I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
8
,
No
.
1
,
A
p
r
il
20
2
5
,
p
p
.
243
~
253
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
8
.
i
1
.
pp
243
-
2
5
3
243
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
A
multi
-
sca
le con
v
o
lutiona
l neural
network a
nd
disc
rete
wa
v
elet
trans
form ba
sed
retinal ima
g
e com
press
io
n
Da
lila
Chik
ha
o
ui
1
,
M
o
ha
m
m
ed
B
ela
dg
ha
m
1
,
M
o
ha
m
ed
B
ena
is
s
a
2
,
Ab
delm
a
lik
T
a
leb
-
Ahm
ed
3
1
I
n
f
o
r
mat
i
o
n
P
r
o
c
e
ss
i
n
g
a
n
d
Te
l
e
c
o
m
mu
n
i
c
a
t
i
o
n
La
b
o
r
a
t
o
r
y
(
LTI
T)
,
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
T
e
c
h
n
o
l
o
g
y
-
U
n
i
v
e
r
si
t
y
TA
H
R
I
M
o
h
a
mm
e
d
B
e
c
h
a
r
,
B
e
c
h
a
r
,
A
l
g
e
r
i
a
2
I
n
f
o
r
mat
i
o
n
P
r
o
c
e
ss
i
n
g
a
n
d
Te
l
e
c
o
m
mu
n
i
c
a
t
i
o
n
La
b
o
r
a
t
o
r
y
(
LTI
T)
,
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
T
e
c
h
n
o
l
o
g
y
-
U
n
i
v
e
r
si
t
y
A
b
o
u
B
e
k
r
B
e
l
k
a
i
d
Tl
e
mce
n
,
T
l
e
m
c
e
n
,
A
l
g
e
r
i
a
3
P
o
l
y
t
e
c
h
n
i
c
U
n
i
v
e
r
si
t
y
o
f
H
a
u
t
s
-
de
-
F
r
a
n
c
e
,
Li
l
l
e
U
n
i
v
e
r
si
t
y
,
C
N
R
S
,
V
a
l
e
n
c
i
e
n
n
e
s,
F
r
a
n
c
e
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
May
19
,
2
0
2
4
R
ev
is
ed
Oct
15
,
2
0
2
4
Acc
ep
ted
Oct
30
,
2
0
2
4
Th
e
d
iffere
n
t
a
p
p
li
c
a
ti
o
n
s
o
f
m
e
d
ica
l
ima
g
e
s
h
a
v
e
c
o
n
tri
b
u
ted
si
g
n
ifi
c
a
n
t
l
y
to
th
e
g
ro
wi
n
g
a
m
o
u
n
t
o
f
ima
g
e
d
a
ta.
As
a
re
su
lt
,
c
o
m
p
re
ss
io
n
t
e
c
h
n
iq
u
e
s
b
e
c
o
m
e
e
ss
e
n
ti
a
l
to
a
l
lo
w
re
a
l
-
ti
m
e
tran
sm
issio
n
a
n
d
st
o
ra
g
e
wit
h
in
li
m
it
e
d
n
e
two
rk
b
a
n
d
wid
t
h
a
n
d
sto
ra
g
e
sp
a
c
e
.
De
e
p
lea
rn
in
g
,
p
a
rti
c
u
larly
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s
(C
NN
)
h
a
v
e
m
a
rk
e
d
ra
p
id
a
d
v
a
n
c
e
s
in
m
a
n
y
c
o
m
p
u
ter
v
isio
n
tas
k
s
a
n
d
h
a
v
e
p
ro
g
re
ss
iv
e
ly
d
ra
w
n
a
tt
e
n
t
io
n
f
o
r
b
e
in
g
u
se
d
in
ima
g
e
c
o
m
p
re
ss
io
n
.
T
h
e
re
fo
r
e
,
we
p
re
se
n
t
a
m
e
th
o
d
fo
r
c
o
m
p
re
ss
in
g
re
ti
n
a
l
ima
g
e
s
b
a
se
d
o
n
d
e
e
p
C
NN
a
n
d
d
isc
re
te
wa
v
e
let
tran
sfo
rm
(DWT
).
To
f
u
rth
e
r
e
n
h
a
n
c
e
CNN
c
a
p
a
b
il
it
ies
,
m
u
lt
i
-
sc
a
le
c
o
n
v
o
lu
t
io
n
s
a
r
e
in
tro
d
u
c
e
d
i
n
t
o
th
e
n
e
tw
o
rk
a
rc
h
it
e
c
tu
re
.
In
th
is
p
ro
p
o
se
d
m
e
th
o
d
,
m
u
lt
isc
a
le
CNN
s
a
re
u
se
d
to
e
x
trac
t
u
se
fu
l
f
e
a
tu
re
s
to
p
r
o
v
i
d
e
a
c
o
m
p
a
c
t
re
p
r
e
se
n
tatio
n
a
t
th
e
e
n
c
o
d
i
n
g
sta
g
e
a
n
d
g
u
a
ra
n
tee
a
b
e
tt
e
r
re
c
o
n
str
u
c
ti
o
n
q
u
a
l
it
y
o
f
th
e
ima
g
e
a
t
th
e
d
e
c
o
d
i
n
g
sta
g
e
.
Ba
se
d
o
n
c
o
m
p
re
ss
io
n
e
ffici
e
n
c
y
a
n
d
re
c
o
n
stru
c
ted
ima
g
e
q
u
a
li
t
y
,
a
wid
e
ra
n
g
e
o
f
e
x
p
e
rime
n
t
s
h
a
v
e
b
e
e
n
c
o
n
d
u
c
ted
t
o
v
a
li
d
a
te
th
e
p
ro
p
o
s
e
d
tec
h
n
i
q
u
e
p
e
rfo
rm
a
n
c
e
c
o
m
p
a
re
d
with
p
o
p
u
lar
ima
g
e
c
o
m
p
re
ss
io
n
sta
n
d
a
rd
s
a
n
d
e
x
isti
n
g
d
e
e
p
lea
rn
in
g
-
b
a
se
d
m
e
th
o
d
s.
At
a
c
o
m
p
re
ss
io
n
ra
ti
o
(CR)
o
f
8
0
,
th
e
p
r
o
p
o
se
d
m
e
th
o
d
a
c
h
iev
e
d
a
n
a
v
e
ra
g
e
p
e
a
k
sig
n
a
l
-
to
-
n
o
ise
r
a
ti
o
(
P
S
NR)
v
a
lu
e
o
f
3
8
.
9
8
d
B
a
n
d
9
6
.
8
%
sim
il
a
rit
y
in
term
s
o
f
m
u
lt
i
-
sc
a
le
stru
c
tu
ra
l
sim
il
a
rit
y
(M
S
-
S
S
IM
),
d
e
m
o
n
stra
ti
n
g
it
s
e
ffe
c
ti
v
e
n
e
ss
.
K
ey
w
o
r
d
s
:
C
NN
Dee
p
lear
n
in
g
Dis
cr
ete
wav
elet
tr
an
s
f
o
r
m
s
I
m
ag
e
co
m
p
r
ess
io
n
R
etin
al
im
ag
e
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
:
Dalila
C
h
ik
h
ao
u
i
I
n
f
o
r
m
atio
n
Pro
ce
s
s
in
g
an
d
T
elec
o
m
m
u
n
icatio
n
L
ab
o
r
ato
r
y
(
L
T
I
T
)
,
Dep
ar
tm
en
t
o
f
E
lectr
ical
E
n
g
in
ee
r
in
g
Facu
lty
o
f
T
ec
h
n
o
lo
g
y
-
Un
iv
e
r
s
ity
T
AHRI
Mo
h
am
m
ed
B
ec
h
ar
B
ec
h
ar
,
Alg
er
ia
E
m
ail: c
h
ik
h
ao
u
i
.
d
alila@
u
n
iv
-
b
ec
h
ar
.
d
z
1.
I
NT
RO
D
UCT
I
O
N
As
p
ar
t
o
f
m
ed
ical
p
lan
n
in
g
in
h
ea
lth
ca
r
e
s
y
s
tem
s
,
m
ed
ical
im
ag
in
g
p
lay
s
a
s
ig
n
if
ican
t
r
o
le
in
d
ig
ital
im
ag
e
p
r
o
ce
s
s
in
g
.
Me
d
ical
im
ag
in
g
h
as
m
ad
e
ex
ten
s
i
v
e
u
s
e
o
f
r
etin
al
im
ag
es.
T
h
e
d
iag
n
o
s
tic
p
r
o
ce
s
s
f
o
r
s
ev
er
al
ey
e
d
is
ea
s
es,
in
clu
d
in
g
s
tr
o
k
e,
d
iab
etic
r
etin
o
p
at
h
y
,
an
d
g
lau
c
o
m
a,
d
e
p
en
d
s
h
e
av
ily
o
n
th
e
clar
ity
o
f
r
etin
al
im
ag
es
[
1
]
.
I
n
ad
d
itio
n
,
th
e
d
if
f
ic
u
lty
o
f
d
ia
g
n
o
s
i
n
g
s
o
m
e
r
etin
al
d
is
ea
s
es
also
r
e
q
u
ir
es
ad
v
an
ce
d
tech
n
o
lo
g
ies
th
at
allo
w
m
u
lt
ip
le
s
p
ec
ialized
m
ed
ical
tea
m
s
to
ef
f
ec
tiv
ely
s
h
ar
e,
e
x
c
h
an
g
e,
r
etr
iev
e,
an
d
p
r
o
ce
s
s
d
ata.
R
etin
al
im
ag
es
ar
e
f
u
r
t
h
er
u
s
ed
f
o
r
r
esear
c
h
p
u
r
p
o
s
es
to
h
el
p
in
th
e
au
to
m
atic
lab
elin
g
o
f
d
i
s
e
a
s
e
d
v
e
s
s
e
l
s
i
.
e
.
,
a
u
t
o
m
a
t
i
c
d
i
a
g
n
o
s
t
i
c
[
2
]
,
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
[
3
]
,
a
u
t
o
m
a
t
i
c
r
e
t
i
n
a
l
v
e
s
s
e
l
s
e
g
m
e
n
t
a
t
i
o
n
[
4
]
.
T
h
e
d
if
f
e
r
en
t
ap
p
licatio
n
s
o
f
r
etin
al
im
ag
es,
wh
eth
er
f
o
r
m
ed
ical
p
r
ac
tice
o
r
r
esear
c
h
,
h
av
e
m
ar
k
e
d
an
in
cr
ea
s
e
in
m
ed
ical
d
ata
ac
q
u
ir
ed
b
y
th
e
latest
im
ag
in
g
tech
n
o
lo
g
ies,
in
wh
ich
t
h
e
s
ize
o
f
th
e
im
ag
es
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
8
,
No
.
1
,
Ap
r
il
20
2
5
:
2
4
3
-
2
5
3
244
in
cr
ea
s
es
as
th
e
r
eso
lu
tio
n
r
eq
u
ir
em
en
t
in
cr
ea
s
es.
T
h
u
s
,
s
to
r
ag
e
an
d
tr
an
s
m
is
s
io
n
s
y
s
tem
s
ar
e
af
f
ec
ted
.
B
ec
au
s
e
o
f
th
e
lar
g
e
s
ize
o
f
th
e
d
ata
an
d
th
e
n
ee
d
f
o
r
s
to
r
ag
e
ca
p
ac
ity
,
it
is
es
s
en
tial
to
co
m
p
r
ess
th
e
d
ata
f
o
r
p
r
eser
v
atio
n
in
Ho
s
p
ital I
n
f
o
r
m
at
io
n
Sy
s
tem
s
(
HI
S)
o
r
p
ictu
r
e
ar
ch
iv
in
g
an
d
co
m
m
u
n
icatio
n
s
y
s
tem
s
(
PA
C
S)
f
o
r
f
u
tu
r
e
r
e
f
er
en
ce
.
A
v
ar
iet
y
o
f
m
ec
h
an
is
m
s
f
o
r
e
f
f
icien
t
d
ata
p
r
o
ce
s
s
in
g
ar
e
im
p
o
r
t
an
t
as
s
to
r
ag
e
h
as
b
ec
o
m
e
o
n
e
o
f
th
e
g
r
ea
test
ch
allen
g
es.
Mo
r
eo
v
er
,
th
e
p
r
i
m
ar
y
g
o
al
o
f
telem
ed
ici
n
e
is
to
en
ab
le
ef
f
icien
t
r
em
o
te
a
n
aly
s
is
o
f
m
ed
ical
i
m
ag
es.
Sy
s
tem
s
tr
an
s
m
itti
n
g
lar
g
e
am
o
u
n
ts
o
f
m
e
d
ical
i
m
ag
in
g
d
ata
ca
u
s
e
co
m
p
licated
tr
an
s
m
is
s
io
n
p
r
o
ce
s
s
es.
T
r
an
s
m
is
s
io
n
is
m
ain
l
y
m
an
ag
e
d
b
y
r
e
d
u
cin
g
b
it
r
a
tes
in
r
esp
o
n
s
e
to
a
lim
ited
ch
an
n
el
b
an
d
wid
t
h
[
5
]
.
Hen
ce
,
it
is
im
p
o
r
tan
t
to
d
ev
elo
p
ad
eq
u
ate
co
m
p
r
ess
io
n
tech
n
iq
u
es
th
at
r
ed
u
ce
th
e
s
ize
o
f
im
a
g
e
d
ata
wh
ile
p
r
eser
v
in
g
a
r
ea
s
o
n
ab
le
lev
el
o
f
clin
ical
f
id
elity
in
o
r
d
er
to
o
v
er
c
o
m
e
th
e
lim
ited
b
an
d
wid
th
a
n
d
s
to
r
a
g
e
r
eso
u
r
ce
s
.
At
p
r
esen
t,
a
v
a
r
iety
o
f
l
o
s
s
y
co
m
p
r
ess
io
n
tech
n
iq
u
es
f
o
r
m
ed
ical
im
ag
es
h
av
e
b
ee
n
d
e
v
elo
p
ed
in
th
e
liter
atu
r
e.
T
h
ese
tech
n
iq
u
es
ar
e
p
r
im
ar
ily
d
iv
id
ed
in
to
two
ca
teg
o
r
ies:
co
m
p
r
ess
io
n
tech
n
iq
u
es
b
ased
o
n
c
o
n
v
e
n
t
i
o
n
a
l
a
l
g
o
r
i
t
h
m
s
(
n
o
n
-
d
e
e
p
-
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
s
)
a
n
d
c
o
m
p
r
e
s
s
i
o
n
t
e
c
h
n
i
q
u
e
s
b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
[
6
]
.
Gen
er
ally
,
co
n
v
en
tio
n
al
co
m
p
r
ess
io
n
ap
p
r
o
ac
h
es
(
n
o
n
-
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
)
ar
e
r
ea
li
ze
d
b
y
co
m
b
in
in
g
d
if
f
e
r
en
t
tr
an
s
f
o
r
m
s
jo
in
tly
with
a
q
u
an
tizatio
n
s
tep
an
d
en
tr
o
p
y
co
d
in
g
m
eth
o
d
[
7
]
,
[
8
]
.
I
n
ad
d
itio
n
,
co
n
v
en
tio
n
al
ap
p
r
o
ac
h
es
h
a
v
e
b
ee
n
em
p
lo
y
ed
f
o
r
m
e
d
ical
im
ag
e
co
m
p
r
ess
io
n
.
Hän
s
g
en
et
a
l.
