
C
S
I
Q
[
3
6
]
30
1
5
0
1
5
0
1
5
0
1
5
0
D
M
O
S
[
0
,
1
]
S
I
Q
A
D
[
3
7
]
20
1
4
0
1
4
0
1
4
0
1
4
0
D
M
O
S
[
0
,
1
0
0
]
TI
D
2
0
1
3
[
3
8
]
25
1
2
5
1
2
5
1
2
5
1
2
5
M
O
S
[
0
,
9
]
3
.
2
.
I
m
ple
m
ent
a
t
io
n det
a
ils
a
nd
pa
ra
m
et
er
s
et
t
ing
A
f
ew
p
ar
am
eter
s
a
r
e
r
eq
u
ir
e
d
to
b
e
s
et
in
th
is
s
tu
d
y
.
Fo
r
th
e
Gau
s
s
ian
b
lu
r
,
a
P
R
I
is
o
b
tain
ed
b
y
em
p
lo
y
in
g
a
3
×
3
a
v
er
ag
e
f
ilter
; f
o
r
wh
ite
n
o
is
e,
a
b
lu
e
PR
I
is
g
en
er
ated
b
y
c
o
m
b
i
n
in
g
Gau
s
s
ian
n
o
is
e
with
an
av
er
ag
e
v
ar
ia
n
ce
o
f
0
a
n
d
0
.
5
v
ar
ian
ce
;
an
d
f
o
r
J
PEG
an
d
J
P2
K,
th
e
J
PEG
en
co
d
er
in
MA
T
L
AB
is
u
tili
ze
d
.
T
h
e
"im
wr
ite"
f
u
n
ctio
n
is
u
til
ized
in
p
ar
ticu
la
r
.
Settin
g
th
e
"Qu
ality
"
p
ar
am
eter
o
f
"im
wr
ite"
to
th
e
ab
s
o
lu
te
v
alu
e
0
co
r
r
esp
o
n
d
s
to
th
e
q
u
an
tizatio
n
tab
le
QT
.
I
t
r
e
p
r
e
s
en
ts
th
e
m
o
s
t
s
tr
in
g
en
t
co
m
p
r
ess
io
n
th
at
ca
n
b
e
ac
h
iev
ed
b
y
th
e
en
co
d
er
.
C
o
r
n
er
d
etec
tio
n
is
ac
co
m
p
lis
h
ed
u
s
in
g
th
e
MA
T
L
AB
ap
p
licatio
n
o
f
th
e
m
in
im
u
m
eig
en
v
alu
e
m
eth
o
d
[
3
9
]
.
T
h
e
f
u
n
ctio
n
in
q
u
esti
o
n
is
th
e
"c
o
r
n
er
"
f
u
n
ctio
n
.
I
n
p
r
ac
tice,
th
e
r
e
is
n
o
r
estrictio
n
o
n
th
e
u
t
m
o
s
t
n
u
m
b
er
o
f
d
etec
ted
co
r
n
er
s
,
as
th
e
v
alu
e
is
s
et
to
a
n
e
x
tr
em
ely
h
ig
h
v
alu
e.
T
h
e
v
alu
e
o
f
0
:
0
0
1
is
s
p
ec
if
ied
f
o
r
th
e
"Qu
ality
L
ev
el"
p
ar
am
eter
,
wh
ich
d
ete
r
m
in
es
th
e
m
in
im
u
m
co
r
n
e
r
q
u
ality
.
I
n
o
r
d
e
r
to
en
h
an
ce
th
e
d
etec
tio
n
o
f
co
r
n
er
s
f
o
r
th
e
p
u
r
p
o
s
e
o
f
d
escr
ib
i
n
g
s
tr
u
ctu
r
es,
th
e
q
u
ality
p
ar
a
m
eter
is
in
itialized
to
an
ex
tr
em
ely
lo
w
v
alu
e
.
All
ex
p
er
im
en
ts
h
av
e
b
ee
n
ca
r
r
i
ed
o
u
t
o
n
a
co
m
p
u
ter
with
a
2
.
4
GHz
I
n
tel
C
o
r
e
i5
-
2
4
3
0
M
C
PU a
n
d
8
GB
R
AM
.
3
.
3
.
Ana
ly
s
ing
t
he
perf
o
r
m
a
nce
o
f
t
he
pro
po
s
ed
m
et
ho
d
T
h
is
s
ec
tio
n
ass
ess
es
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
s
u
g
g
ested
tech
n
iq
u
e
tak
in
g
in
to
ac
co
u
n
t
f
o
u
r
ty
p
ical
d
is
to
r
tio
n
k
in
d
s
(
GB
,
W
N,
J
PEG,
an
d
J
P2
K)
an
d
f
o
u
r
d
if
f
er
en
t
b
en
c
h
m
ar
k
I
QA
d
atasets
:
L
I
VE
,
SIQ
AD,
C
SIQ
,
an
d
T
I
D2
0
1
3
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
em
p
lo
y
s
a
p
r
etr
ain
ed
I
n
ce
p
tio
n
R
esNetV2
m
o
d
el
an
d
PS
NR
as
a
s
im
ilar
ity
m
ea
s
u
r
e.
As
p
r
esen
ted
in
T
ab
le
3
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
d
em
o
n
s
tr
ates
r
o
b
u
s
t
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
iv
er
s
e
b
en
c
h
m
ar
k
d
atasets
an
d
d
is
to
r
tio
n
ty
p
es,
as
ev
id
e
n
ce
d
b
y
PLCC
an
d
SR
OC
C
v
alu
es.
Ou
r
m
eth
o
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
:
1
5
2
1
-
1
5
3
1
1526
ac
h
iev
es
a
r
em
ar
k
ab
le
PLCC
o
f
0
.
9
8
7
in
th
e
L
I
VE
d
ataset
f
o
r
J
PEG
d
is
to
r
tio
n
an
d
h
ig
h
SR
OC
C
v
alu
es.
A
s
tr
o
n
g
co
r
r
elatio
n
o
b
s
er
v
e
d
in
th
e
L
I
VE
d
ataset
ac
r
o
s
s
all
d
is
to
r
tio
n
ty
p
es,
em
p
h
as
izin
g
th
e
m
eth
o
d
'
s
ef
f
ec
tiv
en
ess
in
ca
p
tu
r
in
g
im
a
g
e
q
u
ality
v
ar
iatio
n
s
.
T
h
e
m
eth
o
d
e
x
h
ib
its
r
eliab
le
p
r
ed
ictio
n
s
in
th
e
C
SIQ
d
ataset,
ex
ce
llin
g
n
o
tab
ly
i
n
J
PEG
d
is
to
r
tio
n
.
Ho
wev
er
,
th
e
T
I
D2
0
1
3
d
ata
s
et
r
ev
ea
ls
v
ar
ied
p
er
f
o
r
m
a
n
ce
,
with
h
ig
h
co
r
r
elatio
n
in
W
N
an
d
J
PEG
d
is
to
r
tio
n
s
b
u
t
co
m
p
ar
ativ
ely
lo
wer
v
alu
es
i
n
J
P2
K
(
PLCC
o
f
0
.
8
7
8
)
.
B
esid
es,
th
e
lo
w
R
MSE
v
alu
es
ac
r
o
s
s
d
atasets
f
u
r
th
er
u
n
d
er
s
co
r
e
th
e
ac
cu
r
ac
y
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
'
s
p
r
ed
ictio
n
s
.
Desp
ite
m
in
o
r
f
lu
ctu
atio
n
s
i
n
PLCC
,
S
R
O
C
C
an
d
R
M
SE
v
alu
es,
th
e
o
v
er
all
s
tr
o
n
g
co
r
r
e
latio
n
an
d
lo
w
e
r
r
o
r
m
etr
ics
r
ev
ea
ls
th
e
m
eth
o
d
'
s
ef
f
ec
tiv
en
ess
in
p
r
o
v
i
d
in
g
r
el
iab
le
im
ag
e
q
u
ality
ass
ess
m
en
ts
ac
r
o
s
s
d
iv
er
s
e
s
ce
n
ar
i
o
s
.
Fig
u
r
e
2
p
r
esen
ts
ex
am
p
les
o
f
th
e
in
p
u
t
d
is
to
r
t
ed
im
ag
es
an
d
th
e
co
r
r
esp
o
n
d
in
g
PR
I
with
th
e
GB
,
W
N,
J
PEG
an
d
J
P2
K
d
is
to
r
tio
n
s
co
n
s
id
er
in
g
d
if
f
er
e
n
t
lev
els
o
f
d
is
to
r
tio
n
s
.
