I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
201
6
,
p
p
.
267
4
~
2
6
8
1
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
6i
6
.
1
1
3
0
4
2674
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
JE
C
E
Retinal A
rea Seg
m
en
tatio
n using
Ada
ptive Sup
erpi
x
a
la
tion a
nd
its
Cla
ss
ificatio
n
using
RBFN
Ni
m
is
ha
,
Ra
na
G
ill
De
p
a
rt
m
e
n
t
o
f
El
e
c
tro
n
ics
a
n
d
C
o
m
u
n
ica
ti
o
n
E
n
g
in
e
e
rin
g
,
C
h
a
n
d
i
g
a
rh
Un
iv
e
rsit
y
,
G
h
a
ru
a
n
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
y
1
9
,
2
0
1
6
R
ev
i
s
ed
J
u
l
2
8
,
2
0
1
6
A
cc
ep
ted
A
u
g
1
1
,
2
0
1
6
Re
ti
n
a
l
d
ise
a
se
is
th
e
v
e
r
y
i
m
p
o
r
tan
t
issu
e
in
m
e
d
ica
l
f
i
e
ld
.
T
o
d
i
a
g
n
o
se
th
e
d
ise
a
se
,
it
n
e
e
d
s
to
d
e
tec
t
th
e
tr
u
e
re
ti
n
a
l
a
re
a
.
A
rte
f
a
c
ts
li
k
e
e
y
e
li
d
s
a
n
d
e
y
e
las
h
e
s
a
re
c
o
m
e
a
lo
n
g
w
it
h
re
ti
n
a
l
p
a
rt
so
re
m
o
v
a
l
o
f
a
rte
f
a
c
ts
is
th
e
b
ig
tas
k
f
o
r
b
e
tt
e
r
d
iag
n
o
sis
o
f
d
ise
a
se
in
to
th
e
re
ti
n
a
l
p
a
rt.
In
th
is
p
a
p
e
r,
we
h
a
v
e
p
ro
p
o
se
d
t
h
e
se
g
m
e
n
tatio
n
a
n
d
u
se
m
a
c
h
in
e
lea
rn
in
g
a
p
p
r
o
a
c
h
e
s
to
d
e
tec
t
th
e
tru
e
re
ti
n
a
l
p
a
rt.
P
re
p
ro
c
e
ss
in
g
is
d
o
n
e
o
n
th
e
o
rig
in
a
l
i
m
a
g
e
u
sin
g
G
a
m
m
a
No
r
m
a
li
z
a
ti
o
n
w
h
ich
h
e
lp
s
to
e
n
h
a
n
c
e
th
e
im
a
g
e
th
a
t
c
a
n
g
iv
e
s
d
e
tail
in
f
o
rm
a
ti
o
n
a
b
o
u
t
t
h
e
im
a
g
e
.
T
h
e
n
th
e
se
g
m
e
n
tatio
n
is
p
e
r
f
o
r
m
e
d
o
n
th
e
Ga
m
m
a
No
r
m
a
li
z
e
d
i
m
a
g
e
b
y
S
u
p
e
rp
ix
e
l
m
e
th
o
d
.
S
u
p
e
rp
ix
e
l
is
th
e
g
ro
u
p
o
f
p
ix
e
l
in
to
d
if
f
e
re
n
t
re
g
io
n
s
w
h
ich
is
b
a
se
d
o
n
c
o
m
p
a
c
tn
e
ss
a
n
d
re
g
io
n
a
l
siz
e
.
S
u
p
e
rp
ix
e
l
is
u
se
d
to
re
d
u
c
e
t
h
e
c
o
m
p
lex
it
y
o
f
im
a
g
e
p
ro
c
e
ss
in
g
tas
k
a
n
d
p
ro
v
id
e
su
it
a
b
le
p
rim
it
iv
e
i
m
a
g
e
p
a
tt
e
rn
.
T
h
e
n
f
e
a
tu
re
g
e
n
e
ra
ti
o
n
m
u
st
b
e
d
o
n
e
a
n
d
m
a
c
h
in
e
lea
rn
in
g
a
p
p
ro
a
c
h
h
e
l
p
s
to
e
x
trac
t
tru
e
re
ti
n
a
l
a
re
a
.
T
h
e
e
x
p
e
ri
m
e
n
tal
e
v
a
lu
a
ti
o
n
g
iv
e
s th
e
b
e
tt
e
r
re
su
lt
w
it
h
a
c
c
u
ra
c
y
o
f
9
6
%
.
K
ey
w
o
r
d
:
E
x
tr
ac
tio
n
o
f
r
eti
n
al
ar
ea
Featu
r
e
g
e
n
er
atio
n
Ma
ch
i
n
e
lear
n
i
n
g
ap
p
r
o
ac
h
Su
p
er
p
ix
el
(
S
L
I
C
O)
Co
p
y
rig
h
t
©
2
0
1
6
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
ce
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ni
m
i
s
h
a,
Dep
ar
t
m
en
t o
f
E
lectr
o
n
ics a
n
d
C
o
m
m
u
n
icat
io
n
E
n
g
i
n
ee
r
in
g
,
C
h
a
n
d
ig
ar
h
U
n
iv
er
s
it
y
,
Natio
n
al
Hi
g
h
w
a
y
9
5
,
C
h
a
n
d
i
g
ar
h
-
L
u
d
h
ia
n
a
Hig
h
w
a
y
,
Sa
h
i
b
za
d
a
A
j
it Sin
g
h
Na
g
ar
,
P
u
n
j
ab
1
4
0
4
1
3
,
I
n
d
ia
.
E
m
ail:
n
i
m
is
h
a2
5
.
s
in
g
h
@
g
m
a
il.c
o
m
1.
I
NT
RO
D
UCT
I
O
N
No
w
ad
a
y
s
,
B
io
m
etr
ic
s
y
s
te
m
s
ar
e
b
ec
o
m
i
n
g
s
u
i
tab
le
f
o
r
co
n
v
en
tio
n
al
m
e
th
o
d
s
s
u
ch
as
P
I
N,
p
ass
w
o
r
d
etc.
B
io
m
e
tr
ic
tec
h
n
iq
u
es
ar
e
d
ep
en
d
i
n
g
u
p
o
n
b
eh
av
io
u
r
al
o
r
p
h
y
s
io
lo
g
ic
al
tr
ai
t
li
k
e
h
an
d
g
eo
m
etr
y
,
b
lo
o
d
v
ess
el
p
atter
n
s
,
g
ait,
s
i
g
n
at
u
r
e
an
d
ir
is
etc.
I
r
is
p
atter
n
s
co
n
s
is
t o
f
u
n
iq
u
e
i
n
f
o
r
m
atio
n
s
u
c
h
a
s
r
id
g
es,
cr
y
p
ts
,
co
r
o
n
a,
f
u
r
r
o
ws,
r
in
g
s
,
a
zig
za
g
co
llar
ette
a
n
d
f
r
ec
k
le
s
etc.