[
9
]
,
in
v
esti
g
ated
th
e
ef
f
ec
t
o
f
wa
v
elet
co
m
p
r
ess
io
n
o
n
au
t
o
m
a
tic
an
aly
s
is
task
s
an
d
th
e
d
e
g
r
ad
atio
n
o
f
r
etin
al
im
ag
e
q
u
ality
ca
u
s
ed
b
y
d
if
f
er
en
t
co
m
p
r
ess
io
n
r
atio
s
(
C
R
)
.
E
ik
elb
o
o
m
et
a
l.
[
1
0
]
,
th
e
ef
f
ec
t
o
f
J
PEG
an
d
wav
elet
co
m
p
r
ess
io
n
tech
n
iq
u
es
o
n
d
ig
ital
r
etin
al
im
ag
es
q
u
ality
h
as
b
ee
n
in
v
esti
g
ated
.
K
r
iv
en
k
o
et
a
l.
[
1
1
]
,
p
r
o
p
o
s
ed
an
im
ag
e
co
d
er
b
as
ed
o
n
3
2
×
3
2
p
ix
els
b
lo
ck
s
d
i
s
cr
et
e
co
s
in
e
tr
an
s
f
o
r
m
(
DC
T
)
f
o
r
r
etin
al
im
ag
e
co
m
p
r
ess
io
n
.
M
o
o
k
ia
h
et
a
l.
[
1
2
]
,
th
ey
r
ep
o
r
ted
a
q
u
an
titativ
e
ass
ess
m
en
t
o
f
th
e
ef
f
ec
ts
i
n
d
u
ce
d
b
y
th
e
J
PEG
im
ag
e
co
m
p
r
ess
io
n
alg
o
r
ith
m
o
n
au
to
m
atic
v
ess
el
s
eg
m
en
tatio
n
in
d
ig
ital
r
etin
al
im
ag
e
s
.
I
n
th
e
p
r
ev
io
u
s
s
tu
d
ies,
as th
e
co
m
p
r
ess
io
n
r
atio
in
cr
ea
s
es,
co
n
v
en
tio
n
al
co
m
p
r
ess
io
n
m
eth
o
d
s
,
s
u
ch
as JP
E
G,
ca
u
s
e
b
lo
ck
in
g
ar
tifa
cts
o
r
n
o
is
e
th
at
d
e
g
r
ad
es
th
e
q
u
ality
o
f
t
h
e
d
ec
o
d
ed
im
ag
es.
So
m
e
wo
r
k
s
p
r
o
p
o
s
ed
to
o
v
e
r
co
m
e
t
h
e
p
r
o
b
lem
b
y
im
p
lem
en
tin
g
p
o
s
t
-
p
r
o
ce
s
s
in
g
o
r
d
e
n
o
is
in
g
b
a
s
ed
m
eth
o
d
s
f
o
r
r
etin
al
im
ag
e
p
r
o
ce
s
s
in
g
.
F
o
r
ex
am
p
le,
Naz
ar
i
an
d
Po
u
r
g
h
a
s
s
em
[
1
3
]
,
s
u
g
g
ested
a
n
a
p
p
r
o
ac
h
b
ased
o
n
p
r
e
-
p
r
o
ce
s
s
in
g
v
ess
el
ex
tr
ac
tio
n
,
an
d
p
o
s
t
-
p
r
o
ce
s
s
in
g
t
o
en
h
an
ce
d
etails
in
r
etin
al
im
a
g
es
f
o
r
th
e
ex
tr
ac
tio
n
o
f
lar
g
e
an
d
th
in
b
lo
o
d
v
ess
els
u
s
in
g
a
2
D
Gab
o
r
f
ilter
f
o
llo
w
ed
b
y
lin
ea
r
Ho
u
g
h
tr
an
s
f
o
r
m
atio
n
.
J
av
ed
et
a
l.
[
1
4
]
,
a
tech
n
iq
u
e
o
f
ed
g
e
-
b
ased
en
h
an
ce
m
e
n
t
o
f
r
etin
al
im
ag
e
s
was
p
r
esen
ted
.
I
n
th
is
tech
n
iq
u
e
th
e
im
ag
es
ar
e
p
r
o
ce
s
s
ed
an
d
an
aly
ze
d
i
n
th
e
J
PEG
co
m
p
r
ess
ed
d
o
m
ain
to
en
h
an
ce
th
e
ed
g
e
s
f
o
r
d
is
ea
s
e
d
iag
n
o
s
is
p
er
s
p
ec
tiv
e
.
Salih
et
a
l.
[
1
5
]
,
p
r
esen
ted
an
ef
f
ec
tiv
e
r
etin
a
l
im
ag
e
co
m
p
r
ess
io
n
ap
p
r
o
ac
h
f
o
c
u
s
ed
o
n
th
e
ar
ea
o
f
in
ter
est
(
R
OI
)
.
T
h
is
ap
p
r
o
ac
h
in
clu
d
es
p
r
e
-
p
r
o
ce
s
s
in
g
with
an
ad
a
p
tiv
e
m
ed
ian
f
ilter
,
s
eg
m
en
tatio
n
with
en
h
an
ce
d
ad
a
p
tiv
e
f
u
zz
y
c
-
m
ea
n
s
clu
s
ter
in
g
,
co
m
p
r
ess
io
n
with
in
teg
er
m
u
lti
wav
elet
tr
an
s
f
o
r
m
,
a
n
d
s
e
t
p
ar
titi
o
n
in
g
in
h
ier
ar
ch
ical
tr
ee
,
to
ac
h
iev
e
b
etter
im
a
g
e
q
u
ality
.
L
ately
,
d
ee
p
lear
n
i
n
g
m
eth
o
d
s
h
av
e
b
ee
n
s
u
cc
ess
f
u
lly
a
p
p
lie
d
to
im
ag
e
co
m
p
r
ess
io
n
.
T
h
o
s
e
m
eth
o
d
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
to
b
en
ef
it
f
r
o
m
an
e
n
co
d
i
n
g
an
d
d
ec
o
d
i
n
g
m
o
d
u
le
b
u
ilt
o
f
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
.
Usi
n
g
C
N
Ns
th
e
m
o
d
u
le
en
ab
les
d
im
en
s
io
n
ality
r
ed
u
ctio
n
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
d
u
r
in
g
en
co
d
i
n
g
,
an
d
en
h
an
ce
d
r
ec
o
n
s
tr
u
ctio
n
d
u
r
in
g
d
ec
o
d
in
g
.
B
allé
et
a
l.
[
1
6
]
,
p
r
esen
ted
a
d
ee
p
lear
n
in
g
m
o
d
el
f
o
r
im
ag
e
co
m
p
r
ess
io
n
b
y
s
u
cc
ess
iv
ely
ap
p
ly
in
g
co
n
v
o
lu
tio
n
al
lin
ea
r
f
ilter
s
to
n
o
n
lin
ea
r
ac
tiv
atio
n
f
u
n
ctio
n
s
,
wh
ile
th
e
r
o
u
n
d
in
g
q
u
an
tizer
was
r
ep
lac
ed
b
y
a
u
n
if
o
r
m
q
u
an
tizer
to
en
s
u
r
e
an
u
n
i
n
ter
r
u
p
ted
tr
ain
i
n
g
p
r
o
ce
s
s
.
R
ely
in
g
o
n
m
o
d
el
in
[
1
6
]
,
o
th
er
d
ee
p
lear
n
in
g
ar
c
h
itectu
r
es
f
o
r
im
a
g
e
co
m
p
r
ess
io
n
h
a
v
e
b
ee
n
p
r
o
p
o
s
ed
,
s
u
ch
as
th
e
o
n
e
p
r
o
p
o
s
ed
b
y
C
h
en
g
et
a
l.
[
1
7
]
.
I
n
wh
ich
,
th
e
au
th
o
r
s
in
tr
o
d
u
ce
d
r
esid
u
al
b
lo
c
k
s
in
to
th
e
ar
ch
itectu
r
e
to
in
cr
ea
s
e
th
e
r
ec
ep
tiv
e
f
ield
an
d
im
p
r
o
v
e
co
m
p
r
ess
io
n
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
el.
T
h
o
s
e
d
ee
p
lear
n
in
g
-
b
ased
m
o
d
els
h
av
e
o
u
tp
e
r
f
o
r
m
ed
c
o
n
v
en
tio
n
al
co
m
p
r
ess
io
n
m
et
h
o
d
s
.
I
m
ag
e
co
m
p
r
ess
io
n
tech
n
o
lo
g
y
b
ased
o
n
d
ee
p
lear
n
in
g
was
ap
p
lied
t
o
m
ed
ical
im
ag
es.
Kar
et
a
l.
[
1
8
]
,
p
r
o
p
o
s
ed
a
co
n
v
o
l
u
tio
n
al
au
to
en
co
d
er
ar
ch
itectu
r
e
f
o
r
m
e
d
ical
lo
s
s
y
im
ag
e
co
m
p
r
ess
io
n
to
p
r
eser
v
e
d
iag
n
o
s
tically
r
elev
a
n
t
f
ea
tu
r
es
d
u
r
in
g
co
m
p
r
ess
io
n
.
Su
s
h
m
it
et
a
l.
[
1
9
]
s
u
g
g
ested
a
co
n
v
o
lu
tio
n
a
l
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
ar
ch
itectu
r
e
to
lear
n
co
n
tex
tu
alize
d
f
ea
tu
r
es
f
o
r
e
f
f
icien
t
X
-
r
ay
im
ag
e
c
o
m
p
r
e
s
s
io
n
.
A
co
m
p
r
ess
io
n
m
et
h
o
d
f
o
r
r
etin
a
o
p
tical
co
h
er
en
ce
to
m
o
g
r
ap
h
y
(
OC
T
)
im
ag
es
was
d
ev
elo
p
ed
in
[
2
0
]
,
wh
ich
u
s
es
C
NNs
an
d
s
k
ip
co
n
n
ec
tio
n
s
with
q
u
an
tizatio
n
to
p
r
eser
v
e
f
in
e
s
tr
u
ctu
r
e
f
ea
tu
r
es b
etwe
en
th
e
c
o
m
p
r
ess
io
n
an
d
r
ec
o
n
s
tr
u
ctio
n
C
NNs.
T
h
e
p
r
e
v
io
u
s
ly
r
ev
iewe
d
tec
h
n
iq
u
es
h
a
v
e
r
e
v
ea
led
s
o
m
e
s
h
o
r
tco
m
in
g
s
th
at
n
ee
d
to
b
e
ad
d
r
ess
ed
.
Star
tin
g
with
co
n
v
e
n
tio
n
al
c
o
m
p
r
ess
io
n
alg
o
r
ith
m
s
wh
ich
s
u
f
f
er
ed
f
r
o
m
p
o
o
r
p
er
f
o
r
m
a
n
ce
at
h
ig
h
CR
,
th
e
im
ag
e
q
u
ality
was
d
r
asti
ca
lly
d
eg
r
ad
ed
[
2
1
]
.
T
h
er
e
f
o
r
e,
m
u
ch
ef
f
o
r
t
h
as
b
ee
n
f
o
cu
s
ed
o
n
im
p
r
o
v
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
th
ese
co
m
p
r
e
s
s
io
n
ap
p
r
o
ac
h
es
u
s
in
g
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
p
o
s
t
-
p
r
o
ce
s
s
in
g
m
eth
o
d
s
.
Desp
ite
th
ese
p
er
f
o
r
m
an
ce
im
p
r
o
v
em
en
ts
,
th
ese
m
eth
o
d
s
in
v
o
lv
e
co
m
p
u
tatio
n
ally
ex
p
e
n
s
iv
e
an
d
tim
e
-
co
n
s
u
m
in
g
p
r
o
ce
s
s
es
f
o
r
s
o
lv
in
g
o
p
tim
al
s
o
lu
tio
n
s
.
On
th
e
o
th
er
h
an
d
,
d
ee
p
lear
n
i
n
g
-
b
ased
i
m
ag
e
co
m
p
r
ess
io
n
tech
n
iq
u
es
h
a
v
e
d
em
o
n
s
tr
ated
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
.
Ho
we
v
er
,
th
ei
r
ar
c
h
itectu
r
es
m
ay
r
eq
u
ir
e
d
ee
p
er
C
NN
o
r
lar
g
e
m
o
d
els,
r
esu
ltin
g
in
co
m
p
u
tatio
n
s
th
at
m
a
k
e
th
e
l
ea
r
n
in
g
p
r
o
ce
s
s
s
lo
w.
I
n
ad
d
i
tio
n
,
m
o
s
t
o
f
th
eir
f
ea
tu
r
e
ex
tr
ac
tio
n
ar
c
h
itectu
r
e
s
r
ely
o
n
c
o
n
v
o
lu
tio
n
al
la
y
er
s
with
o
n
e
co
n
v
o
l
u
tio
n
ea
c
h
,
w
h
ich
m
ay
lo
s
e
s
o
m
e
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
A
mu
lti
-
s
ca
le
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
a
n
d
d
is
crete
w
a
ve
let
…
(
Da
lila
C
h
ikh
a
o
u
i
)
245
u
s
ef
u
l
f
ea
tu
r
es
in
m
ed
ical
im
ag
in
g
ter
m
s
.
Dee
p
lear
n
i
n
g
m
o
d
els
ar
e
in
co
m
p
atib
le
with
c
o
n
v
en
tio
n
al
co
d
ec
s
,
h
en
ce
th
eir
a
p
p
licatio
n
is
lim
ited
.
I
n
r
esp
o
n
s
e
to
th
e
ab
o
v
e
s
h
o
r
t
co
m
in
g
s
,
th
is
p
ap
er
p
r
o
p
o
s
es
a
m
ed
ical
im
ag
e
co
m
p
r
ess
io
n
tech
n
iq
u
e
b
ased
o
n
a
lo
w
co
m
p
le
x
ity
d
ee
p
lear
n
in
g
m
o
d
el
an
d
a
d
is
cr
ete
wav
elet
tr
an
s
f
o
r
m
(
DW
T
)
b
ased
co
d
ec
.
Mo
tiv
ated
b
y
th
e
ad
v
a
n
tag
es
o
f
C
NNs
lik
e
th
e
ab
ili
ty
to
au
t
o
m
atica
lly
d
etec
t
f
ea
tu
r
es
an
d
th
eir
co
m
p
u
tatio
n
ally
ef
f
icien
t
ch
a
r
ac
ter
is
tics
[
2
2
]
.
A
C
NN
ar
ch
it
ec
tu
r
e
with
a
lo
w
p
ar
am
ete
r
c
o
u
n
t
is
d
esig
n
ed
f
o
r
b
o
th
e
n
co
d
i
n
g
a
n
d
d
ec
o
d
in
g
,
en
h
a
n
ce
d
b
y
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
al
lay
er
s
.
T
h
e
p
r
o
p
o
s
ed
d
ee
p
lear
n
i
n
g
ar
ch
itectu
r
e
is
in
teg
r
ate
d
with
th
e
DW
T
-
b
ased
co
d
ec
f
o
r
ef
f
ec
tiv
e
p
er
f
o
r
m
an
ce
at
h
ig
h
CR
,
o
win
g
to
DW
T
’
s
co
m
p
u
tatio
n
al
e
f
f
icien
cy
a
n
d
co
m
p
ac
t
s
ig
n
al
r
ep
r
esen
tatio
n
o
f
th
e
DW
T
,
wh
ich
is
wid
ely
u
s
ed
in
im
ag
e
co
d
in
g
[
2
3
]
.
T
h
e
m
ain
co
n
tr
ib
u
tio
n
s
o
f
th
is
p
ap
er
ar
e
s
u
m
m
ar
ized
as
f
o
llo
ws.