As
o
n
e
ca
n
s
ee
,
th
e
VE
I
s
o
f
th
e
d
is
to
r
ted
im
ag
es
an
d
PR
I
ar
e
q
u
ite
s
im
ilar
in
th
e
c
ase
o
f
GB
,
J
PE
G
an
d
J
P2
K,
a
n
d
th
e
s
am
e
im
p
o
r
tan
t
r
e
g
io
n
s
ar
e
h
ig
h
lig
h
ted
i
n
th
e
VE
I
s
o
f
th
e
d
is
to
r
ted
im
ag
es
an
d
PR
I
.
I
n
tu
r
n
,
s
o
m
e
s
im
ilar
ities
ap
p
ea
r
in
th
e
V
E
I
s
o
f
th
e
d
is
to
r
ted
im
ag
es
an
d
PR
I
,
b
ec
a
u
s
e
o
f
t
h
e
W
N
in
th
e
Gr
a
d
-
C
AM
.
Fig
u
r
e
3
s
h
o
ws
d
if
f
e
r
en
t
ex
am
p
les
o
f
q
u
ality
s
co
r
es
f
o
r
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
.
As
o
n
e
ca
n
s
ee
,
less
n
o
is
y
im
a
g
es
h
av
e
h
i
g
h
er
p
r
ed
icted
s
co
r
es.
T
h
e
p
r
ed
icted
s
co
r
es p
r
o
v
id
e
g
o
o
d
p
r
ed
ictio
n
s
f
o
r
GB
,
W
N,
J
PEG,
an
d
J
P2
K
d
is
to
r
tio
n
s
.
T
ab
le
3
.
Ass
ess
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
in
ter
m
s
o
f
th
e
PLCC
,
SR
O
C
C
an
d
R
MSE
o
n
th
r
ee
b
en
c
h
m
ar
k
d
atab
ases
o
n
f
o
u
r
d
is
to
r
tio
n
ty
p
es
M
e
t
r
i
c
B
e
n
c
h
mar
k
d
a
t
a
se
t
D
i
st
o
r
t
i
o
n
GB
WN
JP
EG
JP
2
K
P
LC
C
LI
V
E
0
.
9
7
2
0
.
8
9
1
0
.
9
8
7
0
.
9
3
9
C
S
I
Q
0
.
9
4
2
0
.
8
4
5
0
.
9
5
6
0
.
9
3
4
TI
D
2
0
1
3
0
.
9
3
6
0
.
9
5
6
0
.
8
7
8
0
.
8
2
6
S
R
O
C
C
LI
V
E
0
.
9
1
9
0
.
9
9
2
0
.
9
7
4
0
.
9
5
5
C
S
I
Q
0
.
9
4
2
0
.
9
5
4
0
.
9
6
3
0
.
9
4
3
TI
D
2
0
1
3
0
.
9
6
0
0
.
9
3
9
0
.
9
7
4
0
.
9
4
8
R
M
S
E
LI
V
E
8
.
0
6
2
6
.
8
4
9
6
.
0
3
7
7
.
9
4
6
C
S
I
Q
0
.
1
1
0
0
.
4
9
7
0
.
0
8
7
0
.
0
9
7
TI
D
2
0
1
3
0
.
4
7
3
0
.
2
9
8
0
.
4
0
0
0
.
5
9
6
Fig
u
r
e
2
.
E
x
am
p
les o
f
th
e
in
p
u
t d
is
to
r
ted
im
ag
es a
n
d
th
e
co
r
r
esp
o
n
d
i
n
g
PR
I
with
th
e
GB
,
W
N.
J
PE
G
an
d
J
P2
K
d
is
to
r
tio
n
s
co
n
s
id
er
in
g
d
if
f
er
en
t le
v
els o
f
d
is
to
r
tio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
No
-
r
eferen
ce
ima
g
e
q
u
a
lity a
s
s
es
s
men
t b
a
s
ed
o
n
visu
a
l
…
(
B
a
s
ma
A
h
med
)
1527
Fig
u
r
e
3
.
E
x
am
p
les o
f
th
e
q
u
a
lity
s
co
r
es o
f
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
with
d
if
f
er
en
t
d
is
to
r
tio
n
t
y
p
es
3
.
4
.
Abla
t
io
n study
First,
we
s
tu
d
y
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
d
if
f
er
en
t
d
ee
p
p
r
e
-
tr
ai
n
ed
f
ea
tu
r
e
ex
tr
ac
to
r
.
T
a
b
le
4
d
em
o
n
s
tr
ates
th
e
ef
f
icac
y
o
f
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
with
d
i
f
f
er
en
t
d
ee
p
p
r
e
-
tr
ain
ed
f
ea
tu
r
e
ex
tr
ac
to
r
s
ac
r
o
s
s
v
ar
io
u
s
b
en
ch
m
ar
k
d
atasets
an
d
d
is
to
r
tio
n
ty
p
es.
No
tewo
r
th
y
is
th
e
c
o
n
s
is
ten
tly
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
u
tili
zin
g
I
n
ce
p
tio
n
R
esNet
V2
,
ev
id
en
t
in
h
ig
h
PLCC
v
alu
es.
Fo
r
in
s
tan
ce
,
in
t
h
e
L
I
VE
d
atas
et,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ac
h
iev
es
a
r
em
ar
k
a
b
le
PLCC
o
f
0
.
9
7
2
f
o
r
J
PEG
d
is
to
r
tio
n
,
s
u
r
p
ass
in
g
b
o
th
V
GG1
6
(
0
.
9
6
3
)
a
n
d
Dar
k
N
et1
9
(
0
.
9
6
9
)
.
Similar
ly
,
in
th
e
C
SIQ
d
ataset,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
attain
s
a
PLCC
o
f
0
.
9
5
6
f
o
r
J
PEG
d
is
to
r
tio
n
,
o
u
tp
e
r
f
o
r
m
in
g
V
GG1
6
(
0
.
9
3
1
)
a
n
d
Dar
k
N
et1
9
(
0
.
9
2
8
)
.
T
h
ese
n
u
m
er
ical
ex
am
p
les
u
n
d
e
r
s
co
r
e
th
e
m
eth
o
d
'
s
r
o
b
u
s
tn
ess
an
d
h
ig
h
lig
h
t
th
e
im
p
ac
tf
u
l
r
o
le
o
f
th
e
I
n
ce
p
tio
n
R
esNetV2
f
ea
tu
r
e
ex
tr
ac
to
r
in
co
n
s
is
ten
tly
d
eliv
er
in
g
ac
cu
r
ate
im
ag
e
q
u
ality
ass
es
s
m
en
ts
ac
r
o
s
s
d
iv
er
s
e
d
atasets
an
d
d
is
to
r
tio
n
ty
p
es.
T
ab
le
4
.
An
aly
zi
n
g
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
with
d
if
f
er
e
n
t d
ee
p
p
r
e
-
t
r
ain
ed
f
ea
tu
r
e
e
x
tr
ac
to
r
s
B
e
n
c
h
mar
k
d
a
t
a
b
a
se
F
e
a
t
u
r
e
e
x
t
r
a
c
t
o
r
D
i
st
o
r
t
i
o
n
GB
WN
JP
EG
JP
2
K
LI
V
E
V
G
G
1
6
0
.
9
6
3
0
.
9
8
8
0
.
9
7
8
0
.
9
7
6
D
a
r
k
N
e
t
1
9
0
.
9
6
9
0
.
9
8
0
0
.
9
8
3
0
.
9
7
1
P
r
o
p
o
se
d
(
I
n
c
e
p
t
i
o
n
R
e
sN
e
t
V
2
)
0
.
9
7
2
0
.
8
9
1
0
.
9
8
7
0
.
9
3
9
C
S
I
Q
V
G
G
1
6
0
.
9
3
1
0
.
7
5
0
0
.
8
8
7
0
.
8
4
6
D
a
r
k
N
e
t
1
9
0
.
9
2
8
0
.
8
2
1
0
.
9
3
9
0
.
9
0
8
P
r
o
p
o
se
d
(
I
n
c
e
p
t
i
o
n
R
e
sN
e
t
V
2
)
0
.
9
4
2
0
.
8
4
5
0
.
9
5
6
0
.
9
3
4
TI
D
2
0
1
3
V
G
G
1
6
0
.
9
2
5
0
.
9
3
1
0
.
9
6
3
0
.
9
2
8
D
a
r
k
N
e
t
1
9
0
.
9
1
8
0
.
9
4
4
0
.
9
6
0
0
.
8
4
4
P
r
o
p
o
se
d
(
I
n
c
e
p
t
i
o
n
R
e
sN
e
t
V
2
)
0
.
9
3
6
0
.
9
5
6
0
.
8
7
8
0
.
8
2
6
Seco
n
d
,
we
s
tu
d
y
th
e
im
p
ac
t
o
f
th
e
s
elec
tio
n
o
f
th
e
s
im
ilar
ity
m
ea
s
u
r
e
o
n
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
.