T
h
e
ac
cu
r
ac
y
o
f
I
r
is
r
ec
o
g
n
itio
n
d
ep
en
d
s
u
p
o
n
its
s
e
g
m
e
n
tat
io
n
.
I
r
is
s
eg
m
e
n
tatio
n
i
s
af
f
ec
t
ed
b
y
th
e
r
eg
io
n
o
f
e
y
ela
s
h
e
s
an
d
e
y
elid
s
.
T
h
e
e
y
elas
h
es
h
id
e
e
y
elid
b
o
u
n
d
ar
ies s
o
e
y
elid
d
etec
tio
n
is
d
i
f
f
ic
u
lt.
B
u
t
e
y
ela
s
h
es c
a
n
b
e
d
ete
cted
an
d
eli
m
in
a
ted
b
y
t
h
e
w
a
v
elet
tr
an
s
f
o
r
m
[
1
]
.
T
h
e
d
is
ea
s
es
lik
e
ag
e
-
r
elat
ed
m
u
s
c
u
lar
d
eg
e
n
er
atio
n
(
A
MD
)
an
d
d
iab
etic
r
eti
n
o
p
ath
y
(
D
R
)
af
f
ec
t
a
lar
g
e
n
u
m
b
er
o
f
p
o
p
u
latio
n
a
n
d
also
it
i
s
e
x
p
ec
ted
to
th
ese
d
is
ea
s
es
m
u
s
t
b
e
in
cr
ea
s
ed
i
n
o
u
r
co
m
i
n
g
f
u
t
u
r
e.
Gen
er
all
y
,
Di
g
ital
f
u
n
d
u
s
p
h
o
to
g
r
ap
h
y
is
u
s
ed
to
s
cr
ee
n
a
n
d
id
en
t
if
y
th
e
n
a
tu
r
e
o
f
r
eti
n
a
r
elate
d
co
n
d
itio
n
w
h
ic
h
is
p
o
s
s
ib
le
to
allo
w
i
m
ag
e
s
to
r
ag
e,
in
a
n
o
n
in
v
a
s
iv
e
ex
a
m
in
a
tio
n
a
n
d
f
o
r
t
h
e
tr
an
s
m
is
s
io
n
a
t
d
if
f
er
en
t
lo
ca
tio
n
.
So
,
th
e
d
ig
i
tal
r
eti
n
al
i
m
ag
e
s
ar
e
ex
a
m
in
ed
b
y
an
ex
p
er
t
h
u
m
a
n
g
r
ad
er
i.e
,
Op
to
m
etr
i
s
ts
a
n
d
Op
th
al
m
o
g
i
s
ts
w
h
ic
h
p
er
f
o
r
m
th
e
w
h
o
le
p
r
o
ce
s
s
th
at
is
ti
m
e
co
n
s
u
m
i
n
g
an
d
d
i
f
f
icu
l
t
[
2
]
.
Au
to
m
a
ted
an
al
y
s
i
s
o
f
r
etin
al
i
m
a
g
e
s
h
av
e
t
h
e
ab
ilit
y
to
r
ed
u
ce
th
e
ti
m
e
a
n
d
also
it
m
u
s
t
b
e
d
etec
ted
th
e
p
r
o
b
le
m
o
f
th
e
r
etin
a
l
p
ar
t
v
er
y
ea
s
il
y
.
R
eti
n
al
ar
ea
o
b
tain
ed
f
r
o
m
t
h
e
i
m
a
g
i
n
g
i
n
s
tr
u
m
e
n
t
s
s
u
c
h
a
s
f
u
n
d
u
s
ca
m
er
a
a
n
d
s
ca
n
n
in
g
laser
o
p
th
al
m
o
s
co
p
e
(
S
L
O)
wh
ich
co
n
tai
n
s
tr
u
c
tu
r
e
o
f
r
eti
n
a
l a
r
ea
w
it
h
ar
tef
ac
t
s
(
e
y
ela
s
h
e
s
an
d
e
y
elid
s
)
[
3
]
.
R
e
m
o
v
al
o
f
ar
te
f
ac
t
is
t
h
e
i
m
p
o
r
tan
t
s
tep
b
ef
o
r
e
th
e
d
etec
ti
o
n
o
f
r
eti
n
al
d
is
ea
s
es.
E
x
tr
an
e
o
u
s
o
b
j
ec
t
lik
e
e
y
las
h
es,
d
u
s
t
an
d
e
y
elid
s
o
n
th
e
o
p
tical
s
u
r
f
ac
es
m
a
y
co
m
e
in
f
o
cu
s
an
d
also
ap
p
ea
r
b
r
ig
h
t.
E
x
tr
ac
tio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
R
etin
a
l A
r
ea
S
eg
men
t
a
tio
n
u
s
i
n
g
A
d
a
p
tive
S
u
p
erp
ixa
la
tio
n
a
n
d
its
cla
s
s
ifica
tio
n
u
s
in
g
R
B
F
N
(
N
imi
s
h
a
)
2675
o
f
tr
u
e
r
etin
al
ar
ea
f
r
o
m
t
h
e
S
L
O
i
m
a
g
es
is
i
m
p
o
r
tan
t
f
o
r
th
e
d
iag
n
o
s
is
o
f
e
y
e
d
is
ea
s
es.
T
h
is
is
v
er
y
d
if
f
ic
u
l
t
to
d
if
f
er
en
tiate
b
et
w
ee
n
t
h
e
ar
tef
ac
ts
a
n
d
th
e
tr
u
e
r
etin
al
ar
e
a.
T
h
e
SL
O
i
m
a
g
es
o
f
r
etin
a
t
h
at
is
o
b
tain
ed
f
r
o
m
o
p
to
s
[
4
]
w
it
h
t
h
e
w
id
t
h
o
f
2
0
0
º
ap
p
r
o
x
i
m
atel
y
.
F
ig
u
r
e
1
s
h
o
w
s
t
h
at
t
h
e
r
eti
n
al
i
m
a
g
es
w
h
ic
h
i
s
ca
p
t
u
r
ed
u
s
i
n
g
S
L
O
an
d
f
u
n
d
u
s
ca
m
er
a
.
W
ith
th
e
r
eti
n
al
ar
ea
e
y
elas
h
an
d
e
y
elid
s
tr
u
ctu
r
es
ar
e
also
i
m
ag
ed
d
u
e
to
t
h
e
lar
g
e
f
ield
o
f
v
ie
w
(
FOV)
i
n
th
e
S
L
O
i
m
ag
e
s
.
I
f
e
y
elid
s
an
d
e
y
elas
h
es
ar
e
r
e
m
o
v
ed
t
h
en
an
al
y
s
i
s
o
f
r
et
in
a
l
ar
ea
alo
n
g
w
it
h
t
h
e
d
ia
g
n
o
s
i
s
o
f
d
is
ea
s
e
s
in
th
e
v
i
s
ib
le
r
eti
n
a
ca
n
b
e
d
o
n
e.