I
n
itially
,
to
im
p
r
o
v
e
th
e
co
m
p
r
ess
io
n
p
er
f
o
r
m
an
ce
,
th
e
m
u
lti
-
s
ca
le
C
NN
(
MS
-
C
NN)
o
n
th
e
en
co
d
in
g
s
id
e
d
er
iv
es
a
n
o
p
tim
al
co
m
p
ac
t
r
ep
r
esen
tatio
n
th
at
h
o
ld
s
im
p
o
r
tan
t
s
tr
u
ctu
r
al
d
ata
f
r
o
m
th
e
o
r
ig
in
al
im
ag
e.
W
h
ile
o
n
th
e
d
ec
o
d
in
g
s
id
e,
th
e
s
ec
o
n
d
MS
-
C
NN
allo
ws
ac
cu
r
ate
r
ec
o
n
s
tr
u
ctio
n
o
f
th
e
o
u
t
p
u
t
im
ag
e.
Seco
n
d
,
a
DW
T
-
b
ased
im
ag
e
co
d
ec
r
esid
in
g
b
etwe
en
th
e
en
co
d
in
g
M
S
-
C
NN
an
d
d
ec
o
d
in
g
M
S
-
C
NN
ca
n
b
e
ef
f
ec
tiv
ely
u
tili
ze
d
b
y
tak
in
g
a
co
m
p
ac
t
r
ep
r
esen
tatio
n
as
in
p
u
t
f
o
r
f
u
r
th
er
c
o
m
p
r
ess
io
n
.
T
h
ir
d
ly
,
we
p
r
esen
t
a
lear
n
i
n
g
s
tr
ateg
y
f
o
r
t
h
e
MS
-
C
NN
s
,
wh
ich
o
v
er
c
o
m
es
th
e
p
r
o
b
lem
o
f
tr
ain
in
g
in
ter
r
u
p
tio
n
ca
u
s
ed
b
y
n
o
n
-
d
if
f
er
en
tiab
le
q
u
an
tizatio
n
in
th
e
DW
T
c
o
d
ec
.
As
d
em
o
n
s
tr
ated
b
y
th
e
ex
p
er
im
en
t
al
r
esu
lts
,
th
e
p
r
o
p
o
s
ed
tec
h
n
iq
u
e
o
u
tp
e
r
f
o
r
m
s
ex
is
tin
g
tech
n
iq
u
es
an
d
s
tan
d
ar
d
co
m
p
r
ess
io
n
tec
h
n
iq
u
es
i
n
ter
m
s
o
f
s
ev
er
al
m
etr
ics.
C
o
n
n
ec
tin
g
th
e
d
ee
p
lear
n
in
g
m
o
d
el
with
DW
T
co
d
ec
u
s
i
n
g
a
c
o
m
p
ac
t
i
n
ter
m
ed
iate
r
ep
r
esen
tatio
n
a
llo
ws
th
e
p
r
o
p
o
s
ed
co
m
p
r
ess
io
n
tech
n
iq
u
e
to
ex
h
ib
it
co
m
p
atib
ilit
y
with
o
th
er
av
ailab
le
im
ag
e
co
d
in
g
s
tan
d
ar
d
s
.
T
o
th
e
b
est
o
f
o
u
r
k
n
o
wled
g
e
,
th
is
is
th
e
f
ir
s
t
s
tu
d
y
to
u
s
e
MS
-
C
NNs
to
en
h
an
ce
th
e
co
m
p
r
ess
io
n
p
er
f
o
r
m
a
n
ce
o
f
co
n
v
en
tio
n
al
DW
T
-
b
ased
co
d
ec
s
an
d
ac
h
iev
e
h
ig
h
CR
with
ac
cu
r
ate
m
ed
ical
im
ag
e
r
ec
o
n
s
tr
u
ctio
n
.
T
h
e
s
u
cc
ee
d
in
g
p
ar
t
o
f
th
e
ar
ticle
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws:
a
d
escr
ip
tio
n
o
f
th
e
k
ey
co
m
p
o
n
e
n
ts
o
f
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
is
g
iv
en
i
n
s
ec
tio
n
2
.
I
n
s
ec
ti
o
n
3
p
r
o
v
id
es
a
co
m
p
ar
is
o
n
an
d
d
is
cu
s
s
io
n
o
f
s
im
u
latio
n
r
esu
lts
.
Fin
ally
,
co
n
clu
s
io
n
is
p
r
esen
ted
in
s
ec
tio
n
4
.
2.
M
E
T
H
O
D
Data
co
m
p
r
ess
io
n
b
ased
o
n
d
ee
p
lear
n
in
g
is
a
p
r
o
m
is
in
g
r
e
s
ea
r
ch
ar
ea
.
T
h
ese
tech
n
iq
u
es
s
p
ec
ialize
in
m
an
y
asp
ec
ts
,
in
clu
d
in
g
tr
ain
in
g
a
n
d
lear
n
in
g
ab
i
liti
es
[
2
4
]
.
L
o
s
s
y
co
m
p
r
es
s
io
n
b
y
r
ed
u
cin
g
d
im
en
s
io
n
ality
is
o
n
e
o
f
th
e
m
ajo
r
ca
teg
o
r
ies
o
f
co
m
p
r
ess
io
n
b
ased
o
n
d
ee
p
lear
n
in
g
te
ch
n
iq
u
es,
in
w
h
ich
p
er
f
o
r
m
an
ce
is
co
m
p
ar
ab
le
t
o
o
r
e
v
en
b
etter
th
an
s
tan
d
a
r
d
co
d
ec
s
[
1
6
]
,
[
2
0
]
.
Dim
en
s
i
o
n
ality
r
ed
u
ctio
n
is
ac
co
m
p
lis
h
ed
b
y
lear
n
in
g
an
in
v
er
tib
le
m
a
p
p
in
g
b
etwe
en
th
e
q
u
a
n
tized
co
m
p
ac
t
r
ep
r
esen
tatio
n
an
d
th
e
o
r
ig
in
al
d
ata.
T
h
is
p
r
o
ce
s
s
r
elies
m
ain
ly
o
n
d
ee
p
ar
ch
ite
ctu
r
es
o
f
C
NNs
wh
ich
allo
w
ef
f
icien
t
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
ex
h
ib
it
a
g
o
o
d
r
ep
r
esen
tativ
e
ab
ilit
y
[
2
5
]
.
U
s
u
ally
,
a
C
NN
is
ca
s
ca
d
ed
at
b
o
th
th
e
en
co
d
in
g
an
d
d
ec
o
d
in
g
e
n
d
s
w
h
en
b
u
ild
in
g
th
ese
d
ee
p
lear
n
in
g
m
o
d
els.
I
n
v
iew
o
f
th
is
,
o
u
r
p
r
esen
ted
lo
s
s
y
co
m
p
r
ess
io
n
tech
n
i
q
u
e
in
v
o
lv
es two
MS
-
C
NN
s
an
d
a
DW
T
im
ag
e
co
d
ec
,
as sh
o
wn
in
Fig
u
r
e
1
.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
c
o
m
p
r
ess
io
n
tech
n
iq
u
e
o
v
er
all
d
esig
n
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
8
,
No
.
1
,
Ap
r
il
20
2
5
:
2
4
3
-
2
5
3
246
Acc
o
r
d
in
g
t
o
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
,
th
e
i
n
p
u
t
im
a
g
e
will
u
n
d
er
g
o
th
e
f
ir
s
t
MS
-
C
NN
r
esid
in
g
at
th
e
en
co
d
in
g
wh
ich
g
e
n
er
ates
a
co
m
p
ac
t
r
ep
r
esen
tatio
n
th
at
p
r
eser
v
es
th
e
s
tr
u
ct
u
r
al
i
n
f
o
r
m
atio
n
o
f
th
e
in
p
u
t.
T
h
e
g
e
n
er
ated
c
o
m
p
ac
t
r
ep
r
e
s
en
tatio
n
is
f
u
r
th
e
r
en
c
o
d
ed
,
s
in
ce
it
allo
ws
th
e
DW
T
b
ased
co
d
ec
to
ac
h
iev
e
ef
f
icien
t
co
m
p
r
ess
io
n
with
a
h
ig
h
CR
.
On
th
e
d
ec
o
d
in
g
s
id
e,
a
s
ec
o
n
d
MS
-
C
N
N
i
s
ap
p
lied
in
o
r
d
er
to
p
r
o
d
u
ce
a
m
o
r
e
ac
cu
r
ate
an
d
h
ig
h
-
q
u
ality
r
ec
o
n
s
tr
u
c
te
d
im
ag
e.
T
h
e
two
n
etwo
r
k
s
co
o
p
er
ate
to
co
m
p
r
ess
im
ag
es
at
a
v
er
y
lo
w
b
it
r
ate
wh
ile
m
ain
tain
in
g
h
ig
h
q
u
alit
y
.
Un
lik
e
d
ee
p
lear
n
in
g
m
o
d
e
ls
with
m
illi
o
n
s
o
f
p
ar
am
eter
s
,
o
u
r
m
eth
o
d
in
c
o
r
p
o
r
ates
a
DW
T
-
b
ased
co
d
ec
,
k
n
o
wn
f
o
r
its
id
ea
l
p
r
o
p
er
ties
an
d
l
o
w
co
m
p
u
tatio
n
al
c
o
m
p
lex
ity
in
i
m
ag
e
co
m
p
r
ess
io
n
task
s
[
2
3
]
.
T
h
e
f
o
llo
win
g
s
u
b
s
ec
tio
n
s
p
r
o
v
id
e
m
o
r
e
d
etails
ab
o
u
t
th
e
k
ey
co
m
p
o
n
en
ts
o
f
th
e
p
r
o
p
o
s
ed
tech
n
i
q
u
e,
s
u
ch
as
th
e
MS
-
C
NN
ar
ch
itectu
r
e,
DW
T
b
ased
co
d
ec
,
th
e
lo
s
s
f
u
n
ctio
n
s
an
d
tr
ain
in
g
s
tr
ateg
y
.
2
.
1
.
Arch
it
ec
t
ure
o
f
t
he
M
S
-
CNNs
Ou
r
ap
p
r
o
ac
h
f
u
r
th
er
le
v
er
ag
es
C
NN
’
s
r
o
b
u
s
tn
ess
,
b
y
ad
d
in
g
a
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
al
b
lo
ck
s
,
wh
ich
h
av
e
b
ee
n
p
r
ev
i
o
u
s
ly
em
p
lo
y
ed
f
o
r
class
if
icatio
n
[
2
2
]
a
n
d
im
a
g
e
s
u
p
er
-
r
eso
lu
tio
n
[
2
3
]
.
T
h
e
co
n
v
o
l
u
tio
n
al
lay
er
s
a
u
to
m
ati
ca
lly
ex
tr
a
ct
lo
ca
l
f
ea
tu
r
es
o
f
in
p
u
t
im
ag
es
th
r
o
u
g
h
th
e
lea
r
n
in
g
p
r
o
ce
s
s
b
ased
o
n
th
e
g
iv
en
tr
ai
n
in
g
d
ataset.
I
n
Fig
u
r
e
2
,
th
e
m
ain
d
i
f
f
er
en
ce
s
b
etwe
en
m
u
ltis
ca
le
co
n
v
o
lu
tio
n
an
d
b
asic
co
n
v
o
l
u
tio
n
ar
e
illu
s
tr
ated
in
Fig
u
r
e
2
(
a)
an
d
Fig
u
r
e
2
(
b
)
,
r
esp
ec
tiv
ely
.
I
n
g
en
e
r
al,
s
m
all
s
ize
k
er
n
els
ten
d
to
ex
tr
ac
t
f
ea
tu
r
es
with
s
m
aller
s
ca
les,
s
u
ch
as
d
etails,
wh
ile
co
ar
s
e
s
tr
u
ctu
r
es
r
esp
o
n
d
to
k
er
n
els
with
lar
g
e
s
ca
les
[
2
6
]
.
T
h
er
e
f
o
r
e,
em
p
lo
y
in
g
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
s
with
s
m
all
an
d
lar
g
e
k
er
n
el
s
izes
is
f
av
o
r
ab
le
to
g
u
ar
an
tee
a
n
ef
f
icien
t e
x
tr
ac
tio
n
o
f
th
e
d
if
f
er
en
t scale
f
ea
tu
r
es f
o
u
n
d
i
n
m
ed
ical
im
a
g
es.
(
a)
(
b
)
Fig
u
r
e
2
.
Dif
f
e
r
en
ce
b
etwe
en
(
a)
b
asic c
o
n
v
o
lu
tio
n
al
la
y
er
o
f
C
NN
an
d
(
b
)
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
al
lay
e
r
T
h
e
en
co
d
i
n
g
an
d
d
ec
o
d
i
n
g
MS
-
C
NN
s
ar
ch
itectu
r
e
d
etails
ar
e
g
iv
en
in
T
ab
le
1
.
T
h
e
en
co
d
in
g
s
id
e
MS
-
C
NN
in
co
r
p
o
r
ates
b
asic
an
d
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
s
an
d
a
d
o
wn
-
s
am
p
lin
g
o
p
er
ati
o
n
to
r
ed
u
ce
in
p
u
t
d
im
en
s
io
n
ality
.
T
h
e
m
u
lti
-
s
ca
le
co
n
v
o
l
u
tio
n
al
b
lo
c
k
(
MSC
B
)
co
n
s
is
ts
o
f
a
m
u
lti
-
s
ca
le
co
n
v
o
lu
ti
o
n
al
lay
er
with
th
r
ee
p
ar
allel
co
n
v
o
lu
ti
o
n
s
o
f
v
ar
y
in
g
k
er
n
el
s
izes,
p
r
o
d
u
cin
g
f
ea
tu
r
e
m
a
p
s
co
n
c
aten
ated
alo
n
g
th
e
s
p
ec
tr
al
d
im
en
s
io
n
,
as
s
h
o
w
n
in
Fig
u
r
e
3
.
Mo
r
e
o
v
er
,
a
d
o
wn
-
s
am
p
lin
g
o
p
e
r
atio
n
o
cc
u
r
s
b
y
u
s
in
g
s
tr
id
e
co
n
v
o
l
u
tio
n
in
o
r
d
e
r
to
co
n
t
r
ac
t
s
p
atial
d
im
en
s
io
n
b
y
a
f
ac
to
r
o
f
2
.
I
n
g
en
er
al,
th
e
en
co
d
in
g
MS
-
C
NN
co
n
s
is
ts
o
f
2
MSC
B
s
an
d
co
n
v
o
lu
tio
n
al
lay
er
s
f
o
llo
wed
e
ac
h
b
y
a
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
n
o
n
lin
ea
r
ac
tiv
atio
n
f
u
n
ctio
n
,
as
s
h
o
wn
in
T
ab
le
1
.
T
h
e
ar
c
h
itectu
r
e
in
v
o
lv
es
also
s
k
ip
c
o
n
n
ec
tio
n
s
,
wh
ich
h
av
e
b
ee
n
co
n
s
id
er
ed
o
n
e
o
f
th
e
m
o
s
t
ef
f
icien
t
s
o
lu
tio
n
s
f
o
r
tr
ain
in
g
d
ee
p
n
etwo
r
k
s
[
2
4
]
.
T
h
e
s
k
ip
c
o
n
n
ec
tio
n
is
clea
r
ly
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
e
MS
-
C
NN
at
th
e
d
ec
o
d
in
g
s
id
e
is
co
m
p
o
s
ed
o
f
5
weig
h
t
lay
e
r
s
an
d
o
n
e
MSC
B
,
as
s
h
o
wn
in
T
a
b
le
1
.