B
e
s
i
d
e
s
P
S
N
R
,
w
e
a
s
s
e
s
s
t
h
e
i
m
p
a
c
t
s
o
f
o
t
h
e
r
s
i
m
i
l
a
r
i
t
y
m
e
a
s
u
r
e
s
:
G
M
S
D
[
2
9
]
,
P
S
N
R
[
3
0
]
,
NC
C
[
3
0
]
,
FS
I
M
[
3
1
]
,
NAE
[
3
0
]
,
VSI
[
3
2
]
an
d
SS
I
M
[
3
3
]
.
T
ab
le
5
tab
u
lates
th
e
P
L
C
C
v
alu
es
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
d
if
f
er
e
n
t
s
im
ilar
ity
m
ea
s
u
r
es
with
f
o
u
r
co
m
m
o
n
d
is
to
r
tio
n
ty
p
es,
i.
e.
,
GB
,
W
N,
J
PEG
an
d
J
P2
K
d
is
to
r
tio
n
to
g
u
id
e
th
e
s
co
r
e
alig
n
m
e
n
t.
T
h
e
a
v
er
ag
e
o
f
th
e
p
r
ev
i
o
u
s
d
is
to
r
tio
n
f
o
r
ea
ch
d
ataset
is
ca
lcu
lated
.
I
t is n
o
ted
t
h
at
th
e
PS
NR
lead
s
to
th
e
b
est r
esu
lts
,
an
d
th
u
s
it is
u
s
ed
in
th
is
p
r
o
p
o
s
ed
m
eth
o
d
.
T
ab
le
5
.
Per
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
d
if
f
er
e
n
t
s
im
ilar
ity
m
ea
s
u
r
es
S
i
mi
l
a
r
i
t
y
M
e
a
su
r
e
P
LC
C
TI
D
2
0
1
3
C
S
I
Q
LI
V
E
A
v
e
r
a
g
e
P
S
N
R
[
3
0
]
0
.
8
9
9
0
0
.
9
1
9
3
0
.
9
3
2
3
0
.
9
1
6
8
G
M
S
D
[
2
9
]
0
.
7
4
1
7
0
.
8
3
7
8
0
.
7
5
3
8
0
.
7
9
1
1
N
C
C
[
3
0
]
0
.
9
1
9
0
0
.
8
8
5
4
0
.
9
4
2
5
0
.
9
1
5
6
F
S
I
M
[
3
1
]
0
.
9
7
5
5
0
.
8
8
5
5
0
.
8
9
3
5
5
0
.
9
1
8
1
N
A
E
[
3
0
]
0
.
8
4
5
6
0
.
9
0
5
4
0
.
9
3
0
1
0
.
8
9
3
7
V
S
I
[
3
2
]
0
.
7
6
5
4
0
.
7
1
0
6
0
.
8
4
5
4
0
.
7
7
3
8
S
S
I
M
[
3
3
]
0
.
9
4
7
8
0
.
8
4
5
8
0
.
8
2
7
5
0
.
8
7
3
7
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
:
1
5
2
1
-
1
5
3
1
1528
3
.
5
.
Co
m
pa
ring
t
he
pro
po
s
e
d m
et
ho
d wit
h t
ra
ditio
na
l a
n
d dee
p lea
rning
-
ba
s
e
d NR
-
I
Q
A
m
et
ho
ds
T
ab
le
6
p
r
o
v
i
d
es
a
co
m
p
r
eh
en
s
iv
e
co
m
p
ar
is
o
n
b
etwe
en
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
8
tr
ad
it
io
n
al
NR
-
I
QA
m
eth
o
d
s
,
n
am
ely
I
L
-
NI
QE
[
4
0
]
,
QAC
[
4
1
]
,
NI
QE
[
4
2
]
,
L
PS
I
[
4
3
]
,
DI
I
V
I
NE
[
4
4
]
,
B
L
I
I
NDS
-
II
[
4
5
]
,
B
R
I
SQ
UE
[
4
6
]
,
W
aDI
QaM
-
NR
[
1
1
]
,
as
well
a
s
T
S
PR
[
2
2
]
d
ee
p
lear
n
in
g
-
b
ased
NR
-
I
QA
m
eth
o
d
.
I
n
th
e
L
I
VE
d
ataset,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
c
o
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
s
ex
is
tin
g
m
eth
o
d
s
s
u
ch
a
s
QAC,
I
L
-
NI
QE
,
NI
QE
,
L
PS
I
,
DI
I
VI
NE
,
B
L
I
I
NDS
-
I
I
,
B
R
I
SQUE
,
W
aDI
Qa
M
-
NR
,
an
d
T
SP
R
,
ac
h
iev
in
g
a
r
em
ar
k
ab
le
PLCC
o
f
0
.
9
7
2
f
o
r
J
PEG
d
is
to
r
tio
n
.
Similar
ly
,
in
th
e
C
SIQ
d
atas
et,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ex
ce
ls
ag
ain
s
t
v
ar
io
u
s
m
eth
o
d
s
,
in
clu
d
i
n
g
I
L
-
N
I
QE
,
NI
QE
,
L
PS
I
,
DI
I
VI
NE
,
B
L
I
I
NDS
-
I
I
,
B
R
I
SQUE
,
W
aDI
Qa
M
-
NR
,
an
d
T
SP
R
,
with
a
PLCC
o
f
0
.
9
4
2
f
o
r
J
PEG
d
is
to
r
tio
n
.
I
n
th
e
T
I
D2
0
1
3
d
ataset,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
co
n
tin
u
es
to
d
em
o
n
s
tr
ate
s
tr
o
n
g
p
er
f
o
r
m
a
n
ce
,
s
u
r
p
ass
in
g
tr
ad
itio
n
al
m
eth
o
d
s
an
d
r
e
v
ea
lin
g
co
m
p
eti
tiv
e
r
esu
lts
ag
ain
s
t
d
ee
p
lear
n
in
g
-
b
ased
m
eth
o
d
s
.
T
h
ese
r
esu
lts
h
ig
h
lig
h
t
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
in
p
r
o
v
id
in
g
ac
cu
r
ate
an
d
r
eliab
le
n
o
-
r
ef
e
r
en
ce
im
ag
e
q
u
ality
ass
ess
m
en
ts
ac
r
o
s
s
a
d
iv
er
s
e
r
an
g
e
o
f
d
is
to
r
tio
n
s
a
n
d
d
atasets
,
p
o
s
itio
n
in
g
it f
av
o
r
ab
ly
am
o
n
g
ex
is
tin
g
s
tate
-
of
-
th
e
-
ar
t m
eth
o
d
s
.
T
a
b
l
e
6
.
C
o
m
p
a
r
i
s
o
n
b
e
t
w
e
e
n
t
h
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
a
n
d
t
r
a
d
i
t
i
o
n
a
l
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
N
R
-
I
Q
A
m
e
t
h
o
d
s
B
e
n
c
h
mar
k
d
a
t
a
b
a
se
M
e
t
h
o
d
D
i
st
o
r
t
i
o
n
GB
WN
JP
EG
JP
2
K
LI
V
E
Q
A
C
[
4
1
]
0
.
9
1
1
0
.
9
2
8
0
.
9
4
4
0
.
8
6
6
IL
-
N
I
Q
E
[
4
0
]
0
.
9
3
3
0
.
9
8
7
0
.
9
5
9
0
.
9
0
5
N
I
Q
E
[
4
2
]
0
.
9
4
5
0
.
9
7
6
0
.
9
5
2
0
.
9
2
6
LPS
I
[
4
3
]
0
.
9
1
5
0
.
9
6
5
0
.
9
7
5
0
.
9
3
6
D
I
I
V
I
N
E
[
4
4
]
0
.
9
2
1
0
.
9
8
4
0
.
9
1
0
0
.
9
1
3
B
LI
I
N
D
S
-
I
I
[
4
5
]
0
.
9
3
8
0
.
9
8
0
0
.
9
6
8
0
.
9
3
5
B
R
I
S
Q
U
E
[
4
6
]
0
.
9
5
1
0
.
9
8
5
0
.
9
7
3
0
.
9
2
3
W
a
D
I
Q
a
M
-
N
R
[
1
1
]
0
.
9
6
1
0
.
9
8
9
0
.
9
5
3
0
.
9
6
2
TSP
R
[
2
2
]
0
.
9
6
8
0
.
9
9
5
0
.
9
8
0
0
.
9
7
3
P
r
o
p
o
se
d
0
.
9
7
2
0
.
8
9
1
0
.
9
8
7
0
.
9
3
9
C
S
I
Q
Q
A
C
[
4
1
]
0
.
8
5
7
0
.
8
7
8
0
.
9
3
8
0
.
8
9
5
IL
-
N
I
Q
E
[
4
0
]
0
.
8
9
4
0
.
8
6
4
0
.
9
5
5
0
.
9
2
6
N
I
Q
E
[
4
2
]
0
.
9
2
5
0
.
8
1
1
0
.
9
3
5
0
.
9
2
6
LPS
I
[
4
3
]
0
.