I
n
th
i
s
p
ap
er
,
w
e
h
a
v
e
d
o
n
e
t
h
e
s
eg
m
e
n
tatio
n
p
ar
t a
n
d
clas
s
i
f
i
er
co
n
s
tr
u
ct
io
n
w
h
ic
h
i
s
h
e
lp
t
o
ex
tr
ac
t t
h
e
r
eti
n
al
ar
ea
.
T
h
en
d
iag
n
o
s
i
s
o
f
r
etin
a
l
p
ar
t c
an
b
e
ea
s
il
y
d
o
n
e.
Fig
u
r
e
1
.
(
a)
A
Fu
n
d
u
s
I
m
a
g
e
,
(
b
)
A
n
S
L
O
Im
a
g
e
w
it
h
T
r
u
e
R
e
ti
n
al
A
r
ea
[
3
]
T
h
e
m
ai
n
f
o
u
r
s
tep
s
ar
e
d
o
n
e
i
n
t
h
i
s
p
ap
er
:
a.
P
r
e
p
r
o
ce
s
s
in
g
ta
s
k
o
f
S
L
O
i
m
ag
e
s
.
P
r
ep
r
o
ce
s
s
in
g
i
s
d
o
n
e
w
it
h
Ga
m
m
a
No
r
m
aliza
tio
n
.
b
.
Gen
er
atio
n
o
f
S
u
p
er
p
ix
el
o
f
th
e
Ga
m
m
a
No
r
m
al
ized
i
m
a
g
e.
Su
p
er
p
ix
el
w
i
ll
b
e
g
en
er
at
ed
b
y
s
i
m
p
le
li
n
ea
r
iter
ativ
e
cl
u
s
ter
i
n
g
(
S
L
I
C
)
.
c.
Af
ter
g
e
n
er
atin
g
t
h
e
s
u
p
er
p
ix
el,
f
ea
t
u
r
e
g
e
n
er
atio
n
(
te
x
t
u
r
al
f
ea
t
u
r
e
an
d
g
r
ad
ien
t
f
ea
t
u
r
e)
is
d
o
n
e.
d
.
Fo
r
class
if
icatio
n
o
f
tr
u
e
r
et
in
al
ar
ea
f
r
o
m
SLO
i
m
a
g
e
s
,
co
n
s
tr
u
ctio
n
o
f
class
if
ier
i
s
i
m
p
o
r
tan
t.
T
h
e
p
ap
e
r
is
ar
r
an
g
ed
a
s
f
o
llo
w
s
.
Sec
tio
n
I
I
s
h
o
w
s
o
u
r
p
r
o
p
o
s
ed
m
et
h
o
d
s
w
h
ic
h
co
n
tain
p
r
ep
r
o
ce
s
s
in
g
,
Su
p
er
p
ix
el
g
en
er
at
io
n
,
f
ea
tu
r
e
g
en
er
atio
n
a
n
d
co
n
s
tr
u
ctio
n
o
f
clas
s
i
f
ier
.
Sectio
n
I
I
I
p
r
o
v
id
e
s
th
e
e
x
p
er
i
m
e
n
tal
an
d
v
i
s
u
a
l
r
es
u
lt
o
f
o
u
r
p
r
o
p
o
s
ed
m
et
h
o
d
.
An
d
at
la
s
t
s
ec
ti
o
n
I
V
s
h
o
w
s
t
h
e
d
is
c
u
s
s
io
n
a
b
o
u
t
r
esu
lt
an
d
t
h
e
co
n
clu
s
io
n
.
2.
P
RO
P
O
SE
D
M
E
T
H
O
D
2
.
1
.
I
m
a
g
e
Acquis
it
io
n
I
m
ag
e
ac
q
u
is
it
io
n
is
t
h
e
v
er
y
b
asic a
n
d
i
m
p
o
r
tan
t
s
tep
.
I
t g
i
v
es t
h
e
ac
q
u
i
s
itio
n
o
f
i
m
ag
e
d
ata
w
i
th
t
h
e
an
n
o
tatio
n
ar
o
u
n
d
th
e
ac
t
u
al
d
etec
tio
n
p
ar
t
o
f
th
e
i
m
a
g
es.
An
y
m
et
h
o
d
o
r
p
r
o
ce
s
s
ca
n
b
e
ap
p
lied
o
n
th
e
i
m
a
g
es.
I
m
a
g
e
s
ca
n
b
e
co
llect
ed
f
r
o
m
v
ar
io
u
s
f
ield
li
k
e
b
io
m
ed
ical,
s
atell
ite,
p
lan
t
etc.
T
h
ese
i
m
a
g
es
ca
n
b
e
o
b
tain
ed
u
s
i
n
g
a
n
y
d
ig
i
tal
eq
u
ip
m
en
t
li
k
e
ca
m
er
a,
lap
to
p
,
m
o
b
ile
p
h
o
n
es
etc.
Si
x
s
tan
d
ar
d
i
m
a
g
es
h
a
v
e
b
ee
n
tak
en
f
r
o
m
t
h
e
d
ataset
[
9
]
.
Fo
r
th
e
f
i
n
d
in
g
o
f
r
es
u
l
ts
,
i
m
a
g
e
ac
q
u
is
itio
n
m
u
s
t
b
e
r
eq
u
ir
ed
i
n
t
h
e
f
ield
o
f
i
m
a
g
e
p
r
o
ce
s
s
in
g
[
5
]
.
2
.
2
.
G
a
mm
a
No
r
m
a
liza
t
io
n
Ga
m
m
a
No
r
m
al
izatio
n
i
s
a
n
o
n
-
li
n
ea
r
o
p
er
atio
n
w
h
ich
i
s
u
s
ed
to
co
n
tr
o
l
th
e
to
tal
b
r
ig
h
t
n
es
s
o
f
t
h
e
g
iv
e
n
i
m
a
g
e.
I
t is also
k
n
o
w
n
as Ga
m
m
a
co
r
r
ec
tio
n
o
r
P
o
w
e
r
la
w
tr
an
s
f
o
r
m
atio
n
a
n
d
is
d
ef
i
n
ed
as
=
(
1
)
w
h
er
e
ɣ
is
g
a
m
m
a
.
Gen
er
all
y
,
Ga
m
m
a
v
al
u
e
r
an
g
e
is
i
n
b
et
w
ee
n
0
an
d
1
i.e
,
0
≤
ɣ
≤
1
,
w
h
e
n
ɣ
≤
1
th
en
th
e
o
u
tp
u
t
i
m
a
g
e
i
s
b
r
ig
h
ter
a
n
d
it
is
ca
lled
a
s
a
n
en
co
d
i
n
g
g
a
m
m
a
w
h
er
ea
s
i
f
ɣ
≥
1
th
e
n
t
h
e
o
u
t
p
u
t
i
m
ag
e
is
d
ar
k
er
an
d
it
i
s
ca
lled
as
d
ec
o
d
in
g
g
a
m
m
a
an
d
i
f
ɣ
=1
t
h
en
it
i
s
lin
ea
r
.
[
6
]
Var
io
u
s
s
tep
o
f
e
n
h
a
n
ce
m
en
t
ca
n
b
e
g
en
er
ated
f
o
r
th
e
d
if
f
er
e
n
t
v
a
lu
es
o
f
ɣ
.