I
n
d
etail,
th
r
ee
ty
p
es
o
f
lay
er
s
ar
e
p
r
esen
t
in
th
e
ar
c
h
itectu
r
e,
n
o
t
ab
ly
,
d
ec
o
n
v
o
lu
tio
n
lay
er
s
,
u
p
-
s
am
p
lin
g
lay
er
an
d
m
u
lti
-
s
ca
le
d
ec
o
n
v
o
lu
tio
n
.
T
h
e
co
m
p
ac
t
r
e
p
r
esen
tatio
n
u
n
d
er
g
o
es
t
h
e
r
e
v
er
s
e
p
r
o
ce
s
s
at
th
e
d
ec
o
d
in
g
s
id
e
to
s
ec
u
r
e
ac
cu
r
ate
r
esto
r
atio
n
o
f
th
e
o
r
ig
in
al
im
a
g
e.
T
ab
le
1
.
Ar
c
h
itectu
r
e
d
etails o
f
th
e
MS
-
C
NNs a
t th
e
en
co
d
in
g
an
d
d
ec
o
d
i
n
g
s
id
es
MS
-
C
N
N
a
t
e
n
c
o
d
i
n
g
MS
-
C
N
N
a
t
d
e
c
o
d
i
n
g
C
o
n
v
(
3
×
3
,
3
2
,
st
r
i
d
e
=
1
)
,
R
e
L
U
D
e
c
o
n
v
(
3
×
3
,
3
2
,
st
r
i
d
e
=
2
)
,
R
e
L
U
M
S
C
B
(
3
2
)
M
S
C
B
(
6
4
)
C
o
n
v
(
3
×
3
,
3
2
,
st
r
i
d
e
=
1
)
,
R
e
L
U
D
e
c
o
n
v
(
3
×
3
,
1
2
8
,
st
r
i
d
e
=
1
)
,
R
e
L
U
M
S
C
B
(
3
2
)
D
e
c
o
n
v
(
3
×
3
,
6
4
,
st
r
i
d
e
=
1
)
,
R
e
L
U
C
o
n
v
(
3
×
3
,
6
4
,
st
r
i
d
e
=
2
)
,
R
e
L
U
D
e
c
o
n
v
(
3
×
3
,
6
4
,
st
r
i
d
e
=
1
)
,
R
e
L
U
C
o
n
v
(
3
×
3
,
9
6
,
st
r
i
d
e
=
1
)
,
R
e
L
U
D
e
c
o
n
v
(
3
×
3
,
3
,
s
t
r
i
d
e
=
1
)
,
R
e
L
U
C
o
n
v
(
3
×
3
,
9
6
,
st
r
i
d
e
=
1
)
,
R
e
L
U
C
o
n
v
(
3
×
3
,
6
4
,
st
r
i
d
e
=
1
)
,
R
e
L
U
C
o
n
v
(
3
×
3
,
3
,
s
t
r
i
d
e
=
1
)
,
R
e
L
U
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
A
mu
lti
-
s
ca
le
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
a
n
d
d
is
crete
w
a
ve
let
…
(
Da
lila
C
h
ikh
a
o
u
i
)
247
Fig
u
r
e
3
.
Pro
p
o
s
ed
MSC
B
2
.
2
.
DWT
ba
s
ed
co
dec
T
h
e
C
o
h
en
-
Dau
b
ec
h
ies
-
Feau
v
ea
u
9
/7
(
C
DF
9
/7
)
b
io
r
th
o
g
o
n
al
wav
elet
tr
an
s
f
o
r
m
s
er
v
es
as
th
e
f
o
u
n
d
a
t
i
o
n
o
f
t
h
e
o
u
r
c
o
d
e
c
t
o
a
l
l
o
w
e
f
f
i
c
i
e
n
t
c
o
m
p
r
e
s
s
i
o
n
.
F
o
r
i
m
a
g
e
d
e
c
o
r
r
e
l
a
t
i
o
n
,
A
n
t
o
n
i
n
i
a
n
d
B
a
r
l
a
u
d
[
2
7
]
,
d
em
o
n
s
tr
ated
th
e
s
u
p
er
io
r
ity
o
f
th
e
b
i
o
r
th
o
g
o
n
al
wav
elet
tr
an
s
f
o
r
m
9
/7
.
J
PEG
-
2
0
0
0
c
o
d
ec
an
d
o
th
e
r
im
a
g
e
co
d
in
g
m
eth
o
d
s
[
2
3
]
,
[
2
8
]
,
h
a
v
e
wid
ely
r
elie
d
o
n
th
is
tr
an
s
f
o
r
m
.
T
h
e
C
DF
f
am
ily
o
f
s
y
m
m
etr
ic
b
io
r
th
o
g
o
n
al
wav
elets a
r
e
d
is
tin
g
u
is
h
ed
b
y
th
eir
co
m
p
ac
t su
p
p
o
r
t,
b
io
r
th
o
g
o
n
ality
,
s
y
m
m
etr
y
,
an
d
s
im
p
l
icity
.
Acc
o
r
d
in
g
to
T
ab
le
2
,
th
e
C
DF 9
/7
wav
elet
h
as 9
co
ef
f
icien
ts
in
t
h
e
lo
w
p
ass
f
ilter
an
d
7
co
ef
f
icien
ts
in
th
e
h
ig
h
p
ass
f
ilter
.
T
ab
le
2
.
Fil
ter
co
ef
f
icien
ts
o
f
C
DF 9
/7
wav
elet
i
Lo
w
-
p
a
ss fi
l
t
e
r
H
i
g
h
-
p
a
ss
f
i
l
t
e
r
0
+
0
.
6
0
2
9
4
9
0
1
8
2
3
6
3
5
7
9
+
1
.
1
1
5
0
8
7
0
5
2
4
5
7
0
0
±
1
+
0
.
2
6
6
8
6
4
1
1
8
4
4
2
8
7
5
-
0
.
5
9
1
2
7
1
7
6
3
1
1
4
2
5
±
2
−
0
.
0
7
8
2
2
3
2
6
6
5
2
8
9
9
0
-
0
.
0
5
7
5
4
3
5
2
6
2
2
8
5
0
±
3
–
0
.
0
1
6
8
6
4
1
1
8
4
4
2
8
7
5
+
0
.
0
9
1
2
7
1
7
6
3
1
1
4
2
5
±
4
+
0
.
0
2
6
7
4
8
7
5
7
4
1
0
8
1
0
T
h
e
u
s
e
o
f
th
e
wav
elet
f
o
r
co
m
p
r
ess
io
n
is
d
ep
e
n
d
en
t
o
n
t
h
e
q
u
an
tizatio
n
s
tep
.
A
lo
s
s
y
c
o
m
p
r
ess
io
n
ap
p
r
o
ac
h
u
s
es q
u
an
tizatio
n
,
w
h
ich
is
ad
ju
s
ted
to
attain
th
e
d
esire
d
CR
.
T
h
e
q
u
an
tized
wav
elet
co
ef
f
icien
ts
ar
e
s
u
b
s
eq
u
en
tly
e
n
tr
o
p
y
en
c
o
d
ed
v
ia
ar
ith
m
etic
c
o
d
in
g
[
2
9
]
.
B
y
co
d
i
n
g
th
e
m
o
s
t
f
r
eq
u
e
n
t
s
y
m
b
o
ls
with
f
ewe
r
b
its
,
it
is
m
o
r
e
ef
f
icie
n
t
th
an
co
d
in
g
th
em
all
with
th
e
s
am
e
b
its
n
u
m
b
er
.
E
n
tr
o
p
y
e
n
c
o
d
in
g
,
s
p
ec
if
ically
ar
ith
m
etic
en
co
d
in
g
,
p
r
o
v
i
d
es
lo
s
s
le
s
s
co
m
p
r
ess
io
n
s
in
ce
t
h
e
o
r
ig
in
al
d
ata
ca
n
b
e
r
ec
o
v
er
ed
in
th
e
d
ec
o
d
e
r
s
tag
e
with
o
u
t
af
f
ec
tin
g
d
ee
p
lear
n
in
g
m
o
d
els.
T
h
er
ef
o
r
e,
we
d
id
n
o
t
in
clu
d
e
th
e
ar
ith
m
etic
co
d
in
g
in
th
e
tr
ain
in
g
o
f
n
etwo
r
k
s
in
o
r
d
e
r
to
m
i
n
im
ize
u
n
n
ec
ess
ar
y
c
o
m
p
lex
ity
.
Ou
r
c
o
d
ec
b
ased
o
n
DW
T
ca
n
b
e
s
u
m
m
ar
ized
in
to
t
h
e
f
o
llo
win
g
s
tep
s
:
−
Dec
o
m
p
o
s
itio
n
o
f
th
e
in
p
u
t c
o
m
p
ac
t r
ep
r
esen
tatio
n
u
s
in
g
2
D
wav
elet
tr
an
s
f
o
r
m
s
(
C
DF 9
/
7
)
.
−
Qu
an
tizatio
n
o
f
t
h
e
wav
elet
co
ef
f
icien
ts
.
−
L
o
s
s
less
co
m
p
r
ess
io
n
u
s
in
g
ar
ith
m
etic
en
co
d
in
g
.
2
.
3
.
L
o
s
s
f
un
ct
io
ns
a
nd
o
ptim
iza
t
io
n
T
h
e
o
b
jectiv
e
is
to
o
p
tim
ize
b
o
th
MS
-
C
NNs
to
ac
h
iev
e
an
e
f
f
icien
t
co
m
p
r
ess
io
n
an
d
a
b
et
ter
im
ag
e
q
u
ality
r
ec
o
n
s
tr
u
ctio
n
.
I
n
o
r
d
e
r
to
o
p
tim
ize
o
u
r
m
o
d
el,
a
lo
s
s
ter
m
n
ee
d
s
to
b
e
m
i
n
im
ized
o
v
er
th
e
p
ar
am
ete
r
s
o
f
th
e
p
r
o
p
o
s
ed
n
etwo
r
k
s
.
T
h
e
d
is
to
r
tio
n
b
etwe
en
th
e
in
p
u
t
an
d
r
ec
o
n
s
tr
u
cted
im
ag
es
r
e
p
r
esen
ts
th
e
lo
s
s
,
an
d
it c
an
b
e
ex
p
r
ess
ed
as:
(
,
)
=
1
∑
‖
(
,
(
(
,
)
)
)
−
‖
2
=
1
)
(
1
)
i
n
th
e
(
1
)
,
m
ea
n
s
q
u
ar
e
er
r
o
r
(
MSE
)
is
u
s
ed
in
th
e
lo
s
s
f
u
n
ctio
n
as
a
d
is
to
r
tio
n
ter
m
,
with
d
en
o
tin
g
th
e
in
p
u
t
im
ag
e.
(
.
)
an
d
(
.
)
in
d
icate
o
f
th
e
MS
-
C
NN
at
th
e
en
co
d
in
g
s
id
e
an
d
MS
-
C
NN
at
th
e
d
ec
o
d
in
g
s
id
e
,
with
,
as
th
eir
v
ar
ia
b
les,
r
esp
ec
tiv
ely
,
wh
er
ea
s
d
e
n
o
tes
th
e
DW
T
b
ased
co
d
ec
.
T
h
e
in
p
u
t
im
ag
e
wen
t
th
r
o
u
g
h
s
tag
es
o
f
co
m
p
r
ess
io
n
,
n
am
ely
,
MS
-
C
NN
f
o
r
co
m
p
ac
t
r
e
p
r
esen
tatio
n
an
d
DW
T
co
d
ec
,
th
en
s
ec
o
n
d
MS
-
C
NN
f
o
r
r
ec
o
n
s
tr
u
ctin
g
th
e
im
ag
e.
Ho
wev
er
,
t
h
e
r
o
u
n
d
in
g
f
u
n
ctio
n
in
co
r
p
o
r
ated
in
th
e
DW
T
-
b
ased
co
d
ec
ca
n
n
o
t
b
e
d
if
f
er
e
n
tiated
wh
en
p
er
f
o
r
m
in
g
th
e
b
ac
k
p
r
o
p
ag
atio
n
alg
o
r
ith
m
.
T
o
ad
d
r
ess
th
is
is
s
u
e,
th
e
tr
ain
in
g
will
b
e
p
er
f
o
r
m
e
d
in
two
p
h
ases
.
T
h
e
f
ir
s
t
p
h
ase
in
v
o
lv
es
tr
ain
in
g
b
o
t
h
n
e
two
r
k
s
with
o
u
t
th
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
8
,
No
.
1
,
Ap
r
il
20
2
5
:
2
4
3
-
2
5
3
248
DW
T
co
d
ec
,
wh
er
ea
s
th
e
s
ec
o
n
d
p
h
ase
in
v
o
lv
es
f
in
etu
n
in
g
th
e
n
etwo
r
k
o
n
th
e
d
e
co
d
i
n
g
s
id
e
tak
in
g
in
to
co
n
s
id
er
atio
n
th
e
co
d
ec
.
2
.
3
.
1
.
MS
-
CNNs t
ra
ini
ng
Ass
u
m
in
g
a
co
llectio
n
o
f
in
p
u
t
im
ag
es
u
n
d
er
g
o
es
f
ir
s
t
an
M
S
-
C
NN
to
lear
n
an
o
p
tim
u
m
c
o
m
p
ac
t
r
ep
r
esen
tatio
n
a
n
d
r
eser
v
e
th
e
s
tr
u
ctu
r
al
in
f
o
r
m
atio
n
.
T
h
e
r
ec
o
n
s
tr
u
ctio
n
MS
-
C
NN
is
th
en
em
p
l
o
y
ed
to
r
ec
o
v
er
t
h
e
d
ec
o
d
ed
im
ag
e
w
ith
h
ig
h
q
u
ality
,
h
en
ce
t
h
e
m
ea
n
s
q
u
ar
ed
er
r
o
r
lo
s
s
f
u
n
ctio
n
u
s
ed
f
o
r
tr
ai
n
in
g
ca
n
b
e
d
e
f
in
ed
as sh
o
wn
in
(
2
)
:
1
(
,
)
=
1
∑
‖
(
,
(
,
)
)
−
‖
2
=
1
(
2
)
wh
er
e
an
d
d
en
o
te
th
e
tr
ain
a
b
le
v
ar
iab
le,
wh
e
r
ea
s
N
d
en
o
tes
th
e
b
atch
s
ize.
2
.
3
.
2
.
MS
-
CNN
f
ine
-
t
un
ing
Du
r
in
g
MS
-
C
NN
r
ec
o
n
s
tr
u
cti
o
n
,
th
e
o
u
tp
u
t
im
ag
e
is
r
ec
o
n
s
tr
u
cted
in
a
way
th
at
clo
s
ely
r
ep
licates
th
e
in
p
u
t
im
ag
e.
T
h
er
e
f
o
r
e,
t
h
e
d
ec
o
d
ed
co
m
p
ac
t
r
ep
r
esen
tatio
n
d
e
r
iv
ed
f
r
o
m
co
m
p
r
ess
io
n
n
etwo
r
k
E
th
en
DW
T
b
ased
co
d
ec
D
will
b
e
p
ass
ed
th
r
o
u
g
h
en
co
d
in
g
M
S
-
C
NN
R
to
lear
n
m
o
r
e
ac
c
u
r
ate
r
ec
o
n
s
tr
u
ctio
n
.