9
3
0
0
.
6
8
7
0
.
9
6
9
0
.
9
1
8
D
I
I
V
I
N
E
[
4
4
]
0
.
8
9
9
0
.
8
8
8
0
.
8
2
4
0
.
8
9
6
B
LI
I
N
D
S
-
I
I
[
4
5
]
0
.
8
9
3
0
.
7
7
4
0
.
9
3
8
0
.
9
1
5
B
R
I
S
Q
U
E
[
4
6
]
0
.
9
2
8
0
.
9
3
8
0
.
9
4
6
0
.
8
9
7
W
a
D
I
Q
a
M
-
N
R
[
1
1
]
0
.
9
6
6
0
.
9
1
8
0
.
9
4
3
0
.
8
7
0
TSP
R
[
2
2
]
0
.
9
0
3
0
.
9
5
2
0
.
9
4
5
0
.
8
7
6
P
r
o
p
o
se
d
0
.
9
4
2
0
.
8
4
5
0
.
9
5
6
0
.
9
3
4
TI
D
2
0
1
3
Q
A
C
[
4
1
]
0
.
8
4
8
0
.
7
9
7
0
.
8
6
9
0
.
8
0
9
IL
-
N
I
Q
E
[
4
0
]
0
.
8
4
8
0
.
8
8
4
0
.
9
0
0
0
.
8
9
0
N
I
Q
E
[
4
2
]
0
.
8
1
9
0
.
8
2
7
0
.
8
9
3
0
.
9
0
7
LPS
I
[
4
3
]
0
.
8
3
6
0
.
7
7
5
0
.
9
5
4
0
.
9
1
6
D
I
I
V
I
N
E
[
4
4
]
0
.
8
4
8
0
.
8
6
0
0
.
6
6
4
0
.
9
0
1
B
LI
I
N
D
S
-
I
I
[
4
5
]
0
.
8
4
9
0
.
6
4
8
0
.
8
7
7
0
.
9
2
0
B
R
I
S
Q
U
E
[
4
6
]
0
.
8
4
8
0
.
8
5
1
0
.
9
0
0
0
.
9
1
8
W
a
D
I
Q
a
M
-
N
R
[
1
1
]
0
.
8
2
1
0
.
8
3
2
0
.
9
7
6
0
.
9
3
8
TSP
R
[
2
2
]
0
.
9
0
2
0
.
9
3
8
0
.
9
4
9
0
.
9
6
4
P
r
o
p
o
se
d
0
.
9
3
6
0
.
9
5
6
0
.
8
7
8
0
.
8
2
6
3
.
6
.
Co
m
pu
t
a
t
io
na
l c
o
m
plex
it
y
o
f
t
he
pro
po
s
ed
m
et
ho
d
T
ab
le
7
g
iv
es
in
f
o
r
m
atio
n
ab
o
u
t
th
e
co
m
p
u
tatio
n
al
co
m
p
lex
ity
o
f
th
e
s
u
g
g
ested
ap
p
r
o
ac
h
co
m
p
ar
ed
to
th
e
I
QA
m
o
d
els
th
at
ar
e
cu
r
r
en
tly
in
u
s
e.
T
h
e
r
ep
o
r
ted
a
v
er
ag
e
r
u
n
n
i
n
g
tim
es
(
i
n
s
ec
o
n
d
s
p
er
im
a
g
e)
a
r
e
b
ased
o
n
im
ag
es
r
an
d
o
m
ly
p
ick
ed
u
p
f
r
o
m
th
e
T
I
D2
0
1
3
d
at
aset
o
n
a
co
m
p
u
ter
with
a
2
.
4
GHz
I
n
tel
C
o
r
e
i5
-
2
4
3
0
M
C
PU
an
d
8
GB
R
AM
,
with
an
im
ag
e
r
eso
lu
tio
n
o
f
5
1
2
×3
8
4
.
T
h
e
r
esu
lts
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
m
eth
o
d
h
as
a
r
u
n
n
i
n
g
tim
e
o
f
5
.
1
s
ec
o
n
d
s
p
er
im
ag
e.
W
h
ile
n
o
t
th
e
f
astes
t
am
o
n
g
th
e
m
o
d
els
lis
ted
,
it
ac
h
iev
es
co
m
p
etitiv
e
r
esu
lts
,
esp
ec
ially
wh
en
co
m
p
a
r
ed
to
o
th
er
estab
lis
h
ed
I
QA
m
o
d
els.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
r
ep
o
r
t
h
e
r
e
th
e
r
u
n
tim
e
o
f
an
u
n
o
p
tim
ized
v
er
s
io
n
o
f
o
u
r
co
d
e.
Op
tim
izin
g
th
e
co
d
e
o
f
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
will si
g
n
if
ican
tly
r
ed
u
ce
th
e
r
u
n
tim
e.
T
ab
le
7
.
C
o
m
p
u
tatio
n
al
co
m
p
l
ex
ity
o
f
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
M
e
t
h
o
d
Ti
me
(
s)
N
I
Q
E
[
4
2
]
3
.
3
3
5
2
B
R
I
S
Q
U
E
[
4
6
]
1
.
3
5
8
D
I
I
V
I
N
E
[
4
4
]
1
5
.
1
2
0
P
r
o
p
o
se
d
5
.
1
2
3
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
No
-
r
eferen
ce
ima
g
e
q
u
a
lity a
s
s
es
s
men
t b
a
s
ed
o
n
visu
a
l
…
(
B
a
s
ma
A
h
med
)
1529
3
.
7
.
L
im
it
a
t
io
ns
o
f
t
he
pro
p
o
s
ed
m
et
ho
d
T
h
e
cu
r
r
e
n
t
s
tu
d
y
,
w
h
ile
s
h
o
wca
s
in
g
th
e
p
r
o
p
o
s
ed
m
eth
o
d
'
s
h
ig
h
p
er
f
o
r
m
an
ce
ac
r
o
s
s
v
ar
io
u
s
d
is
to
r
tio
n
ty
p
es
a
n
d
d
atasets
,
ac
k
n
o
wled
g
es
s
ev
er
al
lim
itatio
n
s
.
Firstl
y
,
th
e
s
tu
d
y
f
o
c
u
s
es
o
n
c
o
n
s
tr
u
ctin
g
a
s
in
g
le
r
esto
r
atio
n
m
o
d
el
f
o
r
P
R
I
b
y
s
ep
ar
ately
ad
d
r
ess
in
g
t
h
e
f
o
u
r
in
d
iv
id
u
al
d
is
to
r
tio
n
t
y
p
es.
T
h
is
ap
p
r
o
ac
h
m
ay
lim
it th
e
m
o
d
el'
s
ab
ilit
y
t
o
h
an
d
le
th
e
co
m
p
le
x
ities
o
f
m
ix
ed
m
u
lti
-
d
is
to
r
tio
n
s
.
Ad
d
r
ess
in
g
th
is
l
im
itatio
n
co
u
ld
en
h
an
ce
th
e
m
eth
o
d
'
s
v
er
s
atility
in
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
s
wh
er
e
im
ag
es
o
f
ten
s
u
f
f
er
f
r
o
m
m
u
ltip
le
d
is
to
r
tio
n
s
s
im
u
ltan
eo
u
s
ly
.
Ad
d
itio
n
ally
,
th
e
s
tu
d
y
p
o
i
n
ts
o
u
t
th
at
th
e
q
u
ality
o
f
d
ata
g
en
er
ated
b
y
th
e
PR
I
is
clo
s
ely
tied
to
th
e
ty
p
e
o
f
d
is
to
r
tio
n
.
T
h
is
lim
itatio
n
r
a
is
es
q
u
esti
o
n
s
ab
o
u
t
th
e
g
en
er
aliza
b
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ac
r
o
s
s
a
b
r
o
ad
er
r
a
n
g
e
o
f
d
is
to
r
tio
n
ty
p
es.
Fu
tu
r
e
r
esear
ch
co
u
l
d
ex
p
lo
r
e
ex
ten
d
in
g
th
e
m
eth
o
d
to
h
an
d
le
a
m
o
r
e
d
iv
er
s
e
s
et
o
f
d
is
to
r
tio
n
s
an
d
ass
ess
it
s
p
er
f
o
r
m
an
ce
in
s
ce
n
ar
i
o
s
in
v
o
lv
in
g
m
ix
e
d
d
is
to
r
tio
n
s
.
Fu
r
th
er
m
o
r
e,
t
h
e
s
tu
d
y
n
o
tes
th
at
th
e
r
ep
o
r
ted
r
u
n
tim
e
o
f
5
.
1
s
ec
o
n
d
s
is
b
ased
o
n
a
n
u
n
o
p
tim
ized
v
er
s
io
n
o
f
th
e
co
d
e.