Ga
m
m
a
No
r
m
a
lizat
io
n
ca
n
b
r
ig
h
te
n
t
h
e
in
ten
s
itie
s
o
f
a
n
y
i
m
a
g
e.
I
t
ca
n
also
b
e
u
s
ed
f
o
r
en
h
a
n
ce
th
e
co
n
tr
a
s
t
o
f
i
m
a
g
es
w
h
ic
h
h
as
lo
w
i
n
te
n
s
i
t
y
v
al
u
e.
I
f
ɣ
<
1
,
th
en
t
h
e
g
a
m
m
a
co
r
r
ec
tio
n
f
o
r
m
s
th
e
s
m
all
r
a
n
g
e
o
f
d
ar
k
p
ix
el
v
a
lu
e
to
th
e
lar
g
e
r
an
g
e
an
d
th
e
lar
g
er
r
an
g
e
o
f
b
r
ig
h
t
p
ix
el
v
alu
e
i
n
to
t
h
e
s
m
all
r
an
g
e.
I
t c
an
b
e
w
r
itte
n
as
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
2
6
7
4
–
2
6
8
1
2676
S=c
(
2
)
w
h
er
e
c
an
d
ɣ
ar
e
p
o
s
itiv
e
co
n
s
tan
t
v
al
u
e.
Fro
m
t
h
e
ab
o
v
e
eq
u
at
io
n
,
s
v
er
s
u
s
r
ca
n
b
e
p
lo
tted
f
o
r
d
if
f
er
en
t
v
a
lu
e
o
f
ɣ
w
h
i
c
h
is
s
h
o
w
n
i
n
t
h
e
F
ig
u
r
e
2
.
T
h
e
ab
o
v
e
Fig
u
r
e
2
s
h
o
w
s
t
h
at
t
h
e
v
alu
e
s
ɣ
>1
h
av
e
th
e
o
p
p
o
s
ite
ef
f
ec
t
a
s
t
h
e
v
alu
e
s
ɣ
<1
.
W
h
en
c=
ɣ
=1
,
it
r
ed
u
ce
s
th
e
id
en
ti
t
y
tr
an
s
f
o
r
m
atio
n
.
So
,
t
h
e
g
a
m
m
a
co
r
r
ec
tio
n
is
u
s
ed
to
r
ec
tify
th
e
p
o
w
er
la
w
r
esp
o
n
s
e
p
h
e
n
o
m
e
n
a
[
7
]
.
I
n
t
h
is
p
ap
er
,
Ga
m
m
a
ad
j
u
s
t
m
e
n
t
i
s
u
s
ed
to
b
r
in
g
th
e
m
ea
n
in
te
n
s
it
y
o
f
th
e
i
m
ag
e
to
th
e
tar
g
et
v
al
u
e.
So
,
Ga
m
m
a
c
an
b
e
ca
lcu
lated
b
y
ɣ
=
)
-
/
)
-
(
3
)
w
h
er
e,
=
m
ea
n
i
n
ten
s
it
y
o
f
o
r
ig
in
a
l
i
m
a
g
e
,
=
m
ea
n
i
n
te
n
s
it
y
o
f
tar
g
et
i
m
a
g
e
an
d
=
m
ea
n
in
te
n
s
it
y
o
f
tar
g
et
i
m
a
g
e
.
I
n
t
h
is
p
ap
er
,
is
s
et
to
8
0
f
o
r
th
e
i
m
ag
e
v
is
u
aliza
tio
n
an
d
t
h
u
s
th
e
g
a
m
m
a
ad
j
u
s
t
m
e
n
t is
=
(
4
)
Fig
u
r
e
2
.
P
lo
t
o
f
E
q
u
ati
o
n
(
2
)
f
o
r
Var
io
u
s
Va
l
u
e
o
f
Ɣ(
c=
1
f
o
r
all)
[
7
]
2
.
3
.
Su
perp
ix
el
G
ener
a
t
io
n
Af
ter
t
h
e
g
a
m
m
a
n
o
r
m
aliza
t
io
n
,
th
e
n
e
x
t
s
tep
i
s
to
g
e
n
er
ate
th
e
s
u
p
er
p
ix
el.
T
h
e
alg
o
r
it
h
m
w
h
ic
h
is
u
s
ed
to
g
r
o
u
p
all
t
h
e
p
i
x
els
in
to
th
e
d
i
f
f
er
e
n
t
r
e
g
io
n
i
s
ca
lle
d
s
u
p
er
p
ix
el
al
g
o
r
ith
m
w
h
ich
is
u
s
ed
to
ca
lc
u
lat
e
th
e
f
ea
tu
r
e
o
f
t
h
e
i
m
ag
e
a
n
d
it
is
also
u
s
ed
to
r
ed
u
ce
th
e
co
m
p
lex
i
t
y
o
f
th
e
i
m
a
g
e
p
r
o
ce
s
s
in
g
ta
s
k
.
Su
p
er
p
ix
el
ar
e
u
s
ed
to
ca
p
tu
r
e
i
m
ag
e
r
ed
u
n
d
a
n
c
y
a
n
d
s
u
p
p
o
r
t
th
e
co
n
v
en
ien
t
p
r
i
m
iti
v
e
i
m
ag
e
p
atter
n
.
[
3
]
I
n
th
is
p
ap
er
,
Su
p
er
p
ix
el
is
g
en
er
ated
b
y
t
h
e
ad
ap
tiv
e
SL
I
C
(
SL
I
C
O)
w
h
ic
h
is
s
i
m
ilar
to
th
e
S
L
I
C
.
SL
I
C
u
s
e
s
s
o
m
e
co
m
p
ac
t
n
es
s
p
ar
a
m
e
ter
w
h
ic
h
i
s
c
h
o
s
en
b
y
u
s
er
f
o
r
e
v
er
y
s
u
p
er
p
ix
el
in
th
e
i
m
ag
e.
I
f
i
m
a
g
e
i
s
s
m
o
o
t
h
at
p
ar
ticu
lar
r
eg
io
n
b
u
t
i
n
t
h
e
o
t
h
er
r
eg
io
n
,
it
is
h
i
g
h
l
y
te
x
t
u
r
ed
,
th
en
t
h
e
S
L
I
C
g
en
er
ate
s
m
o
o
t
h
r
e
g
u
lar
-
s
ized
s
u
p
er
p
ix
el
s
at
t
h
e
s
m
o
o
th
r
eg
io
n
an
d
a
t
t
h
e
h
i
g
h
l
y
te
x
tu
r
ed
r
e
g
io
n
,
S
L
I
C
p
r
o
d
u
ce
s
h
ig
h
l
y
ir
r
eg
u
lar
s
u
p
er
p
ix
el
s
.
So
,
it
b
ec
o
m
es
m
o
r
e
co
m
p
licated
to
s
elec
t
th
e
r
ig
h
t
p
ar
am
e
ter
f
o
r
ea
ch
i
m
ag
e.