T
h
e
p
ar
am
eter
̂
was
f
ix
ed
wh
i
le
th
e
en
co
d
in
g
n
etwo
r
k
p
ar
a
m
eter
was
o
p
tim
ized
,
th
e
lo
s
s
f
u
n
ctio
n
u
s
e
d
f
o
r
f
in
e
-
t
u
n
in
g
th
e
MS
-
C
NN
ca
n
b
e
f
o
r
m
u
lated
as:
2
(
)
=
1
∑
‖
(
,
(
(
̂
,
)
)
)
−
‖
2
=
1
(
3
)
2
.
4
.
E
v
a
lua
t
i
o
n m
et
rics
I
n
o
r
d
er
to
ca
r
r
y
o
u
t
a
q
u
a
n
titativ
e
ass
ess
m
en
t
o
f
o
u
r
m
eth
o
d
’
s
p
er
f
o
r
m
an
ce
,
we
ad
o
p
ted
ev
alu
atio
n
m
etr
ics
b
ased
o
n
im
a
g
e
q
u
ali
ty
r
ec
o
n
s
tr
u
ctio
n
an
d
th
e
ef
f
i
cien
cy
o
f
c
o
m
p
r
ess
io
n
.
T
h
e
r
ec
o
n
s
tr
u
cted
im
a
g
e
q
u
ality
is
ev
alu
ated
u
s
in
g
t
h
e
p
ea
k
s
ig
n
al
-
to
-
n
o
is
e
r
atio
(
P
SNR
)
,
MSE
an
d
m
u
ltis
ca
le
s
tr
u
ctu
r
al
s
im
ilar
ity
(
MS
-
SSI
M)
.
T
h
e
MSE
an
d
PS
NR
m
ea
s
u
r
e
d
is
to
r
tio
n
b
e
twee
n
th
e
o
r
ig
in
al
an
d
r
ec
o
n
s
tr
u
cted
im
ag
es
to
ev
alu
ate
v
is
u
al
q
u
ality
[
3
0
]
,
as g
iv
en
in
(
4
)
an
d
(
5
)
,
=
1
∑
∑
[
(
,
)
−
̂
(
,
)
]
2
−
1
=
0
−
1
=
0
(
4
)
=
10
l
og
10
(
2
)
(
5
)
wh
er
e
an
d
̂
ar
e
th
e
in
p
u
t
a
n
d
r
ec
o
n
s
tr
u
cted
im
ag
es
r
esp
ec
tiv
ely
.
M,
N
ar
e
t
h
e
n
u
m
b
er
o
f
r
o
ws
an
d
co
lu
m
n
s
o
f
t
h
e
im
ag
e,
wh
ile
is
th
e
m
ax
im
u
m
v
alu
e
o
f
p
ix
el
in
th
e
im
ag
e.
Similar
ity
is
a
r
eso
lu
tio
n
im
ag
e
q
u
ality
ass
ess
m
en
t
m
eth
o
d
wh
ich
co
m
p
u
tes
r
elativ
e
q
u
a
lity
s
co
r
es
b
etwe
en
a
r
ef
er
en
ce
r
ec
o
n
s
tr
u
cted
im
ag
e
[
3
0
]
.
Me
asu
r
em
e
n
t
s
at
d
if
f
er
en
t scale
s
ca
n
b
e
co
m
b
in
ed
to
o
b
tain
a
n
o
v
er
all
MS
-
SS
I
M
ev
alu
atio
n
,
as sh
o
wn
in
(
6
)
,
−
=
(
,
̂
)
.
∏
=
1
(
,
̂
)
(
,
̂
)
(
6
)
wh
er
e
an
d
̂
r
ep
r
esen
t
th
e
o
r
ig
i
n
al
im
ag
e
an
d
th
e
r
ec
o
n
s
tr
u
ct
ed
im
ag
e
r
esp
ec
tiv
ely
.
d
en
o
t
es
th
e
h
ig
h
est
s
ca
le,
(
,
̂
)
,
(
,
̂
)
,
(
,
̂
)
r
ef
er
to
th
e
lu
m
in
a
n
ce
,
co
n
tr
ast,
a
n
d
s
tr
u
ct
u
r
e
c
o
m
p
ar
is
o
n
s
at
th
e
j
-
th
s
ca
le,
r
esp
ec
tiv
ely
.
I
m
ag
e
co
m
p
r
es
s
io
n
ef
f
icien
c
y
is
ev
alu
ated
b
y
C
R
,
b
its
p
er
p
ix
el
(
b
p
p
)
,
a
n
d
s
p
ac
e
s
av
in
g
s
(
SS
s
)
.
T
h
e
C
R
an
d
b
p
p
g
iv
e
a
s
tr
aig
h
t
n
o
tio
n
o
f
co
m
p
r
ess
io
n
d
eg
r
ee
ass
o
ciate
d
with
th
e
am
o
u
n
t
o
f
d
ata
[
3
1
]
,
[
3
2
]
.
T
h
e
SS
s
is
an
o
th
er
m
etr
ic
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
co
m
p
r
ess
io
n
tech
n
iq
u
e,
it
in
d
icate
s
th
e
g
ain
ed
am
o
u
n
t
o
f
s
to
r
a
g
e
s
p
ac
e
f
r
o
m
s
av
in
g
th
e
co
m
p
r
ess
ed
d
ata
[
3
3
]
.
T
h
e
C
R
,
b
p
p
,
a
n
d
SS
s
ar
e
g
iv
en
in
(
7
)
-
(
9
)
,
r
esp
ec
tiv
ely
:
=
S
i
ze
of
un
co
m
p
r
es
s
ed
i
m
ag
e
S
i
ze
of
co
m
p
r
es
s
e
d
i
m
ag
e
(
7
)
=
ℎ
ℎ
(
8
)
=
(
1
−
)
×
100
(
9
)
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
A
mu
lti
-
s
ca
le
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
a
n
d
d
is
crete
w
a
ve
let
…
(
Da
lila
C
h
ikh
a
o
u
i
)
249
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
th
is
s
ec
tio
n
,
th
e
r
esu
lts
o
f
o
u
r
ex
p
e
r
im
en
ts
ar
e
p
r
esen
ted
an
d
d
is
cu
s
s
ed
,
s
h
o
wca
s
in
g
th
e
ef
f
icac
y
o
f
o
u
r
m
et
h
o
d
.
T
h
e
im
p
lem
en
tatio
n
o
f
th
e
e
x
p
er
im
en
ts
h
as
b
ee
n
ca
r
r
ied
o
u
t
o
n
th
e
NVI
DI
A
T
esla
K8
0
GPU
p
r
o
v
id
e
d
b
y
Go
o
g
le
C
o
lab
.
Ker
as
with
a
T
en
s
o
r
Flo
w
b
ac
k
en
d
ar
e
u
s
ed
to
b
u
ild
o
u
r
n
etwo
r
k
ar
ch
itectu
r
e
.
T
ab
le
3
p
r
o
v
i
d
es
a
s
u
m
m
ar
y
o
f
th
e
m
ain
s
im
u
latio
n
p
ar
a
m
eter
s
o
f
th
e
MS
-
C
NNs
th
at
wer
e
u
s
ed
in
th
e
ex
p
er
im
en
t.
Fo
r
o
p
tim
izatio
n
p
u
r
p
o
s
es
an
d
to
m
i
n
im
ize
th
e
lo
s
s
f
u
n
ctio
n
s
we
u
s
ed
th
e
A
d
am
o
p
tim
izer
[
3
4
]
.
T
h
e
lear
n
in
g
r
ate
is
in
itialized
b
y
1
.
0
E
−3
v
alu
e
an
d
r
ed
u
ce
d
u
s
in
g
a
lear
n
in
g
r
ate
s
ch
ed
u
ler
b
y
a
f
ac
to
r
o
f
2
b
ased
o
n
m
etr
ic
im
p
r
o
v
em
en
t.
T
h
e
n
etwo
r
k
s
wer
e
tr
ain
ed
f
o
r
4
0
0
ep
o
ch
s
an
d
f
in
e
-
tu
n
ed
f
o
r
a
n
o
th
e
r
5
0
ep
o
ch
s
.
T
ab
le
3
.
T
h
e
m
ain
s
im
u
latio
n
p
ar
am
eter
s
o
f
t
h
e
ex
p
e
r
im
en
t
P
a
r
a
me
t
e
r
V
a
l
u
e
Le
a
r
n
i
n
g
r
a
t
e
1
.
0
E−
3
t
o
1
.
0
E−
6
Ep
o
c
h
s
4
5
0
B
a
t
c
h
si
z
e
8
I
n
p
u
t
si
z
e
1
2
8
×
1
2
8
Fo
r
ex
p
er
im
en
tal
p
u
r
p
o
s
es,
we
u
tili
ze
d
r
etin
al
im
ag
es
f
r
o
m
two
p
u
b
licly
av
ailab
le
d
atasets
.
T
h
e
f
ir
s
t,
th
e
d
ig
ital
r
etin
al
i
m
ag
es
f
o
r
v
ess
el
ex
tr
ac
tio
n
(
DR
I
VE
)
d
atab
ase
[
3
5
]
,
c
o
n
ta
in
s
4
0
co
lo
r
f
u
n
d
u
s
im
ag
es
with
a
r
eso
lu
tio
n
o
f
5
6
5
×5
8
4
p
ix
els.
W
e
also
r
an
d
o
m
ly
s
elec
ted
4
0
im
ag
es
f
r
o
m
th
e
OR
I
GA
-
lig
h
t
d
atab
ase
[
3
6
]
,
w
h
ich
in
clu
d
es
6
5
0
h
ig
h
-
r
eso
lu
tio
n
im
ag
es
f
r
o
m
th
e
Sin
g
ap
o
r
e
Ma
lay
E
y
e
Stu
d
y
(
SiME
S).
T
o
p
r
e
v
en
t
m
em
o
r
y
o
v
er
f
lo
w
an
d
o
p
tim
ize
th
e
im
ag
e
c
o
m
p
r
ess
io
n
m
o
d
el,
ea
c
h
im
a
g
e
was
cr
o
p
p
ed
to
p
atch
es
o
f
1
2
8
×1
2
8
p
ix
els
an
d
n
o
r
m
alize
d
b
e
f
o
r
e
c
o
m
p
r
es
s
io
n
.
T
h
e
d
ataset
wa
s
th
en
d
i
v
id
ed
in
to
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
in
g
s
ets,
with
8
0
% f
o
r
tr
ain
in
g
an
d
2
0
% f
o
r
v
alid
atio
n
an
d
test
in
g
.
T
est
d
ataset
im
ag
es
wer
e
u
til
ized
to
ev
alu
ate
th
e
p
r
o
p
o
s
ed
co
m
p
r
ess
io
n
m
eth
o
d
b
y
co
m
p
ar
in
g
it
with
J
PEG,
J
PEG2
0
0
0
,
an
d
ex
is
tin
g
d
ee
p
lear
n
in
g
-
b
ased
m
eth
o
d
s
.
J
PEG
an
d
J
PEG2
0
0
0
wer
e
s
elec
ted
b
ec
au
s
e
o
f
th
eir
tr
a
n
s
f
o
r
m
r
elian
ce
,
with
J
PEG2
0
0
0
u
s
in
g
th
e
C
DF
9
/7
wav
elet
also
em
p
lo
y
ed
in
o
u
r
c
o
d
ec
.
Am
o
n
g
t
h
e
d
ee
p
lear
n
in
g
m
eth
o
d
s
,
B
allé
et
a
l.
[
1
6
]
was
ch
o
s
en
f
o
r
its
s
tate
-
of
-
th
e
-
a
r
t
s
tatu
s
an
d
l
o
wer
co
m
p
lex
ity
co
m
p
ar
e
d
to
m
o
d
els
s
u
ch
in
[
1
7
]
.
T
o
f
u
r
th
er
e
v
alu
ate
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
in
teg
r
ated
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
lay
e
r
s
in
o
u
r
MS
-
C
NN
ar
ch
itectu
r
e,
we
co
n
d
u
cte
d
a
co
m
p
a
r
is
o
n
ex
p
e
r
im
en
t
in
wh
ich
we
tr
ain
ed
an
o
th
er
a
r
ch
itectu
r
e
b
ased
o
n
C
NN
with
o
u
t
th
e
m
u
lti
-
s
ca
le
co
n
v
o
l
u
tio
n
s
th
at
wer
e
r
ep
lace
d
b
y
s
eq
u
en
tially
s
tack
ed
s
im
p
le
co
n
v
o
lu
tio
n
s
in
th
e
ar
c
h
itectu
r
e
b
ased
o
n
C
NN.
Un
d
er
h
ig
h
CR
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
’
s
r
ec
o
n
s
tr
u
ctio
n
was
e
v
alu
ated
,
with
r
esu
lts
s
h
o
wn
i
n
T
ab
le
4
.
T
h
e
av
er
ag
e
MSE
,
PS
NR
,
C
R
,
an
d
SS
s
v
alu
es
o
f
th
e
im
ag
e
p
atch
es
wer
e
6
.
5
1
,
4
0
.
8
6
d
B
,
5
3
.
0
3
,
an
d
9
7
.
9
5
%,
r
esp
ec
tiv
ely
.
Desp
ite
h
ig
h
co
m
p
r
ess
io
n
,
o
u
r
m
eth
o
d
ac
h
ie
v
ed
a
h
ig
h
PS
NR
,
in
d
icatin
g
g
o
o
d
r
etin
al
im
ag
e
r
ec
o
n
s
tr
u
ctio
n
q
u
ality
.
Ad
d
iti
o
n
ally
,
a
s
p
ac
e
-
s
av
in
g
p
er
ce
n
tag
e
n
ea
r
1
0
0
%
d
em
o
n
s
tr
at
es
th
e
co
m
p
r
ess
io
n
ef
f
icien
cy
f
o
r
s
y
s
tem
s
r
eq
u
ir
in
g
m
ed
ical
d
ata
s
to
r
a
g
e.
T
ab
le
4
.
T
h
e
p
er
f
o
r
m
a
n
ce
ev
a
lu
atio
n
o
f
t
h
e
p
r
o
p
o
s
ed
c
o
m
p
r
ess
io
n
m
eth
o
d
o
n
r
etin
a
im
a
g
e
p
atch
es
I
mag
e
p
a
t
c
h
e
s
M
e
a
su
r
e
M
S
E
P
S
N
R
(
d
b
)
CR
S
S
s (%)
P
a
t
c
h
e
1
2
.
0
5
4
5
.
0
2
6
3
.
8
9
8
.
4
4
P
a
t
c
h
e
2
9
.
3
3
8
.
4
5
5
6
.
1
2
9
8
.
2
2
P
a
t
c
h
e
3
8
.
3
8
3
8
.
9
5
2
.
1
3
9
8
.
0
8
P
a
t
c
h
e
4
1
0
.
1
7
3
8
.
0
6
2
9
.
3
7
9
6
.
5
9
P
a
t
c
h
e
5
2
.
6
7
4
3
.
8
7
6
3
.
7
2
9
8
.
4
3
M
e
a
n
6
.
5
1
4
0
.
8
6
5
3
.
0
3
9
7
.
9
5
W
e
co
n
d
u
cted
ex
p
er
im
e
n
tal
co
m
p
ar
is
o
n
s
to
ass
ess
th
e
p
r
o
p
o
s
ed
m
eth
o
d
’
s
p
er
f
o
r
m
a
n
ce
ag
ain
s
t o
th
er
co
m
p
r
ess
io
n
tech
n
iq
u
es
as
d
etailed
in
tab
les
b
elo
w.