Ack
n
o
wl
ed
g
in
g
th
is
,
th
e
au
th
o
r
s
s
u
g
g
e
s
t
th
at
o
p
tim
izatio
n
co
u
ld
s
ig
n
if
ican
tly
r
ed
u
ce
th
e
r
u
n
tim
e,
e
n
h
an
cin
g
th
e
co
m
p
u
tatio
n
al
ef
f
icien
cy
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
T
h
e
d
is
cu
s
s
io
n
em
p
h
asizes th
at
th
e
m
eth
o
d
,
d
esp
ite
th
e
cu
r
r
en
t
r
u
n
tim
e,
s
tr
ik
es
a
b
alan
ce
b
etwe
en
co
m
p
u
tatio
n
al
e
f
f
icien
cy
an
d
p
r
ec
is
io
n
,
m
ak
in
g
it
well
-
s
u
ited
f
o
r
ap
p
licatio
n
s
p
r
io
r
itizin
g
h
ig
h
er
a
cc
u
r
ac
y
in
im
a
g
e
q
u
ality
ass
e
s
s
m
en
t
o
v
er
f
aster
p
r
o
ce
s
s
in
g
tim
es.
4.
CO
NCLU
SI
O
N
AND
F
U
T
U
RE
WO
RK
T
h
is
p
ap
er
h
as
p
r
esen
ted
a
n
o
v
el
NR
-
I
QA
m
eth
o
d
b
ased
o
n
d
ee
p
v
is
u
al
e
x
p
lan
atio
n
s
.
Sp
ec
if
ically
,
th
e
m
ain
s
tep
s
o
f
o
u
r
m
eth
o
d
ar
e
g
en
er
atin
g
a
PR
I
f
r
o
m
th
e
in
p
u
t
d
eg
r
ad
e
d
im
ag
e,
ex
tr
a
ctin
g
d
ee
p
lear
n
in
g
f
ea
tu
r
es
th
r
o
u
g
h
d
ee
p
tr
an
s
f
e
r
lear
n
in
g
,
g
e
n
er
atin
g
VE
I
f
r
o
m
th
e
i
n
p
u
t
d
eg
r
a
d
ed
im
ag
e
an
d
its
PR
I
u
s
in
g
Gr
ad
-
C
AM
,
an
d
ca
lcu
latin
g
th
e
d
is
tan
ce
b
etwe
en
th
e
VE
I
o
f
th
e
d
is
to
r
ted
in
p
u
t im
ag
e
a
n
d
t
h
e
VE
I
o
f
PR
I
.
T
h
e
ef
f
icac
y
p
r
o
p
o
s
ed
m
eth
o
d
h
as
b
ee
n
ass
ess
ed
o
n
v
ar
io
u
s
b
en
ch
m
a
r
k
d
atasets
(
L
I
VE
,
C
SIQ
,
T
I
D2
0
1
3
)
a
n
d
d
is
to
r
tio
n
t
y
p
e
s
(
W
h
ite
No
is
e,
Gau
s
s
ian
b
lu
r
,
J
PEG,
an
d
J
PEG2
0
0
0
)
,
f
in
d
in
g
t
h
at
it
ac
h
ie
v
es
co
m
p
etitiv
e
p
er
f
o
r
m
a
n
ce
.
Dif
f
er
en
t
p
r
e
-
tr
ain
ed
C
NN
n
etwo
r
k
s
h
av
e
b
ee
n
em
p
lo
y
ed
t
o
ex
tr
ac
t
VE
I
s
lik
e
VGG1
6
,
I
n
ce
p
tio
n
R
esNetV2
an
d
Dar
k
n
et
1
9
,
f
in
d
in
g
th
at
I
n
ce
p
tio
n
R
esNetV2
f
ea
tu
r
e
ex
tr
ac
to
r
y
ield
ed
th
e
b
est
r
esu
lts
.
T
h
e
co
m
p
ar
is
o
n
s
with
s
ev
er
al
I
QA
m
o
d
els,
in
clu
d
in
g
QAC,
I
L
-
NI
QE
,
NI
Q
E
,
L
PS
I
,
DI
I
VI
NE
,
B
L
I
I
NDS
-
I
I
,
B
R
I
SQUE
,
W
a
DI
QaM
-
NR
,
an
d
T
SP
R
,
u
n
d
er
s
co
r
es th
e
m
eth
o
d
'
s
s
u
p
er
io
r
it
y
.
On
e
o
f
th
e
m
ain
im
p
licatio
n
s
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
t
h
at
it
en
h
an
ce
s
in
ter
p
r
etab
ilit
y
o
f
d
ee
p
lear
n
in
g
-
b
ased
NR
-
I
QA
b
y
cr
ea
tin
g
h
ea
t
m
ap
s
th
at
h
ig
h
lig
h
t
s
ig
n
if
ican
t
ar
ea
s
co
n
tr
ib
u
tin
g
to
th
e
q
u
ality
s
co
r
e
u
tili
zin
g
Gr
ad
-
C
AM
.
Al
s
o
,
o
u
r
m
eth
o
d
a
p
p
r
o
ac
h
allo
ws
f
o
r
a
m
o
r
e
in
f
o
r
m
ed
m
o
d
el
d
esig
n
an
d
en
s
u
r
es
a
m
o
r
e
ef
f
ec
tiv
e
an
d
tr
u
s
two
r
th
y
NR
-
I
QA
s
o
lu
tio
n
,
o
v
er
c
o
m
in
g
c
h
allen
g
es
o
f
lim
ited
in
ter
p
r
etab
ilit
y
an
d
ass
u
m
p
tio
n
s
p
r
ev
alen
t
in
ex
is
tin
g
m
eth
o
d
o
lo
g
ies.
Ad
d
itio
n
ally
,
o
u
r
m
eth
o
d
d
e
m
o
n
s
tr
ates
b
r
o
ad
e
r
ap
p
licab
ilit
y
ac
r
o
s
s
v
ar
i
o
u
s
d
is
to
r
tio
n
ty
p
es
an
d
d
atasets
,
p
r
o
v
id
i
n
g
a
co
m
p
r
eh
en
s
iv
e
s
t
u
d
y
to
v
alid
ate
its
ef
f
ec
tiv
en
ess
.
T
h
e
p
r
o
p
o
s
ed
I
QA
m
eth
o
d
h
as
p
r
ac
tical
ap
p
licatio
n
s
in
v
ar
io
u
s
f
ield
s
s
u
ch
as:
d
ig
ital
im
ag
in
g
an
d
Ph
o
to
g
r
ap
h
y
,
im
p
r
o
v
in
g
th
e
q
u
ality
co
n
tr
o
l
o
f
m
ed
i
ca
l
im
ag
in
g
,
an
d
o
p
tim
izin
g
im
ag
e
co
m
p
r
ess
io
n
alg
o
r
ith
m
s
to
m
ain
tain
h
ig
h
v
i
s
u
al
q
u
ality
.
Fu
tu
r
e
r
esear
ch
d
ir
ec
tio
n
s
in
v
o
lv
e
ex
ten
d
i
n
g
th
e
p
r
o
p
o
s
ed
m
eth
o
d
to
h
an
d
le
a
b
r
o
a
d
er
r
an
g
e
o
f
d
is
to
r
tio
n
ty
p
es
an
d
o
p
tim
izi
n
g
th
e
co
m
p
u
tatio
n
al
e
f
f
icien
cy
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
T
h
is
e
x
p
an
s
io
n
wo
u
ld
co
n
t
r
ib
u
te
to
a
m
o
r
e
c
o
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
'
s
ap
p
licab
ilit
y
in
r
ea
l
-
wo
r
ld
,
d
y
n
am
ic
s
ettin
g
s
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
d
ee
p
lear
n
i
n
g
h
av
e
s
ig
n
if
ica
n
tly
im
p
r
o
v
e
d
im
ag
e
q
u
ality
ass
ess
m
en
t
m
eth
o
d
s
.
Ou
r
f
in
d
i
n
g
s
p
r
o
v
id
e
co
n
cl
u
s
iv
e
ev
id
e
n
c
e
th
at
th
e
p
r
o
p
o
s
ed
NR
-
I
QA
m
eth
o
d
,
b
ased
o
n
d
ee
p
v
is
u
al
ex
p
lan
atio
n
s
,
ef
f
ec
tiv
ely
en
h
an
ce
s
in
ter
p
r
etab
ilit
y
an
d
ac
cu
r
ac
y
ac
r
o
s
s
v
ar
io
u
s
d
is
to
r
tio
n
t
y
p
es a
n
d
d
atasets
,
o
u
tp
e
r
f
o
r
m
in
g
tr
ad
itio
n
al
m
o
d
els.
RE
F
E
R
E
NC
E
S
[
1
]
M
.
Li
u
,
J
.
H
u
a
n
g
,
D
.
Z
e
n
g
,
X
.
D
i
n
g
,
a
n
d
J
.