B
u
t
in
th
e
S
L
I
C
O,
th
e
u
s
er
n
ee
d
n
o
t
to
s
elec
t
to
s
elec
t
th
e
co
m
p
ac
t
n
es
s
p
ar
am
eter
.
S
L
I
C
O
ad
ap
tiv
el
y
s
ele
ct
th
e
co
m
p
ac
tn
e
s
s
p
ar
am
eter
f
o
r
ea
ch
an
d
e
v
er
y
s
u
p
er
p
ix
el
d
i
f
f
e
r
en
tl
y
.
Fo
r
b
o
t
h
te
x
t
u
r
ed
an
d
n
o
n
-
tex
t
u
r
ed
r
eg
io
n
s
,
S
L
I
C
O
u
s
ed
to
g
e
n
er
ate
r
eg
u
lar
s
h
ap
ed
s
u
p
er
p
ix
els,
An
d
also
S
L
I
C
O
is
v
er
y
f
ast
as
co
m
p
ar
ed
t
o
S
L
I
C
.
I
n
t
h
e
g
i
v
en
Fig
u
r
e
3
,
th
e
i
m
ag
e
s
i
n
t
h
e
t
o
p
r
o
w
s
h
o
w
s
a
co
n
s
ta
n
t
co
m
p
ac
tn
e
s
s
f
ac
to
r
f
o
r
ev
er
y
s
u
p
er
p
ix
e
ls
b
u
t
in
th
e
b
o
tto
m
r
o
w
i
m
ag
e
s
s
h
o
w
S
L
I
C
O
o
u
tp
u
t
w
h
ic
h
i
s
ad
ap
tiv
el
y
ch
o
s
e
n
t
h
e
co
m
p
ac
t
n
es
s
f
ac
to
r
f
o
r
ev
er
y
s
u
p
er
p
ix
el
[
8
]
.
T
h
e
t
w
o
m
ai
n
p
o
in
ts
ar
e
t
h
er
e
f
o
r
ad
ap
tatio
n
o
f
s
u
p
er
p
ix
el
g
e
n
er
atio
n
:
a.
B
y
li
m
iti
n
g
th
e
s
ea
r
c
h
s
p
a
ce
at
th
e
r
eg
io
n
w
h
ich
is
p
r
o
p
o
r
tio
n
al
to
th
e
s
u
p
er
p
ix
el
s
i
ze
is
r
ed
u
ce
d
b
y
th
e
n
u
m
b
er
o
f
d
is
ta
n
ce
ca
lcu
latio
n
.
b
.
B
y
co
m
b
i
n
i
n
g
t
h
e
s
p
atial
an
d
co
lo
u
r
p
r
o
x
i
m
it
y
,
th
e
w
ei
g
h
ted
d
is
tan
ce
ca
n
b
e
m
ea
s
u
r
ed
.
A
n
d
th
at
w
ei
g
h
ted
d
is
tan
ce
u
s
e
to
co
n
tr
o
l o
v
er
th
e
co
m
p
ac
t
n
es
s
an
d
th
e
s
ize
o
f
Su
p
er
p
ix
el
[
9
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
R
etin
a
l A
r
ea
S
eg
men
t
a
tio
n
u
s
i
n
g
A
d
a
p
tive
S
u
p
erp
ixa
la
tio
n
a
n
d
its
cla
s
s
ifica
tio
n
u
s
in
g
R
B
F
N
(
N
imi
s
h
a
)
2677
Fig
u
r
e
3
.
U
p
p
er
f
ig
u
r
e
s
h
o
w
s
SL
I
C
an
d
lo
w
er
f
i
g
s
h
o
w
s
S
L
I
C
O
[
8
]
2
.
3
.
1
.
Alg
o
rit
h
m
(1
)
SL
I
C
is
s
i
m
p
le
alg
o
r
it
h
m
to
u
s
e.
I
n
th
is
al
g
o
r
ith
m
o
n
l
y
o
n
e
p
ar
am
eter
is
u
s
ed
i.e
,
k
w
h
ic
h
h
as
eq
u
al
s
ized
S
u
p
er
p
ix
els.
T
h
e
clu
s
ter
i
n
g
p
r
o
ce
d
u
r
e
s
tar
ts
w
i
th
an
i
n
it
ializatio
n
s
tep
f
o
r
th
e
co
lo
u
r
i
m
ag
e
s
i
n
C
I
E
L
A
B
co
lo
u
r
s
p
ac
e
w
h
er
e
k
s
tar
tin
g
cl
u
s
ter
ce
n
tr
e
=
[
]
th
at
i
s
s
a
m
p
led
o
n
t
h
e
co
m
m
o
n
g
r
id
s
p
ac
ed
S
p
ix
els.
T
h
e
g
r
id
i
n
t
er
v
al
S=√
N/
k
is
to
p
r
o
d
u
ce
r
o
u
g
h
l
y
s
a
m
e
s
ized
s
u
p
er
p
ix
el
s
.
T
h
e
ce
n
tr
es
ar
e
ch
an
g
ed
t
h
eir
p
o
s
itio
n
to
th
e
s
m
a
lles
t
g
r
ad
ien
t
p
o
s
itio
n
i
n
t
h
e
3
×3
n
ei
g
h
b
o
u
r
h
o
o
d
.
T
h
e
n
e
x
t
s
tep
is
th
e
ass
i
g
n
m
e
n
t
s
tep
.
I
n
th
i
s
s
tep
e
ac
h
p
ix
e
l
i
is
co
n
n
ec
ted
w
it
h
n
ea
r
est
cl
u
s
ter
ce
n
tr
e
in
w
h
ich
lo
ca
tio
n
i
s
o
v
er
lap
w
it
h
s
ea
r
c
h
r
eg
io
n
.
T
h
e
s
ize
o
f
th
e
s
ea
r
c
h
r
eg
io
n
i
s
li
m
itin
g
w
h
ich
d
ec
r
ea
s
es
t
h
e
n
u
m
b
er
o
f
d
is
ta
n
ce
ca
lcu
latio
n
an
d
h
e
n
ce
th
e
r
esu
lt
in
s
p
ee
d
is
h
i
g
h
er
as
co
m
p
a
r
ed
to
t
r
ad
itio
n
al
k
-
m
ea
n
clu
s
ter
in
g
i
n
w
h
ich
al
l
th
e
clu
s
ter
ce
n
tr
e
co
m
p
ar
ed
w
it
h
ea
ch
p
ix
e
l.
So
th
e
s
p
atial
ex
ten
t
o
f
t
h
e
r
eg
io
n
w
h
o
s
e
s
ize
is
S×S,
s
i
m
ila
r
p
ix
el
ca
n
b
e
s
ea
r
c
h
ed
i
n
t
h
e
r
eg
io
n
o
f
2
S×2
S
o
v
er
t
h
e
s
u
p
e
r
p
ix
el
ce
n
tr
e.