T
ab
les
5
an
d
6
d
em
o
n
s
tr
ate
th
at
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
o
f
co
m
p
r
ess
in
g
m
ed
ical
im
a
g
es is
m
o
r
e
ef
f
icien
t th
a
n
th
e
o
t
h
er
m
eth
o
d
s
,
as e
v
id
en
ce
d
b
y
its
ac
h
iev
ed
m
ax
i
m
u
m
PS
NR
(
d
B
)
an
d
MS
-
SS
I
M,
a
n
d
th
e
lo
w
d
is
to
r
tio
n
in
MSE
.
At
C
R
=5
0
,
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
o
u
t
p
er
f
o
r
m
ed
s
tan
d
ar
d
m
eth
o
d
s
(
J
PEG,
J
PEG
2
0
0
0
)
an
d
th
e
d
ee
p
lear
n
in
g
-
b
ased
a
p
p
r
o
ac
h
B
allé
et
a
l.
[
1
6
]
.
T
ab
le
5
s
h
o
ws
o
u
r
m
eth
o
d
ac
h
iev
ed
s
u
p
e
r
io
r
r
esu
lts
,
with
a
6
.
8
9
MSE
,
2
.
3
2
d
B
h
ig
h
er
PS
NR
,
an
d
0
.
8
%
h
ig
h
er
MS
-
SS
I
M
co
m
p
ar
ed
to
B
allé
et
a
l.
[1
6]
,
wh
ich
its
elf
s
u
r
p
ass
ed
J
PEG.
C
o
m
p
ar
ed
to
J
PEG2
0
0
0
,
o
u
r
m
eth
o
d
s
h
o
wed
g
ain
s
o
f
2
.
1
1
d
B
in
PS
N
R
an
d
0
.
7
%
in
MS
-
SS
I
M.
Ad
d
itio
n
ally
,
th
e
MS
-
C
NN
ap
p
r
o
ac
h
s
lig
h
tly
o
u
tp
er
f
o
r
m
ed
o
u
r
m
eth
o
d
(
wi
th
o
u
t
MSC
B
)
,
1
.
4
5
d
B
PS
NR
g
ain
,
an
d
0
.
7
%
MS
-
SS
I
M
g
a
in
.
At
C
R
=8
0
,
o
u
r
m
eth
o
d
m
ain
tain
ed
r
o
b
u
s
t
p
e
r
f
o
r
m
a
n
ce
co
m
p
ar
ed
to
o
th
er
s
.
T
ab
le
6
r
ep
o
r
ts
a
2
.
6
d
B
PS
NR
an
d
0
.
7
9
%
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
8
,
No
.
1
,
Ap
r
il
20
2
5
:
2
4
3
-
2
5
3
250
MS
-
SS
I
M
g
ain
o
v
er
B
allé
et
a
l.
[
1
6
]
,
an
d
3
.
0
3
d
B
PS
NR
a
n
d
1
.
6
%
MS
-
SS
I
M
im
p
r
o
v
em
en
t
o
v
er
J
PEG2
0
0
0
.
T
h
e
MS
-
C
NN
m
eth
o
d
also
o
u
tp
er
f
o
r
m
e
d
o
u
r
m
eth
o
d
(
with
o
u
t
MSC
B
)
,
with
1
.
8
9
d
B
PS
N
R
g
ain
,
an
d
1
.
0
9
%
MS
-
SS
I
M
g
ain
.
T
h
is
an
aly
s
is
co
n
f
ir
m
s
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
’
s
ef
f
ec
tiv
e
n
ess
in
p
r
eser
v
i
n
g
im
ag
e
q
u
ality
i
n
ter
m
s
o
f
PS
NR
an
d
MS
-
SS
I
M
m
etr
ics,
esp
ec
ially
at
h
ig
h
CR
(
C
R
=5
0
,
C
R
=8
0
)
.
T
ab
le
5
.
Per
f
o
r
m
an
ce
ev
alu
ati
o
n
o
f
d
if
f
er
en
t m
eth
o
d
s
in
ter
m
s
o
f
r
ec
o
n
s
tr
u
cted
im
ag
e
q
u
a
lity
at
C
R
≈
5
0
M
e
t
h
o
d
A
v
e
r
a
g
e
d
m
e
a
s
u
r
e
s
M
S
E
P
S
N
R
(
d
B
)
MS
-
S
S
I
M
(
%)
O
u
r
me
t
h
o
d
6
.
8
9
3
9
.
7
5
9
7
.
4
2
B
a
l
l
é
e
t
a
l
.
[1
6]
1
1
.
7
4
3
7
.
4
3
9
6
.
6
3
JP
EG
2
0
0
0
1
1
.
2
3
3
7
.
6
4
9
6
.
7
2
JP
EG
3
9
.
6
6
3
2
.
1
5
8
8
.
2
7
O
u
r
me
t
h
o
d
(
w
i
t
h
o
u
t
M
S
C
B
)
1
0
.
0
8
3
8
.
3
9
6
.
6
8
T
ab
le
6
.
Per
f
o
r
m
an
ce
ev
alu
ati
o
n
o
f
d
if
f
er
en
t m
eth
o
d
s
in
ter
m
s
o
f
r
ec
o
n
s
tr
u
cted
im
ag
e
q
u
a
lity
at
C
R
≈
8
0
M
e
t
h
o
d
A
v
e
r
a
g
e
d
m
e
a
s
u
r
e
s
M
S
E
P
S
N
R
(
d
B
)
MS
-
S
S
I
M
(
%)
O
u
r
me
t
h
o
d
8
.
2
5
3
8
.
9
7
9
6
.
7
5
B
a
l
l
é
e
t
a
l
.
[1
6]
1
5
.
0
2
3
6
.
3
7
9
5
.
8
JP
EG
2
0
0
0
1
6
.
7
2
3
5
.
9
4
9
5
.
1
5
JP
EG
2
1
8
.
5
7
2
4
.
7
7
76
O
u
r
me
t
h
o
d
(
w
i
t
h
o
u
t
M
S
C
B
)
1
2
.
9
3
7
.
0
8
9
5
.
6
6
T
h
e
v
is
u
al
q
u
ality
c
o
m
p
ar
is
o
n
s
at
0
.
1
6
b
p
p
a
r
e
p
r
o
v
id
e
d
i
n
Fig
u
r
e
4
.
I
t
ca
n
b
e
o
b
s
er
v
e
d
th
at
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
attain
e
d
b
ett
er
s
u
b
jectiv
e
q
u
ality
o
f
r
ec
o
n
s
tr
u
cted
im
ag
es
in
Fig
u
r
e
s
4(
a)
-
4(
b
)
,
co
m
p
a
r
ed
to
B
allé
et
a
l.
[
1
6
]
,
J
PEG2
0
0
,
an
d
J
PEG
s
h
o
wn
in
Fig
u
r
e
4
(
c)
,
Fig
u
r
e
4
(
d
)
,
an
d
Fig
u
r
e
4
(
e)
,
r
esp
ec
tiv
ely
.
I
n
co
m
p
ar
is
o
n
,
J
PEG
m
ar
k
ed
a
lo
w
p
er
f
o
r
m
a
n
ce
with
o
b
v
i
o
u
s
b
lo
ck
i
n
g
ar
tifa
cts
in
th
e
r
ec
o
n
s
tr
u
cted
r
etin
a
l
im
ag
es sh
o
wn
in
Fig
u
r
e
4
(
e)
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
4
.
Vis
u
al
q
u
ality
c
o
m
p
ar
is
o
n
b
etwe
en
th
e
o
r
ig
in
al
a
n
d
r
ec
o
n
s
tr
u
cte
d
im
ag
es u
s
in
g
d
if
f
er
en
t m
et
h
o
d
s
at
0
.
1
6
b
p
p
:
(
a)
o
r
ig
in
al
r
etin
al
im
ag
es a
n
d
th
eir
zo
o
m
e
d
-
in
p
atch
es,
(
b
)
o
u
r
m
et
h
o
d
,
(
c)
B
allé
et
a
l.
[1
6]
,
(
d
)
J
PEG2
0
0
,
a
n
d
(
e)
J
PEG
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
A
mu
lti
-
s
ca
le
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
a
n
d
d
is
crete
w
a
ve
let
…
(
Da
lila
C
h
ikh
a
o
u
i
)
251
W
e
p
r
o
p
o
s
ed
a
n
ef
f
icien
t
co
m
p
r
ess
io
n
tech
n
i
q
u
e
u
tili
zin
g
two
MS
-
C
NNs
an
d
DW
T
co
d
e.
T
h
e
co
m
p
r
ess
io
n
p
er
f
o
r
m
an
c
e
was
ev
alu
ated
u
s
in
g
v
a
r
io
u
s
q
u
ality
m
etr
ics.
Fo
r
ex
a
m
p
le,
at
C
R
=8
0
o
u
r
m
eth
o
d
ac
h
iev
ed
th
e
h
ig
h
e
s
t
PS
N
R
an
d
MS
-
SS
I
M
v
alu
es
o
f
3
8
.
9
8
d
B
an
d
9
6
.
8
6
%,
r
esp
ec
tiv
ely
,
o
u
tp
er
f
o
r
m
in
g
e
v
en
s
tan
d
ar
d
co
m
p
r
ess
io
n
m
eth
o
d
s
.
Par
ticu
lar
ly
,
th
e
J
PEG2
0
0
0
wh
ich
is
b
ased
o
n
th
e
s
am
e
C
DF
9
/7
wav
elet
tr
an
s
f
o
r
m
u
s
ed
in
o
u
r
DW
T
b
ased
co
d
ec
.
Alth
o
u
g
h
in
o
u
r
tech
n
i
q
u
e
d
ee
p
lear
n
in
g
m
o
d
el
M
S
-
C
NN
allo
wed
an
en
h
an
c
ed
co
m
p
r
ess
io
n
p
er
f
o
r
m
an
ce
o
f
DW
T
co
d
ec
,
it
is
im
p
o
r
ta
n
t
to
n
o
te
th
at
o
u
r
co
d
ec
is
b
ased
o
n
a
b
asic
co
m
p
r
ess
io
n
s
ch
em
e
with
a
s
tr
aig
h
tf
o
r
war
d
en
tr
o
p
y
c
o
d
in
g
tech
n
iq
u
e,
c
o
m
p
ar
e
d
to
th
e
J
PEG2
0
0
0
’
s
s
o
p
h
is
ticated
s
ch
em
e
an
d
en
tr
o
p
y
co
d
i
n
g
tech
n
i
q
u
e
[
7
]
,
[
3
7
]
.
I
n
co
n
tr
ast,
th
e
lo
w
p
er
f
o
r
m
an
ce
o
f
J
PEG
is
b
asically
d
u
e
to
th
e
DC
T
,
wh
ic
h
is
less
ef
f
ec
tiv
e
th
an
th
e
wav
elet
tr
an
s
f
o
r
m
s
.
I
n
co
m
p
a
r
is
o
n
to
d
ee
p
lear
n
i
n
g
-
b
ased
m
eth
o
d
s
,
we
p
r
esen
t
in
T
ab
le
7
,
a
b
r
ief
an
d
av
e
r
ag
ed
ex
p
er
im
e
n
tal
r
esu
lts
co
m
p
ar
is
o
n
th
at
in
clu
d
es
o
n
ly
lear
n
in
g
b
ased
co
m
p
r
e
s
s
io
n
m
o
d
els.
T
a
b
le
7
c
o
m
p
ar
es
o
u
r
m
eth
o
d
wit
h
o
th
er
lear
n
i
n
g
-
b
ased
co
m
p
r
ess
io
n
m
o
d
els,
s
h
o
win
g
th
at
o
u
r
m
eth
o
d
ac
h
ie
v
ed
th
e
h
ig
h
est
PS
N
R
with
th
e
f
ewe
s
t
p
ar
am
eter
s
allo
win
g
l
o
w
co
m
p
u
tatio
n
al
co
m
p
lex
ity
.
Alth
o
u
g
h
C
h
en
g
et
a
l.
[
1
7
]
ac
h
iev
ed
a
0
.
0
0
2
h
ig
h
er
MS
-
SS
I
M,
it
r
eq
u
i
r
ed
1
1
.
6
m
illi
o
n
p
ar
am
eter
s
in
th
e
ar
ch
itectu
r
e.
Ou
r
m
eth
o
d
is
also
5
.
2
an
d
1
2
.
1
tim
es
f
aster
th
an
B
allé
et
a
l.
[
1
6
]
an
d
C
h
en
g
et
a
l.
[
1
7
]
,
r
es
p
ec
tiv
ely
.
T
h
u
s
,
it
ca
n
b
e
s
aid
th
at
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
ac
h
iev
es
b
etter
co
d
i
n
g
p
e
r
f
o
r
m
an
ce
a
n
d
r
ec
o
n
s
t
r
u
ctio
n
q
u
ality
with
lo
w
d
e
ep
lear
n
in
g
m
o
d
el
co
m
p
lex
ity
a
n
d
co
m
p
u
tatio
n
tim
e
T
ab
le
7
.
T
h
e
co
m
p
ar
ativ
e
r
esu
lts
o
f
ex
is
tin
g
lear
n
in
g
b
ased
m
eth
o
d
s
in
ter
m
s
PS
NR
,
MS
-
SS
I
M,
n
u
m
b
er
o
f
tr
ai
n
ab
le
p
a
r
am
eter
s
an
d
co
m
p
u
tatio
n
tim
e
at
C
R
=8
0
M
e
t
h
o
d
P
S
N
R
(
d
B
)
MS
-
S
S
I
M
(
%)
N
u
mb
e
r
s
o
f
p
a
r
a
m
e
t
e
r
s
Ex
e
c
u
t
i
o
n
t
i
m
e
(
s)
O
u
r
me
t
h
o
d
3
8
.
9
8
9
6
.
8
6
7
8
2
6
9
4
0
.
2
5
B
a
l
l
é
e
t
a
l
.
[
1
6
]
3
6
.
3
8
9
5
.
8
2
5
8
2
5
3
1
1
.
3
C
h
e
n
g
e
t
a
l
.
[
1
7
]
3
8
.
9
0
97
11
6
2
7
9
1
6
3
.
0
3
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
h
as
d
e
m
o
n
s
tr
ated
ex
ce
p
tio
n
al
p
er
f
o
r
m
an
ce
with
p
r
o
m
is
in
g
r
esu
lts
,
ev
en
w
h
en
co
m
p
ar
ed
to
co
m
p
lex
d
ee
p
l
ea
r
n
in
g
-
b
ased
m
et
h
o
d
s
t
h
at
h
av
e
a
g
r
ea
ter
n
u
m
b
e
r
o
f
tr
ain
ab
le
p
a
r
am
eter
s
.
T
h
e
im
p
licatio
n
o
f
t
h
ese
r
esu
lts
is
th
at
it
o
f
f
er
s
a
s
u
p
e
r
io
r
s
o
lu
tio
n
f
o
r
im
a
g
e
d
ata
r
ed
u
cti
o
n
,
p
a
r
ticu
lar
ly
i
n
s
ce
n
ar
io
s
r
eq
u
ir
in
g
h
i
g
h
CR
.
T
h
is
ad
v
an
ce
m
e
n
t
co
u
l
d
s
ig
n
i
f
ican
tly
im
p
ac
t v
ar
io
u
s
f
ield
s
t
h
at
r
ely
o
n
ef
f
icien
t
im
ag
e
s
to
r
ag
e
an
d
tr
an
s
m
is
s
i
o
n
,
s
u
ch
as
m
ed
ical
im
ag
in
g
,
s
atellite
co
m
m
u
n
icatio
n
s
,
an
d
d
ig
it
al
ar
ch
iv
in
g
.