P
a
i
sl
e
y
,
“
A
mu
l
t
i
sca
l
e
a
p
p
r
o
a
c
h
t
o
d
e
e
p
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
sses
sme
n
t
,
”
I
E
E
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
3
2
,
p
p
.
1
6
5
6
–
1
6
6
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
2
3
.
3
2
4
5
9
9
1
.
[
2
]
S
.
B
i
a
n
c
o
,
L.
C
e
l
o
n
a
,
P
.
N
a
p
o
l
e
t
a
n
o
,
a
n
d
R
.
S
c
h
e
t
t
i
n
i
,
“
O
n
t
h
e
u
s
e
o
f
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
b
l
i
n
d
i
m
a
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
S
i
g
n
a
l
,
I
m
a
g
e
a
n
d
V
i
d
e
o
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
2
,
n
o
.
2
,
p
p
.
3
5
5
–
3
6
2
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
7
6
0
-
0
1
7
-
1
1
6
6
-
8.
[
3
]
W.
-
H
.
K
i
m
,
C
.
H
a
h
m,
A
.
B
a
i
j
a
l
,
N
.
I
m,
I
.
C
h
o
,
a
n
d
J.
K
o
o
,
“
P
i
x
e
l
-
by
-
p
i
x
e
l
m
e
a
n
o
p
i
n
i
o
n
sc
o
r
e
(
p
M
O
S
)
f
o
r
no
-
r
e
f
e
r
e
n
c
e
i
m
a
g
e
q
u
a
l
i
t
y
a
sse
ssm
e
n
t
,
”
a
rXi
v
p
re
p
ri
n
t
a
r
Xi
v
:
2
2
0
6
.
0
6
5
4
1
,
2
0
2
2
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
2
2
0
6
.
0
6
5
4
1
.
[
4
]
S
.
A
t
h
a
r
a
n
d
Z.
W
a
n
g
,
“
D
e
g
r
a
d
e
d
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
3
2
,
p
p
.
8
2
2
–
8
3
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
2
3
.
3
2
3
4
4
9
8
.
[
5
]
C
.
L
i
,
W
.
Y
u
a
n
,
A
.
C
.
B
o
v
i
k
,
a
n
d
X
.
W
u
,
“
N
o
-
r
e
f
e
r
e
n
c
e
b
l
u
r
i
n
d
e
x
u
s
i
n
g
b
l
u
r
c
o
m
p
a
r
i
so
n
s,
”
E
l
e
c
t
r
o
n
i
c
s
L
e
t
t
e
rs
,
v
o
l
.
4
7
,
n
o
.
1
7
,
p
.
9
6
2
,
2
0
1
1
,
d
o
i
:
1
0
.
1
0
4
9
/
e
l
.
2
0
1
1
.
0
9
2
1
.
[
6
]
X
.
M
i
n
,
K
.
G
u
,
G
.
Zh
a
i
,
J.
L
i
u
,
X
.
Y
a
n
g
,
a
n
d
C
.
W
.
C
h
e
n
,
“
B
l
i
n
d
q
u
a
l
i
t
y
a
ssessm
e
n
t
b
a
se
d
o
n
p
se
u
d
o
-
r
e
f
e
r
e
n
c
e
i
m
a
g
e
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
M
u
l
t
i
m
e
d
i
a
,
v
o
l
.
2
0
,
n
o
.
8
,
p
p
.
2
0
4
9
–
2
0
6
2
,
A
u
g
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
T
M
M
.
2
0
1
7
.
2
7
8
8
2
0
6
.
[
7
]
Y
.
Zh
a
n
a
n
d
R
.
Zh
a
n
g
,
“
N
o
-
r
e
f
e
r
e
n
c
e
JP
EG
i
m
a
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
b
a
s
e
d
o
n
b
l
o
c
k
i
n
e
ss
a
n
d
l
u
mi
n
a
n
c
e
c
h
a
n
g
e
,
”
I
EE
E
S
i
g
n
a
l
Pro
c
e
ssi
n
g
L
e
t
t
e
rs
,
v
o
l
.
2
4
,
n
o
.
6
,
p
p
.
7
6
0
–
7
6
4
,
J
u
n
.
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
LS
P
.
2
0
1
7
.
2
6
8
8
3
7
1
.
[
8
]
M
.
O
s
z
u
st
,
“
N
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
w
i
t
h
l
o
c
a
l
g
r
a
d
i
e
n
t
o
r
i
e
n
t
a
t
i
o
n
s
,
”
S
y
m
m
e
t
r
y
,
v
o
l
.
1
1
,
n
o
.
1
,
p
.
9
5
,
Ja
n
.
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
s
y
m
1
1
0
1
0
0
9
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
3
,
Dec
em
b
er
20
24
:
1
5
2
1
-
1
5
3
1
1530
[
9
]
Z.
L
i
a
n
g
,
W
.
L
u
,
Y
.
Z
h
e
n
g
,
W
.
H
e
,
a
n
d
J
.
Y
a
n
g
,
“
T
h
e
c
o
n
t
e
x
t
e
f
f
e
c
t
f
o
r
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
sse
ssm
e
n
t
,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
5
2
1
,
p
p
.
1
7
2
–
1
8
0
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.
2
0
2
2
.
1
1
.
0
2
6
.
[
1
0
]
C
.
F
a
n
,
Y
.
Zh
a
n
g
,
L.
F
e
n
g
,
a
n
d
Q
.
J
i
a
n
g
,
“
N
o
r
e
f
e
r
e
n
c
e
i
m
a
g
e
q
u
a
l
i
t
y
a
s
sessme
n
t
b
a
se
d
o
n
m
u
l
t
i
-
e
x
p
e
r
t
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
6
,
p
p
.
8
9
3
4
–
8
9
4
3
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
E
S
S
.
2
0
1
8
.
2
8
0
2
4
9
8
.
[
1
1
]
S
.
B
o
s
s
e
,
D
.
M
a
n
i
r
y
,
K
.
-
R
.
M
u
l
l
e
r
,
T
.
W
i
e
g
a
n
d
,
a
n
d
W
.
S
a
m
e
k
,
“
D
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
n
o
-
r
e
f
e
r
e
n
c
e
a
n
d
f
u
l
l
-
r
e
f
e
r
e
n
c
e
i
m
a
g
e
q
u
a
l
i
t
y
a
s
s
e
s
s
m
e
n
t
,
”
I
E
E
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
P
r
o
c
e
s
s
i
n
g
,
v
o
l
.
2
7
,
n
o
.
1
,
p
p
.
2
0
6
–
2
1
9
,
J
a
n
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
T
I
P
.
2
0
1
7
.
2
7
6
0
5
1
8
.
[
1
2
]
Q
.
Y
a
n
,
D
.
G
o
n
g
,
a
n
d
Y
.
Zh
a
n
g
,
“
Tw
o
-
st
r
e
a
m c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s f
o
r
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
2
8
,
n
o
.
5
,
p
p
.
2
2
0
0
–
2
2
1
1
,
M
a
y
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
8
.
2
8
8
3
7
4
1
.
[
1
3
]
R
.
S
.
R
a
v
e
l
a
,
M
.
V
.
S
h
i
r
v
a
i
k
a
r
,
a
n
d
C
.
G
r
e
c
o
s,
“
N
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
b
a
s
e
d
o
n
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
i
n
R
e
a
l
-
T
i
m
e
I
m
a
g
e
Pro
c
e
ssi
n
g
a
n
d
D
e
e
p
L
e
a
r
n
i
n
g
2
0
1
9
,
M
a
y
2
0
1
9
,
p
.
1
0
9
9
6
0
4
,
d
o
i
:
1
0
.
1
1
1
7
/
1
2
.
2
5
1
8
4
3
8
.
[
1
4
]
Z.
P
a
n
e
t
a
l
.
,
“
D
A
C
N
N
:
b
l
i
n
d
i
m
a
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
v
i
a
a
d
i
st
o
r
t
i
o
n
-
a
w
a
r
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
f
o
r
Vi
d
e
o
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
3
2
,
n
o
.
1
1
,
p
p
.
7
5
1
8
–
7
5
3
1
,
N
o
v
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TC
S
V
T
.
2
0
2
2
.
3
1
8
8
9
9
1
.
[
1
5
]
M
.
Z
h
o
u
e
t
a
l
.
,
“
A
t
t
e
n
t
i
o
n
a
l
f
e
a
t
u
r
e
f
u
si
o
n
f
o
r
e
n
d
-
to
-
e
n
d
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
s
sme
n
t
,
”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
Br
o
a
d
c
a
st
i
n
g
,
v
o
l
.
6
9
,
n
o
.
1
,
p
p
.
1
4
4
–
1
5
2
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
B
C
.
2
0
2
2
.
3
2
0
4
2
3
5
.
[
1
6
]
K
.
Y
.
L
i
n
a
n
d
G
.