W
h
en
ea
ch
p
ix
e
l
is
co
n
n
ec
ted
w
it
h
th
e
n
ea
r
est
cl
u
s
ter
ce
n
tr
e
t
h
e
n
th
e
cl
u
s
ter
s
ce
n
tr
es
to
t
h
e
m
ea
n
[
l
a
b
x
y
]
v
e
cto
r
ca
n
b
e
ad
ju
s
ted
b
y
an
u
p
d
ate
s
tep
f
o
r
all
th
e
p
ix
els
w
h
ich
b
elo
n
g
s
to
th
e
clu
s
ter
.
A
r
esid
u
al
er
r
o
r
E
is
c
o
m
p
u
ted
b
y
t
h
e
n
o
r
m
b
et
w
ee
n
th
e
p
r
io
r
clu
s
ter
ce
n
tr
e
lo
ca
tio
n
s
a
n
d
th
e
r
ec
en
t
clu
s
ter
ce
n
tr
e
lo
ca
tio
n
s
.
T
h
e
u
p
d
ate
an
d
th
e
ass
i
g
n
m
e
n
t
s
tep
s
m
a
y
b
e
r
ep
ea
ted
iter
ativ
el
y
till
th
e
er
r
o
r
r
ed
u
ce
s
.
At
las
t,
a
p
o
s
t
p
r
o
ce
s
s
in
g
s
tep
is
d
o
n
e
b
y
u
s
i
n
g
th
e
r
ea
s
s
ig
n
i
n
g
d
is
j
o
in
t p
ix
els
to
war
d
s
th
e
n
ea
r
b
y
s
u
p
er
p
ix
els
[
9
]
as sh
o
w
n
i
n
Fi
g
u
r
e
1
.
Fig
u
r
e
4
.
Stan
d
ar
d
K
-
Me
a
n
Se
ar
ch
es a
n
d
S
L
I
C
Sear
ch
es a
L
i
m
i
ted
t
h
e
E
n
tire
I
m
a
g
e
R
eg
io
n
s
[
9
]
2
.
3
.
2
.
Dis
t
a
nce
M
ea
s
ure
SL
I
C
S
u
p
er
p
ix
el
u
s
ed
to
f
o
r
m
a
g
r
o
u
p
in
th
e
p
lan
e
s
p
ac
e
o
f
lab
x
y
co
lo
u
r
i
m
ag
e.
T
h
i
s
i
s
t
h
e
p
r
o
b
lem
f
o
r
d
eter
m
i
n
i
n
g
t
h
e
d
is
ta
n
ce
m
ea
s
u
r
e
D.
D
ca
lcu
lates
t
h
e
d
is
tan
ce
b
et
w
ee
n
th
e
clu
s
ter
ce
n
tr
e
an
d
a
p
ix
el
i.
I
n
th
e
C
I
E
L
A
B
co
lo
u
r
s
p
ac
e
,
th
e
co
lo
u
r
o
f
t
h
e
p
ix
el
s
r
ep
r
esen
t
w
h
o
s
e
r
an
g
e
v
a
lu
e
is
k
n
o
w
n
.
O
n
t
h
e
o
th
er
s
id
e,
th
e
p
ix
el
p
o
s
itio
n
ca
n
u
s
e
th
e
v
al
u
e
r
an
g
e
v
ar
y
a
cc
o
r
d
in
g
to
th
e
i
m
ag
e
s
ize
[
9
]
.
Si
m
p
l
y
D
is
d
e
f
in
ed
as
t
h
e
5
D
E
u
clid
ea
n
d
is
ta
n
ce
in
th
e
s
p
ac
e
o
f
lab
x
y
t
h
at
w
il
l
ca
u
s
e
in
co
n
s
is
te
n
cie
s
in
cl
u
s
ter
in
g
s
u
p
er
p
ix
el
s
ize
s
.
Sp
atial
d
is
tan
ce
s
d
o
m
i
n
ate
co
lo
u
r
p
r
o
x
i
m
i
t
y
f
o
r
lar
g
e
s
u
p
er
p
ix
el
to
g
i
v
e
m
o
r
e
r
ela
tiv
e
v
al
u
e
to
s
p
atia
l
p
r
o
x
i
m
i
t
y
t
h
a
n
co
lo
u
r
.
T
h
e
co
n
v
er
s
e
is
tr
u
e
f
o
r
s
m
all
s
u
p
er
p
ix
el.
I
t
is
e
s
s
e
n
tial
to
n
o
r
m
alize
s
p
atial
p
r
o
x
i
m
it
y
a
n
d
co
lo
u
r
p
r
o
x
i
m
it
y
b
y
t
h
eir
s
p
ec
if
ic
m
ax
i
m
u
m
d
is
tan
ce
s
w
ith
in
t
h
e
cl
u
s
ter
,
a
n
d
an
d
th
e
n
co
m
b
in
e
t
h
e
t
w
o
d
is
tan
ce
s
,
a
f
ter
th
is
D
w
il
l b
e
:
(5
)
(6
)
D’ = √
+
(7
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
2
6
7
4
–
2
6
8
1
2678
T
h
e
m
a
x
i
m
u
m
co
lo
u
r
d
is
tan
c
e
d
eter
m
i
n
atio
n
is
n
o
t
s
tr
ai
g
h
t
f
o
r
w
ar
d
.
T
h
is
p
r
o
b
lem
ca
n
b
e
eli
m
i
n
ated
b
y
f
i
x
in
g
t
o
th
e
co
n
s
ta
n
t
m
.
So
D
’
b
ec
o
m
es
:
D’ = √
+
(8
)
2
.
4
.
F
ea
t
ure
G
en
a
er
a
t
io
n
T
h
e
n
ex
t
s
tep
i
s
f
ea
t
u
r
e
g
e
n
e
r
atio
n
to
d
eter
m
i
n
e
t
h
e
f
ea
t
u
r
e
af
ter
t
h
e
t
h
e
g
e
n
er
atio
n
o
f
s
u
p
er
p
ix
el.
T
h
e
ar
tef
ac
ts
a
n
d
t
h
e
r
ei
n
al
ar
ea
ca
n
d
i
f
f
en
t
iated
b
y
te
x
t
u
r
al
an
d
g
r
ad
ie
n
t b
ased
f
ea
t
u
r
es.
B
y
g
r
ee
n
a
n
d
t
h
e
r
ed
ch
an
n
el
s
o
n
d
i
f
f
er
en
t
s
m
o
o
t
h
i
n
g
s
ca
le
s
,
T
ex
tu
r
al
a
n
d
g
r
ad
ie
n
t
f
ea
t
u
r
es
ar
e
ca
lc
u
lated
[
1
0
]
.
B
lu
e
c
h
an
n
el
m
u
s
t
b
e
ze
r
o
in
S
L
O
i
m
ag
e
s
s
o
t
h
e
r
e
is
n
o
f
ea
t
u
r
e
g
e
n
er
ated
f
o
r
th
e
b
l
u
e
r
eg
io
n
.
T
ex
tu
r
al
f
ea
t
u
r
e
an
d
th
e
g
r
ad
ien
t
f
ea
t
u
r
e
ar
e
d
escr
ib
ed
b
el
o
w
:
a.