T
h
e
m
eth
o
d
’
s
a
b
ilit
y
to
m
ai
n
tain
g
o
o
d
im
ag
e
q
u
ality
ev
en
at
h
ig
h
co
m
p
r
ess
io
n
lev
el
s
s
u
g
g
ests
p
o
ten
tial
ap
p
licatio
n
s
in
b
an
d
wid
th
-
c
o
n
s
tr
ain
ed
en
v
ir
o
n
m
en
ts
o
r
s
y
s
tem
s
with
lim
i
ted
s
to
r
ag
e
ca
p
ac
ity
.
Fu
r
th
er
m
o
r
e
,
th
is
d
ev
elo
p
m
e
n
t
m
a
y
lead
to
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
in
r
ea
l
-
tim
e
im
ag
e
p
r
o
ce
s
s
in
g
a
p
p
li
ca
tio
n
s
,
wh
er
e
d
ata
s
ize,
q
u
ality
an
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
ar
e
cr
itical
f
ac
to
r
s
.
I
n
th
e
f
u
tu
r
e,
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
ca
n
b
e
en
h
a
n
ce
d
.
Fir
s
t,
b
y
d
esig
n
in
g
b
etter
d
ee
p
l
ea
r
n
in
g
a
r
ch
ite
ctu
r
es
an
d
o
p
tim
izatio
n
s
tr
ateg
ies
f
o
r
im
ag
e
co
m
p
r
ess
io
n
task
s
.
Fu
r
th
er
m
o
r
e,
t
h
e
tech
n
iq
u
e
’
s
co
m
p
atib
ilit
y
with
co
n
v
en
tio
n
al
co
d
ec
s
,
d
u
e
t
o
its
u
s
e
o
f
a
DW
T
-
b
ased
c
o
d
ec
,
o
p
en
s
u
p
p
o
s
s
ib
ilit
ies
f
o
r
o
th
er
c
o
n
v
e
n
tio
n
al
co
d
ec
s
to
b
e
u
s
ed
in
co
n
ju
n
ctio
n
with
it.
4.
CO
NCLU
SI
O
N
I
n
th
is
p
a
p
er
,
we
h
av
e
i
n
tr
o
d
u
ce
d
a
r
etin
al
im
ag
e
co
m
p
r
e
s
s
io
n
m
eth
o
d
b
ased
o
n
MS
-
C
NNs
an
d
DW
T
.
T
o
ac
h
iev
e
b
etter
im
ag
e
q
u
ality
at
a
h
ig
h
CR
,
two
MS
-
C
NN
s
wer
e
co
n
n
ec
ted
to
g
eth
er
,
th
e
en
co
d
in
g
MS
-
C
NN
i
s
em
p
lo
y
ed
to
g
e
n
er
ate
in
ter
m
ed
iate
c
o
m
p
ac
t
r
ep
r
esen
tatio
n
,
wh
ich
m
ain
t
ain
s
th
e
s
tr
u
ctu
r
al
in
f
o
r
m
atio
n
o
f
th
e
o
r
i
g
in
al
im
ag
e
th
at
will
b
e
co
d
ed
b
y
DW
T
.
Nex
t,
th
e
MS
-
C
NN
e
n
ab
les
h
ig
h
-
q
u
ality
r
ec
o
n
s
tr
u
ctio
n
a
n
d
r
etr
ie
v
al
o
f
th
e
o
r
ig
in
al
im
a
g
e
at
th
e
d
ec
o
d
in
g
s
id
e.
T
h
e
o
b
tain
e
d
e
x
p
er
im
en
tal
r
esu
lts
co
n
f
ir
m
e
d
th
e
s
u
p
er
io
r
ity
o
f
th
e
p
r
o
p
o
s
ed
c
o
m
p
r
ess
io
n
m
eth
o
d
b
ased
o
n
d
if
f
e
r
en
t
p
er
f
o
r
m
a
n
ce
m
etr
ics.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
attain
ed
h
ig
h
er
CR
(
C
R
=8
0
)
wh
ile
m
ain
tain
in
g
an
ac
ce
p
ta
b
le
r
etin
al
im
ag
e
q
u
alit
y
with
an
av
e
r
ag
e
PS
NR
v
alu
e
o
f
3
8
.
9
8
d
B
a
n
d
MS
-
SS
I
M
o
f
9
6
.
8
%.
Hen
ce
,
c
o
n
tr
ib
u
tio
n
to
m
in
im
izin
g
th
e
d
ata
s
ize
an
d
s
av
in
g
s
to
r
a
g
e
a
n
d
tr
a
n
s
m
is
s
io
n
r
eso
u
r
ce
s
wh
i
le
m
ain
tain
in
g
v
is
u
al
q
u
ality
o
f
m
e
d
ical
im
ag
es.
Fu
r
th
er
m
o
r
e
,
o
u
r
p
r
o
p
o
s
ed
te
ch
n
iq
u
e
e
x
h
ib
its
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
m
ak
i
n
g
it
ap
p
li
ca
b
le
in
r
ea
l
-
tim
e
m
ed
ical
im
ag
e
ap
p
licatio
n
s
.
RE
F
E
R
E
NC
E
S
[
1
]
R
.
V
i
j
a
n
d
S
.
A
r
o
r
a
,
“
A
s
y
st
e
ma
t
i
c
su
r
v
e
y
o
f
a
d
v
a
n
c
e
s
i
n
r
e
t
i
n
a
l
i
m
a
g
i
n
g
m
o
d
a
l
i
t
i
e
s
f
o
r
A
l
z
h
e
i
m
e
r
’
s
d
i
sea
se
d
i
a
g
n
o
si
s
,
”
Me
t
a
b
o
l
i
c
Br
a
i
n
D
i
se
a
se
,
v
o
l
.
3
7
,
n
o
.
7
,
p
p
.
2
2
1
3
–
2
2
4
3
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
1
1
-
0
2
2
-
0
0
9
2
7
-
4.
[
2
]
X
.
H
e
,
Y
.
D
e
n
g
,
L.
F
a
n
g
,
a
n
d
Q
.
P
e
n
g
,
“
M
u
l
t
i
-
m
o
d
a
l
r
e
t
i
n
a
l
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
mo
d
a
l
i
t
y
-
sp
e
c
i
f
i
c
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
M
e
d
i
c
a
l
I
m
a
g
i
n
g
,
v
o
l
.
4
0
,
n
o
.
6
,
p
p
.
1
5
9
1
–
1
6
0
2
,
J
u
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
TM
I
.
2
0
2
1
.
3
0
5
9
9
5
6
.
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
8
,
No
.
1
,
Ap
r
il
20
2
5
:
2
4
3
-
2
5
3
252
[
3
]
M
.
Z.
A
t
w
a
n
y
,
A
.
H
.
S
a
h
y
o
u
n
,
a
n
d
M
.
Y
a
q
u
b
,
“
D
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
f
o
r
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
c
l
a
s
si
f
i
c
a
t
i
o
n
:
a
s
u
r
v
e
y
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
2
8
6
4
2
–
2
8
6
5
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
5
7
6
3
2
.
[
4
]
Y
.
Y
u
a
n
,
L.
Zh
a
n
g
,
L.
W
a
n
g
,
a
n
d
H
.
H
u
a
n
g
,
“
M
u
l
t
i
-
l
e
v
e
l
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
f
o
r
r
e
t
i
n
a
l
v
e
sse
l
s
e
g
m
e
n
t
a
t
i
o
n
,
”
I
EEE
J
o
u
rn
a
l
o
f
Bi
o
m
e
d
i
c
a
l
a
n
d
H
e
a
l
t
h
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
2
6
,
n
o
.
1
,
p
p
.
3
1
2
–
3
2
3
,
J
a
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
J
B
H
I
.
2
0
2
1
.
3
0
8
9
2
0
1
.
[
5
]
A
.
C
h
a
a
b
o
u
n
i
,
Y
.
G
a
u
d
e
a
u
,
J.
L
a
mb
e
r
t
,
J.
M
.
M
o
u
r
e
a
u
x
,
a
n
d
P
.
G
a
l
l
e
t
,
“
H
.
2
6
4
me
d
i
c
a
l
v
i
d
e
o
c
o
m
p
r
e
ss
i
o
n
f
o
r
t
e
l
e
m
e
d
i
c
i
n
e
:
Ap
e
r
f
o
r
m
a
n
c
e
a
n
a
l
y
si
s,”
I
rb
m
,
v
o
l
.
3
7
,
n
o
.
1
,
p
p
.
4
0
–
4
8
,
F
e
b
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
r
b
m
.
2
0
1
5
.
0
9
.
0
0
7
.
[
6
]
X
.
L
i
u
e
t
a
l
.
,
“
M
e
d
i
c
a
l
i
ma
g
e
c
o
m
p
r
e
ssi
o
n
b
a
se
d
o
n
v
a
r
i
a
t
i
o
n
a
l
a
u
t
o
e
n
c
o
d
e
r
,
”
Ma
t
h
e
m
a
t
i
c
a
l
Pro
b
l
e
m
s
i
n
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
0
2
2
,
p
p
.
1
–
1
2
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
7
0
8
8
1
3
7
.
[
7
]
D
.
S
.
Ta
u
b
ma
n
a
n
d
M
.
W
.
M
a
r
c
e
l
l
i
n
,
J
PEG
2
0
0
0
I
m
a
g
e
C
o
m
p
re
ssi
o
n
F
u
n
d
a
m
e
n
t
a
l
s
,
S
t
a
n
d
a
rd
s
a
n
d
Pr
a
c
t
i
c
e
.
B
o
s
t
o
n
,
M
A
:
S
p
r
i
n
g
e
r
U
S
,
2
0
0
2
.
[
8
]
S
.
P
.
R
a
j
a
,
“
W
a
v
e
l
e
t
-
b
a
se
d
i
m
a
g
e
c
o
mp
r
e
ss
i
o
n
e
n
c
o
d
i
n
g
t
e
c
h
n
i
q
u
e
s
-
a
c
o
mp
l
e
t
e
p
e
r
f
o
r
ma
n
c
e
a
n
a
l
y
si
s
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
m
a
g
e
a
n
d
G
r
a
p
h
i
c
s
,
v
o
l
.
2
0
,
n
o
.
2
,
p
.
2
0
5
0
0
0
8
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
4
2
/
S
0
2
1
9
4
6
7
8
2
0
5
0
0
0
8
4
.
[
9
]
P
.
H
ä
n
s
g
e
n
,
P
.
E.
U
n
d
r
i
l
l
,
a
n
d
M
.
J.
C
r
e
e
,
“
T
h
e
a
p
p
l
i
c
a
t
i
o
n
o
f
w
a
v
e
l
e
t
s
t
o
r
e
t
i
n
a
l
i
m
a
g
e
c
o
mp
r
e
ssi
o
n
a
n
d
i
t
s
e
f
f
e
c
t
o
n
a
u
t
o
mat
i
c
mi
c
r
o
a
n
e
u
r
y
sm
a
n
a
l
y
si
s
,
”
C
o
m
p
u
t
e
r
Me
t
h
o
d
s
a
n
d
Pr
o
g
ra
m
s
i
n
Bi
o
m
e
d
i
c
i
n
e
,
v
o
l
.
5
6
,
n
o
.
1
,
p
p
.
1
–
1
0
,
A
p
r
.
1
9
9
8
,
d
o
i
:
1
0
.
1
0
1
6
/
S
0
1
6
9
-
2
6
0
7
(
9
8
)
0
0
0
0
6
-
6.
[
1
0
]
R
.
H
.
E
i
k
e
l
b
o
o
m
e
t
a
l
.
,
“
M
e
t
h
o
d
s
a
n
d
l
i
m
i
t
s
o
f
d
i
g
i
t
a
l
i
m
a
g
e
c
o
m
p
r
e
ssi
o
n
o
f
r
e
t
i
n
a
l
i
m
a
g
e
s
f
o
r
t
e
l
e
me
d
i
c
i
n
e
,
”
I
n
v
e
s
t
i
g
a
t
i
v
e
O
p
h
t
h
a
l
m
o
l
o
g
y
a
n
d
V
i
su
a
l
S
c
i
e
n
c
e
,
v
o
l
.
4
1
,
n
o
.
7
,
p
p
.
1
9
1
6
–
1
9
2
4
,
2
0
0
0
.
[
1
1
]
S
.
K
r
i
v
e
n
k
o
,
V
.
Lu
k
i
n
,
O
.
K
r
y
l
o
v
a
,
a
n
d
V
.
S
h
u
t
k
o
,
“
V
i
s
u
a
l
l
y
l
o
ss
l
e
ss
c
o
m
p
r
e
ss
i
o
n
o
f
r
e
t
i
n
a
i
ma
g
e
s
,
”
i
n
2
0
1
8
I
EEE
3
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
El
e
c
t
r
o
n
i
c
s
a
n
d
N
a
n
o
t
e
c
h
n
o
l
o
g
y
,
EL
N
A
N
O
2
0
1
8
-
Pr
o
c
e
e
d
i
n
g
s
,
A
p
r
.
2
0
1
8
,
p
p
.
2
5
5
–
2
6
0
,
d
o
i
:
1
0
.
1
1
0
9
/
ELN
A
N
O
.
2
0
1
8
.
8
4
7
7
4
5
9
.
[
1
2
]
M
.
R
.
K
.
M
o
o
k
i
a
h
,
S
.
H
o
g
g
,
T
.
M
a
c
G
i
l
l
i
v
r
a
y
,
a
n
d
E.
Tr
u
c
c
o
,
“
O
n
t
h
e
q
u
a
n
t
i
t
a
t
i
v
e
e
f
f
e
c
t
s o
f
c
o
m
p
r
e
ss
i
o
n
o
f
r
e
t
i
n
a
l
f
u
n
d
u
s
i
ma
g
e
s
o
n
mo
r
p
h
o
m
e
t
r
i
c
v
a
s
c
u
l
a
r
mea
s
u
r
e
m
e
n
t
s
i
n
V
A
M
P
I
R
E,
”
C
o
m
p
u
t
e
r
m
e
t
h
o
d
s
a
n
d
p
r
o
g
r
a
m
s
i
n
b
i
o
m
e
d
i
c
i
n
e
,
v
o
l
.
2
0
2
,
p
.
1
0
5
9
6
9
,
A
p
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
m
p
b
.
2
0
2
1
.
1
0
5
9
6
9
.
[
1
3
]
P
.
N
a
z
a
r
i
a
n
d
H
.
P
o
u
r
g
h
a
ssem
,
“
A
n
a
u
t
o
m
a
t
e
d
v
e
ss
e
l
se
g
m
e
n
t
a
t
i
o
n
a
l
g
o
r
i
t
h
m
i
n
r
e
t
i
n
a
l
i
ma
g
e
s
u
si
n
g
2
D
G
a
b
o
r
w
a
v
e
l
e
t
,
”
i
n
I
ra
n
i
a
n
C
o
n
f
e
re
n
c
e
o
n
Ma
c
h
i
n
e
Vi
s
i
o
n
a
n
d
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
MV
I
P
,
S
e
p
.
2
0
1
3
,
p
p
.
1
4
5
–
1
4
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
r
a
n
i
a
n
M
V
I
P
.
2
0
1
3
.
6
7
7
9
9
6
7
.
[
1
4
]
M
.
Ja
v
e
d
,
P
.
N
a
g
a
b
h
u
sh
a
n
,
B
.
B
.
C
h
a
u
d
h
u
r
i
,
a
n
d
S
.
K
.