W
a
n
g
,
“
H
a
l
l
u
c
i
n
a
t
e
d
-
I
Q
A
:
no
-
r
e
f
e
r
e
n
c
e
i
m
a
g
e
q
u
a
l
i
t
y
a
s
s
e
s
s
m
e
n
t
v
i
a
a
d
v
e
r
s
a
r
i
a
l
l
e
a
r
n
i
n
g
,
”
i
n
P
r
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
E
E
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
2
0
1
8
,
p
p
.
7
3
2
–
7
4
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
0
8
3
.
[
1
7
]
F
.
G
a
o
,
J.
Y
u
,
S
.
Zh
u
,
Q
.
H
u
a
n
g
,
a
n
d
Q
.
Ti
a
n
,
“
B
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
p
r
e
d
i
c
t
i
o
n
b
y
e
x
p
l
o
i
t
i
n
g
m
u
l
t
i
-
l
e
v
e
l
d
e
e
p
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
,
”
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
8
1
,
p
p
.
4
3
2
–
4
4
2
,
S
e
p
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
1
8
.
0
4
.
0
1
6
.
[
1
8
]
D
.
V
a
r
g
a
,
“
M
u
l
t
i
-
p
o
o
l
e
d
i
n
c
e
p
t
i
o
n
f
e
a
t
u
r
e
s
f
o
r
n
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
s
sessm
e
n
t
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
0
,
n
o
.
6
,
p
.
2
1
8
6
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
0
0
6
2
1
8
6
.
[
1
9
]
S
.
W
a
n
g
a
n
d
Y
.
Z
h
a
n
g
,
“
G
r
a
d
-
C
A
M
:
u
n
d
e
r
st
a
n
d
i
n
g
A
I
m
o
d
e
l
s
,
”
C
o
m
p
u
t
e
rs
,
M
a
t
e
r
i
a
l
s
&
C
o
n
t
i
n
u
a
,
v
o
l
.
7
6
,
n
o
.
2
,
p
p
.
1
3
2
1
–
1
3
2
4
,
2
0
2
3
,
d
o
i
:
1
0
.
3
2
6
0
4
/
c
mc.
2
0
2
3
.
0
4
1
4
1
9
.
[
2
0
]
B
.
A
h
m
e
d
,
O
.
A
.
O
mer,
A
.
R
a
s
h
e
d
,
D
.
P
u
i
g
,
a
n
d
M
.
A
b
d
e
l
-
N
a
ss
e
r
,
“
R
e
f
e
r
e
n
c
e
l
e
ss
i
m
a
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
u
t
i
l
i
z
i
n
g
d
e
e
p
t
r
a
n
sf
e
r
-
l
e
a
r
n
e
d
f
e
a
t
u
r
e
s,
”
Fr
o
n
t
i
e
rs
i
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
5
6
,
p
p
.
2
4
3
–
2
4
8
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
3
2
3
3
/
F
A
I
A
2
2
0
3
4
5
.
[
2
1
]
X
.
M
i
n
,
G
.
Zh
a
i
,
K
.
G
u
,
Y
.
L
i
u
,
a
n
d
X
.
Y
a
n
g
,
“
B
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
e
st
i
ma
t
i
o
n
v
i
a
d
i
s
t
o
r
t
i
o
n
a
g
g
r
a
v
a
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
Bro
a
d
c
a
s
t
i
n
g
,
v
o
l
.
6
4
,
n
o
.
2
,
p
p
.
5
0
8
–
5
1
7
,
Ju
n
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
TB
C
.
2
0
1
8
.
2
8
1
6
7
8
3
.
[
2
2
]
J.
H
u
,
X
.
W
a
n
g
,
F
.
S
h
a
o
,
a
n
d
Q
.
Ji
a
n
g
,
“
TSP
R
:
d
e
e
p
n
e
t
w
o
r
k
-
b
a
s
e
d
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
u
s
i
n
g
t
w
o
-
si
d
e
p
s
e
u
d
o
r
e
f
e
r
e
n
c
e
i
ma
g
e
s,”
D
i
g
i
t
a
l
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
0
6
,
p
.
1
0
2
8
4
9
,
N
o
v
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
d
s
p
.
2
0
2
0
.
1
0
2
8
4
9
.
[
2
3
]
A
.
R
e
h
m
a
n
,
S
.
N
a
z
,
M
.
I
.
R
a
z
z
a
k
,
F
.
A
k
r
a
m
,
a
n
d
M
.
I
m
r
a
n
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
f
r
a
m
e
w
o
r
k
f
o
r
a
u
t
o
m
a
t
i
c
b
r
a
i
n
t
u
m
o
r
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
u
s
i
n
g
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
,
”
C
i
r
c
u
i
t
s
,
S
y
s
t
e
m
s
,
a
n
d
S
i
g
n
a
l
P
r
o
c
e
s
s
i
n
g
,
v
o
l
.
3
9
,
n
o
.
2
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
0
3
4
-
0
1
9
-
0
1
2
4
6
-
3.
[
2
4
]
F
.
C
h
e
n
,
H
.
F
u
,
H
.
Y
u
,
a
n
d
Y
.
C
h
u
,
“
N
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
b
a
se
d
o
n
a
m
u
l
t
i
t
a
s
k
i
m
a
g
e
r
e
s
t
o
r
a
t
i
o
n
n
e
t
w
o
r
k
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
3
,
n
o
.
1
1
,
p
.
6
8
0
2
,
Ju
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
3
1
1
6
8
0
2
.
[
2
5
]
Y
.
P
e
sk
i
n
e
,
M
.
-
C
.
B
o
u
c
h
e
r
,
a
n
d
F
.
C
h
e
r
i
e
t
,
“
A
n
i
n
t
e
r
p
r
e
t
a
b
l
e
d
a
t
a
-
d
r
i
v
e
n
sco
r
e
f
o
r
t
h
e
a
ss
e
ssm
e
n
t
o
f
f
u
n
d
u
s
i
mag
e
s
q
u
a
l
i
t
y
,
”
i
n
I
m
a
g
e
A
n
a
l
y
s
i
s
a
n
d
Re
c
o
g
n
i
t
i
o
n
:
1
7
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
,
I
C
I
AR
2
0
2
0
,
P
ó
v
o
a
d
e
Va
rz
i
m
,
P
o
rt
u
g
a
l
,
2
0
2
0
,
p
p
.
3
2
5
–
3
3
1
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
030
-
5
0
5
1
6
-
5
_
2
8
.
[
2
6
]
S
.
S
a
r
k
a
r
e
t
a
l
.
,
“
R
L
-
C
A
M
:
v
i
s
u
a
l
e
x
p
l
a
n
a
t
i
o
n
s
f
o
r
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
u
si
n
g
r
e
i
n
f
o
r
c
e
m
e
n
t
l
e
a
r
n
i
n
g
,
”
i
n
I
EEE
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
Pa
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
W
o
r
k
sh
o
p
s
,
2
0
2
3
,
v
o
l
.
2
0
2
3
-
J
u
n
e
,
p
p
.
3
8
6
1
–
3
8
6
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
W
5
9
2
2
8
.
2
0
2
3
.
0
0
4
0
0
.
[
2
7
]
J.
S
h
e
n
g
,
J.
F
a
n
,
P
.
Y
e
,
a
n
d
J
.
C
a
o
,
“
JN
D
M
i
x
:
J
n
d
-
b
a
s
e
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
f
o
r
n
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
i
n
I
C
AS
S
P
2
0
2
3
-
2
0
2
3
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ac
o
u
st
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
(
I
C
A
S
S
P)
,
Ju
n
.
2
0
2
3
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
4
9
3
5
7
.
2
0
2
3
.
1
0
0
9
6
2
3
4
.
[
2
8
]
R
.
R
.
S
e
l
v
a
r
a
j
u
,
M
.
C
o
g
sw
e
l
l
,
A
.
D
a
s,
R
.
V
e
d
a
n
t
a
m,
D
.
P
a
r
i
k
h
,
a
n
d
D
.
B
a
t
r
a
,
“
G
r
a
d
-
C
A
M
:
v
i
s
u
a
l
e
x
p
l
a
n
a
t
i
o
n
s
f
r
o
m
d
e
e
p
n
e
t
w
o
r
k
s
v
i
a
g
r
a
d
i
e
n
t
-
b
a
se
d
l
o
c
a
l
i
z
a
t
i
o
n
,
”
i
n
2
0
1
7
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
(
I
C
C
V)
,
O
c
t
.
2
0
1
7
,
p
p
.
6
1
8
–
6
2
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
1
7
.
7
4
.
[
2
9
]
W
.
X
u
e
,
L.
Z
h
a
n
g
,
X
.
M
o
u
,
a
n
d
A
.
C
.