T
ex
tu
r
al
f
ea
t
u
r
es: T
h
e
te
x
t
u
r
es a
r
e
ex
a
m
i
n
ed
b
y
s
tatis
t
ical
m
et
h
o
d
th
at
i
s
g
r
a
y
le
v
el
co
-
o
cc
u
r
r
en
ce
m
atr
i
x
(
GL
C
M)
.
G
L
C
M
f
u
n
ctio
n
d
eter
m
i
n
e
h
o
w
o
f
te
n
th
e
i
n
te
n
s
it
y
v
al
u
e
i
ta
k
e
p
lace
w
it
h
t
h
e
ad
j
ac
en
t
i
n
te
n
s
it
y
v
al
u
e
j
.
T
h
e
ad
j
ac
en
t
p
ix
el
ar
e
o
b
s
er
v
ed
b
y
v
th
e
f
o
u
r
an
g
les
i.e
,
θ=
0
º
,
4
5
º
,
9
0
º
,
1
3
5
º
.
T
h
is
is
s
h
o
w
n
i
n
Fig
u
r
e
5
.
O
f
f
s
et
v
al
u
e
i
s
t
h
e
p
i
x
el
ad
j
ac
en
c
y
b
y
t
h
e
p
ar
ticu
lar
d
is
ta
n
ce
.
I
n
t
h
is
p
ap
er
th
e
o
f
f
s
et
v
alu
e
is
k
ep
t
a
s
1
.
G
L
C
M
ca
l
cu
late
v
ar
io
u
s
f
ea
t
u
r
es
b
u
t
f
o
r
r
ed
u
ctiu
o
n
o
f
co
m
p
u
tat
io
n
al
c
o
m
p
le
x
it
y
t
h
i
s
p
ap
er
ex
tr
ac
ted
o
n
l
y
f
o
u
r
f
ea
t
u
r
es
f
r
o
m
G
L
C
M
i.e
,
co
n
tr
as
t,
co
r
r
elatio
n
,
en
er
g
y
an
d
h
o
m
o
g
en
i
t
y
w
h
ic
h
i
s
s
h
o
w
n
in
t
h
e
T
ab
le1
.
b.
Gr
ad
ien
t
f
ea
tu
r
e
s
-
Gr
ad
ien
t
f
ea
tu
r
es
ar
e
n
ec
es
s
ar
y
to
ca
lc
u
late
b
ec
a
u
s
e
o
f
t
h
e
n
o
n
-
u
n
i
f
o
r
m
it
y
o
f
t
h
e
ar
tef
ac
ts
.
T
h
e
r
esp
o
n
s
e
o
f
g
r
ad
ien
t
f
ea
t
u
r
e
ca
lcu
lated
f
r
o
m
th
e
g
au
s
s
ia
n
f
ilter
b
an
k
[
1
0
]
.
I
t
co
n
tain
s
Gau
s
s
ia
n
N(
)
,
f
ir
s
t
o
r
d
er
d
er
iv
ati
v
es
(
)
an
d
(
)
an
d
s
ec
o
n
d
o
r
d
er
d
e
r
iv
ativ
e
(
)
,
(
)
an
d
(
)
,
T
h
e
m
ea
n
v
a
lu
e
is
ca
lc
u
l
ated
b
y
th
e
g
a
u
s
s
ia
n
f
ilter
b
an
k
o
v
er
ea
ch
an
d
ev
er
y
s
u
p
er
p
ix
el
in
all
th
e
p
ix
el
s
.
Me
an
s
tan
d
ar
d
d
ev
i
atio
n
an
d
v
ar
ie
n
ce
ar
e
ca
lcu
lat
ed
in
th
i
s
p
ap
er
w
h
ic
h
is
s
h
o
wn
in
T
ab
le
2
.
(a
)
(b
)
Fig
u
r
e
5
.
(
a)
GL
C
M
P
r
o
ce
s
s
Usi
n
g
I
m
ag
e
I
,
(
b
)
GL
C
M
Dir
ec
t
io
n
an
d
O
f
f
s
et.
[
1
]
T
ab
le
1
.
T
ex
tu
r
al
Featu
r
e
E
x
tr
ac
t
u
r
e
b
y
G
L
C
M
F
e
a
t
u
r
e
n
a
me
Eq
u
a
t
i
o
n
D
e
f
i
n
i
t
i
o
n
C
o
n
t
r
a
st
C
o
n
=
∑
∑
M
e
a
su
r
e
t
h
e
l
o
c
a
l
v
a
r
i
a
n
c
e
i
n
G
L
C
M
C
o
r
r
e
l
a
t
i
o
n
C
o
r
r
=
∑
∑
M
e
a
su
r
e
t
h
e
j
o
i
n
t
p
r
o
b
a
b
i
l
i
t
y
o
c
c
u
r
a
n
c
e
o
f
t
h
e
s
p
e
c
i
f
i
e
d
p
i
x
e
l
p
a
i
r
s.
En
e
r
g
y
E=
∑
∑
P
r
o
v
i
d
e
s t
h
e
s
u
m o
f
s
q
u
a
r
e
d
e
l
e
me
n
t
s i
n
t
h
e
G
L
C
M
.
H
o
mo
g
e
n
e
i
t
y
H
o
mo
m=
∑
∑
p
(
i
,
j
)
M
e
a
su
r
e
s t
h
e
c
l
o
se
n
e
ss o
f
t
h
e
d
i
s
t
r
i
b
u
t
i
o
n
o
f
e
l
e
m
e
n
t
s i
n
t
h
e
G
L
C
M
t
o
t
h
e
G
L
C
M
d
i
a
g
o
n
a
l
.
(
i,j
)
r
e
p
r
esen
t
th
e
r
o
w
s
a
n
d
co
lu
m
n
s
,
p
(
i,j
)
d
en
o
te
th
e
elem
en
t
f
r
o
m
G
L
C
M
m
atr
i
x
,
s
h
o
w
s
t
h
e
n
u
m
b
er
o
f
d
if
f
er
e
n
t
g
r
a
y
lev
e
l
in
an
i
m
a
g
e,
an
d
r
ep
r
esen
t
th
e
m
ar
g
i
n
al
p
r
o
b
ab
ilit
y
w
h
ic
h
is
g
e
n
er
ea
ted
b
y
s
u
m
m
atio
n
o
f
r
o
w
s
an
d
co
lu
m
n
s
o
f
G
L
C
M
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
R
etin
a
l A
r
ea
S
eg
men
t
a
tio
n
u
s
i
n
g
A
d
a
p
tive
S
u
p
erp
ixa
la
tio
n
a
n
d
its
cla
s
s
ifica
tio
n
u
s
in
g
R
B
F
N
(
N
imi
s
h
a
)
2679
T
ab
le
2
.
Gr
a
d
ien
t
Featu
r
e
E
x
t
r
ac
tio
n
F
e
a
t
u
r
e
n
a
me
Eq
u
a
t
i
o
n
D
e
f
i
n
i
t
i
o
n
M
e
a
n
µ
=
∑
A
v
e
r
a
g
e
o
f
t
h
e
v
a
l
u
e
s i
n
t
h
e
d
a
t
a
se
t
.