S
i
n
g
h
,
“
E
d
g
e
b
a
sed
e
n
h
a
n
c
e
m
e
n
t
o
f
r
e
t
i
n
a
l
i
m
a
g
e
s
u
si
n
g
a
n
e
f
f
i
c
i
e
n
t
JP
EG
-
c
o
m
p
r
e
sse
d
d
o
ma
i
n
t
e
c
h
n
i
q
u
e
,
”
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
a
n
d
F
u
zzy
S
y
s
t
e
m
s
,
v
o
l
.
3
6
,
n
o
.
1
,
p
p
.
5
4
1
–
5
5
6
,
F
e
b
.
2
0
1
9
,
d
o
i
:
1
0
.
3
2
3
3
/
JI
F
S
-
1
8
8
5
9
.
[
1
5
]
N
.
D
.
S
a
l
i
h
,
A
.
A
b
i
d
,
a
n
d
C
.
Es
w
a
r
a
n
,
“
Ef
f
i
c
i
e
n
t
r
e
t
i
n
a
l
i
m
a
g
e
c
o
mp
r
e
ss
i
o
n
b
a
se
d
o
n
m
o
d
i
f
i
e
d
h
u
f
f
man
a
l
g
o
r
i
t
h
m,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
n
g
i
n
e
e
r
i
n
g
Re
se
a
rc
h
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
2
,
n
o
.
7
,
p
p
.
9
4
2
–
9
4
8
,
2
0
1
9
.
[
1
6
]
J.
B
a
l
l
é
,
D
.
M
i
n
n
e
n
,
S
.
S
i
n
g
h
,
S
.
J.
H
w
a
n
g
,
a
n
d
N
.
J
o
h
n
s
t
o
n
,
“
V
a
r
i
a
t
i
o
n
a
l
i
ma
g
e
c
o
mp
r
e
ssi
o
n
w
i
t
h
a
s
c
a
l
e
h
y
p
e
r
p
r
i
o
r
,
”
6
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
L
e
a
r
n
i
n
g
Re
p
r
e
se
n
t
a
t
i
o
n
s
,
I
C
L
R
2
0
1
8
-
C
o
n
f
e
r
e
n
c
e
T
r
a
c
k
Pro
c
e
e
d
i
n
g
s
,
2
0
1
8
.
[
1
7
]
Z.
C
h
e
n
g
,
H
.
S
u
n
,
M
.
T
a
k
e
u
c
h
i
,
a
n
d
J.
K
a
t
t
o
,
“
D
e
e
p
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
mag
e
c
o
m
p
r
e
ss
i
o
n
,
”
Arx
i
v
,
2
0
1
9
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
1
9
0
6
.
0
9
7
3
1
.
[
1
8
]
A
.
K
a
r
,
S
.
P
.
K
.
K
a
r
r
i
,
N
.
G
h
o
sh
,
R
.
S
e
t
h
u
r
a
m
a
n
,
a
n
d
D
.
S
h
e
e
t
,
“
F
u
l
l
y
c
o
n
v
o
l
u
t
i
o
n
a
l
m
o
d
e
l
f
o
r
v
a
r
i
a
b
l
e
b
i
t
l
e
n
g
t
h
a
n
d
l
o
ssy
h
i
g
h
d
e
n
si
t
y
c
o
m
p
r
e
ssi
o
n
o
f
ma
mm
o
g
r
a
ms
,
”
Ar
x
i
v
,
2
0
1
8
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
1
8
0
5
.
0
6
9
0
9
.
[
1
9
]
A
.
S
.
S
u
s
h
mi
t
,
S
.
U
.
Za
ma
n
,
A
.
I
.
H
u
ma
y
u
n
,
T
.
H
a
sa
n
,
a
n
d
M
.
I
.
H
.
B
h
u
i
y
a
n
,
“
X
-
R
a
y
i
ma
g
e
c
o
m
p
r
e
ss
i
o
n
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
i
n
2
0
1
9
I
EEE
E
MB
S
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
Bi
o
m
e
d
i
c
a
l
& H
e
a
l
t
h
I
n
f
o
rm
a
t
i
c
s (
BH
I
)
,
M
a
y
2
0
1
9
,
p
p
.
1
–
4
,
d
o
i
:
1
0
.
1
1
0
9
/
B
H
I
.
2
0
1
9
.
8
8
3
4
6
5
6
.
[
2
0
]
P
.
G
u
o
,
D
.
L
i
,
a
n
d
X
.
L
i
,
“
D
e
e
p
O
C
T
i
ma
g
e
c
o
mp
r
e
ssi
o
n
w
i
t
h
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,”
Bi
o
m
e
d
i
c
a
l
O
p
t
i
c
s
E
x
p
ress
,
v
o
l
.
1
1
,
n
o
.
7
,
p
.
3
5
4
3
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
3
6
4
/
B
O
E
.
3
9
2
8
8
2
.
[
2
1
]
F
.
Li
u
,
M
.
H
e
r
n
a
n
d
e
z
-
C
a
b
r
o
n
e
r
o
,
V
.
S
a
n
c
h
e
z
,
M
.
W
.
M
a
r
c
e
l
l
i
n
,
a
n
d
A
.
B
i
l
g
i
n
,
“
Th
e
c
u
r
r
e
n
t
r
o
l
e
o
f
i
ma
g
e
c
o
m
p
r
e
ssi
o
n
s
t
a
n
d
a
r
d
s
i
n
me
d
i
c
a
l
i
m
a
g
i
n
g
,
”
I
n
f
o
rm
a
t
i
o
n
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
8
,
n
o
.
4
,
p
.
1
3
1
,
O
c
t
.
2
0
1
7
,
d
o
i
:
1
0
.
3
3
9
0
/
i
n
f
o
8
0
4
0
1
3
1
.
[
2
2
]
J.
G
u
o
a
n
d
H
.
C
h
a
o
,
“
B
u
i
l
d
i
n
g
d
u
a
l
-
d
o
ma
i
n
r
e
p
r
e
se
n
t
a
t
i
o
n
s f
o
r
c
o
m
p
r
e
ss
i
o
n
a
r
t
i
f
a
c
t
s
r
e
d
u
c
t
i
o
n
,
”
2
0
1
6
,
p
p
.
6
2
8
–
6
4
4
.
[
2
3
]
M
.
B
e
l
a
d
g
h
a
m,
Y
.
H
a
b
c
h
i
,
M
.
B
e
n
a
i
ssa
,
a
n
d
A
.
T
a
l
e
b
-
A
h
me
d
,
“
M
e
d
i
c
a
l
v
i
d
e
o
c
o
m
p
r
e
ssi
o
n
u
si
n
g
b
a
n
d
e
l
e
t
b
a
se
d
o
n
l
i
f
t
i
n
g
sc
h
e
me
a
n
d
S
P
I
H
T
c
o
d
i
n
g
:
I
n
sea
r
c
h
o
f
h
i
g
h
v
i
s
u
a
l
q
u
a
l
i
t
y
,
”
I
n
f
o
rm
a
t
i
c
s
i
n
Me
d
i
c
i
n
e
U
n
l
o
c
k
e
d
,
v
o
l
.
1
7
,
p
.
1
0
0
2
4
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
mu
.
2
0
1
9
.
1
0
0
2
4
4
.
[
2
4
]
O
.
H
.
N
a
g
o
o
r
,
J.
W
h
i
t
t
l
e
,
J
.
D
e
n
g
,
B
.
M
o
r
a
,
a
n
d
M
.
W
.
J
o
n
e
s,
“
M
e
d
ZI
P
:
3
D
me
d
i
c
a
l
i
ma
g
e
s
l
o
ssl
e
ss
c
o
mp
r
e
sso
r
u
s
i
n
g
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
(
LST
M
)
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
,
Ja
n
.
2
0
2
0
,
p
p
.
2
8
7
4
–
2
8
8
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
P
R
4
8
8
0
6
.
2
0
2
1
.
9
4
1
3
3
4
1
.
[
2
5
]
L.
T
h
e
i
s
,
W
.
S
h
i
,
A
.
C
u
n
n
i
n
g
h
a
m
,
a
n
d
F
.
H
u
s
z
á
r
,
“
L
o
ss
y
i
m
a
g
e
c
o
m
p
r
e
ssi
o
n
w
i
t
h
c
o
m
p
r
e
ssi
v
e
a
u
t
o
e
n
c
o
d
e
r
s,
”
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
rn
i
n
g
Re
p
res
e
n
t
a
t
i
o
n
s
,
I
C
L
R
2
0
1
7
-
C
o
n
f
e
re
n
c
e
T
r
a
c
k
P
ro
c
e
e
d
i
n
g
s
,
2
0
1
7
.
[
2
6
]
Y
.
W
a
n
g
,
G
.
W
a
n
g
,
C
.
C
h
e
n
,
a
n
d
Z.
P
a
n
,
“
M
u
l
t
i
-
s
c
a
l
e
d
i
l
a
t
e
d
c
o
n
v
o
l
u
t
i
o
n
o
f
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
f
o
r
i
mag
e
d
e
n
o
i
s
i
n
g
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
7
8
,
n
o
.
1
4
,
p
p
.
1
9
9
4
5
–
1
9
9
6
0
,
Ju
l
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
0
1
9
-
7
3
7
7
-
y.
[
2
7
]
M
.
A
n
t
o
n
i
n
i
,
M
.
B
a
r
l
a
u
d
,
P
.
M
a
t
h
i
e
u
,
a
n
d
I
.
D
a
u
b
e
c
h
i
e
s,
“
I
mag
e
c
o
d
i
n
g
u
si
n
g
w
a
v
e
l
e
t
t
r
a
n
sf
o
r
m,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ssi
n
g
,
v
o
l
.
1
,
n
o
.
2
,
p
p
.
2
0
5
–
2
2
0
,
A
p
r
.
1
9
9
2
,
d
o
i
:
1
0
.
1
1
0
9
/
8
3
.
1
3
6
5
9
7
.
[
2
8
]
M
.
B
e
l
a
d
g
h
a
m,
A
.
B
e
ssa
i
d
,
M
.
A
b
d
e
l
m
o
u
n
a
i
m,
a
n
d
T.
A
b
d
e
l
ma
l
i
k
,
“
I
mp
r
o
v
i
n
g
q
u
a
l
i
t
y
o
f
m
e
d
i
c
a
l
i
ma
g
e
c
o
mp
r
e
ss
i
o
n
u
si
n
g
b
i
o
r
t
h
o
g
o
n
a
l
C
D
F
w
a
v
e
l
e
t
b
a
s
e
d
o
n
l
i
f
t
i
n
g
sc
h
e
m
e
a
n
d
S
P
I
H
T
c
o
d
i
n
g
,
”
S
e
rb
i
a
n
J
o
u
rn
a
l
o
f
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
8
,
n
o
.
2
,
p
p
.
1
6
3
–
1
7
9
,
2
0
1
1
,
d
o
i
:
1
0
.
2
2
9
8
/
S
JE
E1
1
0
2
1
6
3
B
.
[
2
9
]
H
.
S
c
h
w
a
r
z
e
t
a
l
.
,
“
Q
u
a
n
t
i
z
a
t
i
o
n
a
n
d
e
n
t
r
o
p
y
c
o
d
i
n
g
i
n
t
h
e
v
e
r
s
a
t
i
l
e
v
i
d
e
o
c
o
d
i
n
g
(
V
V
C
)
st
a
n
d
a
r
d
,
”
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
1
,
n
o
.
1
0
,
p
p
.
3
8
9
1
–
3
9
0
6
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
TC
S
V
T
.
2
0
2
1
.
3
0
7
2
2
0
2
.
[
3
0
]
Q
.
H
a
o
e
t
a
l
.
,
“
H
i
g
h
si
g
n
a
l
-
to
-
n
o
i
se
r
a
t
i
o
r
e
c
o
n
s
t
r
u
c
t
i
o
n
o
f
l
o
w
b
i
t
-
d
e
p
t
h
o
p
t
i
c
a
l
c
o
h
e
r
e
n
c
e
t
o
m
o
g
r
a
p
h
y
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
J
o
u
rn
a
l
o
f
Bi
o
m
e
d
i
c
a
l
O
p
t
i
c
s
,
v
o
l
.
2
5
,
n
o
.
1
2
,
N
o
v
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
1
7
/
1
.
j
b
o
.
2
5
.
1
2
.
1
2
3
7
0
2
.
[
3
1
]
J.
C
a
i
,
Z
.
C
a
o
,
a
n
d
L
.
Z
h
a
n
g
,
“
L
e
a
r
n
i
n
g
a
s
i
n
g
l
e
t
u
c
k
e
r
d
e
c
o
mp
o
si
t
i
o
n
n
e
t
w
o
r
k
f
o
r
l
o
ss
y
i
ma
g
e
c
o
m
p
r
e
ss
i
o
n
w
i
t
h
mu
l
t
i
p
l
e
b
i
t
s
-
p
e
r
-
p
i
x
e
l
r
a
t
e
s,
”
I
EE
E
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
9
,
p
p
.
3
6
1
2
–
3
6
2
5
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
2
0
.
2
9
6
3
9
5
6
.
[
3
2
]
J.
W
a
n
g
,
T.
Li
u
,
Q
.
Li
u
,
X
.
H
e
,
H
.
L
u
o
,
a
n
d
W
.
H
e
,
“
C
o
m
p
r
e
ssi
o
n
r
a
t
i
o
mo
d
e
l
i
n
g
a
n
d
e
s
t
i
m
a
t
i
o
n
a
c
r
o
ss
e
r
r
o
r
b
o
u
n
d
s
f
o
r
l
o
ss
y
c
o
m
p
r
e
ss
i
o
n
,
”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
P
a
ra
l
l
e
l
a
n
d
D
i
s
t
ri
b
u
t
e
d
S
y
st
e
m
s
,
v
o
l
.
3
1
,
n
o
.
7
,
p
p
.
1
6
2
1
–
1
6
3
5
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
TPD
S
.
2
0
1
9
.
2
9
3
8
5
0
3
.
[
3
3
]
B
.
S
u
j
i
t
h
a
,
V
.
S
.
P
a
r
v
a
t
h
y
,
E.
L.
L
y
d
i
a
,
P
.
R
a
n
i
,
Z.
P
o
l
k
o
w
s
k
i
,
a
n
d
K
.
S
h
a
n
k
a
r
,
“
O
p
t
i
ma
l
d
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
i
m
a
g
e
c
o
m
p
r
e
ss
i
o
n
t
e
c
h
n
i
q
u
e
f
o
r
d
a
t
a
t
r
a
n
sm
i
ssi
o
n
o
n
i
n
d
u
s
t
r
i
a
l
I
n
t
e
r
n
e
t
o
f
t
h
i
n
g
s
a
p
p
l
i
c
a
t
i
o
n
s
,
”
T
ra
n
s
a
c
t
i
o
n
s
o
n
Em
e
r
g
i
n
g
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
s
T
e
c
h
n
o
l
o
g
i
e
s
,
v
o
l
.
3
2
,
n
o
.
7
,
J
u
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
2
/
e
t
t
.
3
9
7
6
.
[
3
4
]
J.
L
.
B
a
a
n
d
D
.
P
.
K
i
n
g
m
a
,
“
A
d
a
m:
a
m
e
t
h
o
d
f
o
r
st
o
c
h
a
s
t
i
c
o
p
t
i
mi
z
a
t
i
o
n
,
”
3
r
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
r
n
i
n
g
Re
p
r
e
se
n
t
a
t
i
o
n
s
,
I
C
L
R
2
0
1
5
-
C
o
n
f
e
r
e
n
c
e
T
r
a
c
k
Pr
o
c
e
e
d
i
n
g
s
,
p
p
.
1
–
1
5
,
2
0
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.