B
o
v
i
k
,
“
G
r
a
d
i
e
n
t
ma
g
n
i
t
u
d
e
si
m
i
l
a
r
i
t
y
d
e
v
i
a
t
i
o
n
:
a
h
i
g
h
l
y
e
f
f
i
c
i
e
n
t
p
e
r
c
e
p
t
u
a
l
i
ma
g
e
q
u
a
l
i
t
y
i
n
d
e
x
,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
2
3
,
n
o
.
2
,
p
p
.
6
8
4
–
6
9
5
,
F
e
b
.
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
3
.
2
2
9
3
4
2
3
.
[
3
0
]
W
.
A
.
M
u
st
a
f
a
,
H
.
Y
a
z
i
d
,
A
.
S
a
l
i
h
a
h
,
A
.
N
a
s
i
r
,
a
n
d
M
.
V
i
e
w
,
“
A
r
e
v
i
e
w
o
f
i
m
a
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
(
I
Q
A
)
:
S
N
R
,
G
C
F
,
A
D
,
N
A
E,
P
S
N
R
,
M
E,
”
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
Re
se
a
rc
h
i
n
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
1
–
7
,
2
0
1
7
.
[
3
1
]
L.
Z
h
a
n
g
,
L
.
Z
h
a
n
g
,
X
.
M
o
u
,
a
n
d
D
.
Zh
a
n
g
,
“
F
S
I
M
:
a
f
e
a
t
u
r
e
s
i
m
i
l
a
r
i
t
y
i
n
d
e
x
f
o
r
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
2
0
,
n
o
.
8
,
p
p
.
2
3
7
8
–
2
3
8
6
,
A
u
g
.
2
0
1
1
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
1
.
2
1
0
9
7
3
0
.
[
3
2
]
L.
Z
h
a
n
g
,
Y
.
S
h
e
n
,
a
n
d
H
.
L
i
,
“
V
S
I
:
a
v
i
s
u
a
l
sa
l
i
e
n
c
y
-
i
n
d
u
c
e
d
i
n
d
e
x
f
o
r
p
e
r
c
e
p
t
u
a
l
i
m
a
g
e
q
u
a
l
i
t
y
a
sse
ssm
e
n
t
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
2
3
,
n
o
.
1
0
,
p
p
.
4
2
7
0
–
4
2
8
1
,
O
c
t
.
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
4
.
2
3
4
6
0
2
8
.
[
3
3
]
J.
P
e
n
g
e
t
a
l
.
,
“
I
mp
l
e
m
e
n
t
a
t
i
o
n
o
f
t
h
e
s
t
r
u
c
t
u
r
a
l
S
I
M
i
l
a
r
i
t
y
(
S
S
I
M
)
i
n
d
e
x
a
s
a
q
u
a
n
t
i
t
a
t
i
v
e
e
v
a
l
u
a
t
i
o
n
t
o
o
l
f
o
r
d
o
s
e
d
i
st
r
i
b
u
t
i
o
n
e
r
r
o
r
d
e
t
e
c
t
i
o
n
,
”
Me
d
i
c
a
l
Ph
y
si
c
s
,
v
o
l
.
4
7
,
n
o
.
4
,
p
p
.
1
9
0
7
–
1
9
1
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
2
/
m
p
.
1
4
0
1
0
.
[
3
4
]
D
.
V
a
r
g
a
,
“
N
o
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
sses
sme
n
t
u
s
i
n
g
t
h
e
s
t
a
t
i
s
t
i
c
s
o
f
g
l
o
b
a
l
a
n
d
l
o
c
a
l
i
m
a
g
e
f
e
a
t
u
r
e
s,
”
E
l
e
c
t
r
o
n
i
c
s
,
v
o
l
.
1
2
,
n
o
.
7
,
p
.
1
6
1
5
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
2
0
7
1
6
1
5
.
[
3
5
]
H
.
R
.
S
h
e
i
k
h
,
“
LI
V
E
i
ma
g
e
q
u
a
l
i
t
y
a
s
sessme
n
t
d
a
t
a
b
a
se
r
e
l
e
a
se
2
,
”
2
0
0
5
.
[
O
n
l
i
n
e
]
h
t
t
p
:
/
/
l
i
v
e
.
e
c
e
.
u
t
e
x
a
s.
e
d
u
/
r
e
se
a
r
c
h
/
q
u
a
l
i
t
y
.
[
3
6
]
E.
C
.
La
r
s
o
n
a
n
d
D
.
M
.
C
h
a
n
d
l
e
r
,
“
M
o
s
t
a
p
p
a
r
e
n
t
d
i
st
o
r
t
i
o
n
:
f
u
l
l
-
r
e
f
e
r
e
n
c
e
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
a
n
d
t
h
e
r
o
l
e
o
f
s
t
r
a
t
e
g
y
,
”
J
o
u
rn
a
l
o
f
e
l
e
c
t
ro
n
i
c
i
m
a
g
i
n
g
,
v
o
l
.
1
9
,
n
o
.
1
,
p
p
.
0
1
1
0
0
6
--
0
1
1
0
0
6
,
2
0
1
0
.
[
3
7
]
H
.
Y
a
n
g
,
Y
.
F
a
n
g
,
a
n
d
W
.
Li
n
,
“
P
e
r
c
e
p
t
u
a
l
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
o
f
s
c
r
e
e
n
c
o
n
t
e
n
t
i
ma
g
e
s,”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ssi
n
g
,
v
o
l
.
2
4
,
n
o
.
1
1
,
p
p
.
4
4
0
8
–
4
4
2
1
,
N
o
v
.
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
5
.
2
4
6
5
1
4
5
.
[
3
8
]
N
.
P
o
n
o
m
a
r
e
n
k
o
e
t
a
l
.
,
“
I
mag
e
d
a
t
a
b
a
s
e
TI
D
2
0
1
3
:
p
e
c
u
l
i
a
r
i
t
i
e
s
,
r
e
s
u
l
t
s
a
n
d
p
e
r
sp
e
c
t
i
v
e
s,
”
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
:
I
m
a
g
e
C
o
m
m
u
n
i
c
a
t
i
o
n
,
v
o
l
.
3
0
,
p
p
.
5
7
–
7
7
,
J
a
n
.
2
0
1
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
ma
g
e
.
2
0
1
4
.
1
0
.
0
0
9
.
[
3
9
]
Ji
a
n
b
o
S
h
i
a
n
d
To
mas
i
,
“
G
o
o
d
f
e
a
t
u
r
e
s
t
o
t
r
a
c
k
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
I
EEE
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
C
VP
R
-
94
,
1
9
9
4
,
p
p
.
5
9
3
–
6
0
0
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
1
9
9
4
.
3
2
3
7
9
4
.
[
4
0
]
L.
Zh
a
n
g
,
L.
Z
h
a
n
g
,
a
n
d
A
.
C
.
B
o
v
i
k
,
“
A
f
e
a
t
u
r
e
-
e
n
r
i
c
h
e
d
c
o
mp
l
e
t
e
l
y
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
e
v
a
l
u
a
t
o
r
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
2
4
,
n
o
.
8
,
p
p
.
2
5
7
9
–
2
5
9
1
,
A
u
g
.
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
5
.
2
4
2
6
4
1
6
.
[
4
1
]
W
.
X
u
e
,
L.
Zh
a
n
g
,
a
n
d
X
.
M
o
u
,
“
Le
a
r
n
i
n
g
w
i
t
h
o
u
t
h
u
m
a
n
s
c
o
r
e
s fo
r
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
ssessm
e
n
t
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s o
f
t
h
e
I
E
E
E
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
2
0
1
3
,
p
p
.
9
9
5
–
1
0
0
2
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
3
.
1
3
3
.
[
4
2
]
A
.
M
i
t
t
a
l
,
R
.
S
o
u
n
d
a
r
a
r
a
j
a
n
,
a
n
d
A
.
C
.
B
o
v
i
k
,
“
M
a
k
i
n
g
a
‘
c
o
mp
l
e
t
e
l
y
b
l
i
n
d
’
i
m
a
g
e
q
u
a
l
i
t
y
a
n
a
l
y
z
e
r
,
”
I
EE
E
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
L
e
t
t
e
rs
,
v
o
l
.
2
0
,
n
o
.
3
,
p
p
.
2
0
9
–
2
1
2
,
M
a
r
.
2
0
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
LSP
.
2
0
1
2
.
2
2
2
7
7
2
6
.
[
4
3
]
Q
.
W
u
,
Z.
W
a
n
g
,
a
n
d
H
.
L
i
,
“
A
h
i
g
h
l
y
e
f
f
i
c
i
e
n
t
met
h
o
d
f
o
r
b
l
i
n
d
i
ma
g
e
q
u
a
l
i
t
y
a
ss
e
ssm
e
n
t
,
”
i
n
2
0
1
5
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
(
I
C
I
P)
,
S
e
p
.
2
0
1
5
,
p
p
.
3
3
9
–
3
4
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
5
.
7
3
5
0
8
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.