V
a
r
i
a
n
c
e
∑
A
v
e
r
a
g
e
o
f
t
h
e
sq
u
a
r
e
d
d
i
f
f
r
e
n
c
e
s
b
e
t
w
e
e
n
t
h
e
v
a
l
u
e
s
a
n
d
t
h
e
me
a
n
.
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
=
S
q
u
a
r
e
r
o
o
t
o
f
t
h
e
v
a
r
i
e
n
c
e
.
2
.
5
.
Cla
s
s
if
ier
Co
ns
t
ruct
io
n
Af
ter
t
h
e
f
ea
t
u
r
e
g
e
n
er
atio
n
it
is
n
ec
e
s
s
ar
y
to
d
ev
e
lo
p
a
cla
s
s
i
f
ier
u
s
i
n
g
w
h
ic
h
tr
u
e
r
eti
n
a
l
ar
ea
an
d
ar
tef
ac
t c
an
b
e
d
i
f
f
er
e
n
tiate.
S
o
,
R
B
FN h
a
v
e
b
ee
n
ap
p
lied
.
A
r
ad
ial
b
asi
s
f
u
n
ctio
n
n
et
w
o
r
k
is
a
s
p
ec
ial
t
y
p
e
o
f
n
eu
r
al
n
et
w
o
r
k
w
h
ich
p
er
f
o
r
m
s
t
h
e
class
if
ica
tio
n
u
s
i
n
g
t
h
e
in
p
u
t
th
at
i
s
tak
e
n
f
r
o
m
th
e
tr
ain
i
n
g
s
e
t.
A
“
p
r
o
to
t
y
p
e”
is
s
to
r
ed
b
y
ea
c
h
R
B
FN
n
e
u
r
o
n
.
W
h
e
n
a
n
e
w
d
ata
i
n
p
u
t
h
a
s
to
clas
s
i
f
y
th
e
n
th
e
E
u
clid
ea
n
d
is
tan
ce
i
s
co
m
p
u
ted
b
et
w
ee
n
th
e
p
r
o
to
ty
p
e
a
n
d
th
e
d
ata
in
p
u
t
b
y
ea
ch
n
e
u
r
o
n
.
I
n
s
i
m
p
l
e
w
o
r
d
s
,
if
i
n
p
u
t
is
b
elo
n
g
s
to
clas
s
A
p
r
o
to
ty
p
e
t
h
an
cla
s
s
B
p
r
o
to
ty
p
e
th
e
n
t
h
a
t in
p
u
t i
s
clas
s
if
ie
d
as c
lass
A
[
1
2
]
.
Fig
u
r
e
6
.
R
B
F
Net
w
o
r
k
A
r
ch
i
tectu
r
e
[
1
2
]
T
h
e
ab
o
v
e
F
ig
u
r
e
6
s
h
o
w
s
t
h
e
a
r
ch
itect
u
r
e
o
f
R
B
F
n
et
w
o
r
k
.
T
h
is
n
et
w
o
r
k
co
n
s
is
t
s
o
f
i
n
p
u
t
v
ec
to
r
,
R
B
F n
e
u
r
o
n
s
a
n
d
th
e
o
u
tp
u
t v
ec
to
r
.
I
n
p
u
t
v
ec
to
r
:
C
la
s
s
i
f
icatio
n
o
f
in
p
u
t
v
ec
to
r
is
r
eq
u
ir
ed
w
h
ich
is
n
-
d
i
m
en
s
io
n
al
v
ec
to
r
.
T
h
e
R
B
F
n
eu
r
o
n
s
co
n
tai
n
s
t
h
e
e
n
tire
in
p
u
t v
ec
to
r
.
R
B
F
Ne
u
r
o
n
s
:
A
p
r
o
to
t
y
p
e
i
s
s
to
r
ed
in
ea
ch
R
B
F
n
e
u
r
o
n
s
.
R
B
F
n
e
u
r
o
n
s
ar
e
co
m
p
ar
ed
w
it
h
t
h
e
in
p
u
t
v
ec
to
r
a
n
d
its
p
r
o
to
t
y
p
e.
W
h
en
th
e
in
p
u
t is
eq
u
al
to
i
ts
p
r
o
to
ty
p
e
th
e
n
t
h
e
o
u
tp
u
t
is
ta
k
en
as 1
.
W
h
e
n
t
h
e
d
if
f
er
e
n
ce
b
et
w
ee
n
.
in
p
u
t
a
n
d
its
p
r
o
to
t
y
p
e
in
cr
ea
s
es
t
h
e
n
t
h
e
o
u
tp
u
t
r
e
s
p
o
n
s
e
f
all
s
o
f
f
0
.
So
th
e
o
u
tp
u
t
v
al
u
e
lies
b
et
w
ee
n
0
an
d
1
w
h
ic
h
d
ep
en
d
s
u
p
o
n
t
h
e
s
i
m
ilar
it
y
.
T
h
e
R
B
F
n
eu
r
o
n
’
s
s
h
ap
e
is
b
ell
lik
e
s
tr
u
ctu
r
e
w
h
ic
h
is
s
h
o
w
n
i
n
th
e
ab
o
v
e
F
ig
u
r
e
6.
Ou
tp
u
t
n
o
d
es:
A
s
e
t
o
f
n
o
d
es
is
co
n
tain
ed
in
t
h
e
o
u
tp
u
t
n
e
t
w
o
r
k
.
Var
io
u
s
s
co
r
es
ar
e
co
m
p
u
ted
b
y
ea
ch
o
u
tp
u
t
n
o
d
e.
T
h
e
s
co
r
e
is
ca
lcu
lated
u
s
i
n
g
th
e
w
eig
h
t
ed
s
u
m
b
y
t
h
e
ac
tiv
a
tio
n
f
u
n
c
tio
n
f
r
o
m
th
e
R
B
F
n
eu
r
o
n
.
A
n
o
u
tp
u
t
n
o
d
e
is
c
o
n
n
ec
ted
w
i
th
w
ei
g
h
t
v
a
lu
e
o
f
t
h
e
R
B
F
n
e
u
r
o
n
,
b
e
f
o
r
e
ad
d
in
g
t
h
e
ac
t
iv
at
io
n
v
alu
e
o
f
th
e
n
e
u
r
o
n
is
m
u
l
tip
li
ed
w
it
h
th
e
w
ei
g
h
t.
E
ac
h
o
u
tp
u
t
n
o
d
e
co
n
tai
n
it
s
o
w
n
w
ei
g
h
t
b
ec
a
u
s
e
t
h
e
s
co
r
e
f
o
r
all
ca
teg
o
r
y
is
co
m
p
u
te
d
b
y
ea
c
h
o
u
tp
u
t.
T
h
e
o
u
tp
u
t
n
o
d
e
p
r
o
v
id
e
p
o
s
itiv
e
w
eig
h
t
a
s
w
ell
as
n
e
g
ati
v
e
w
ei
g
h
t.
P
o
s
itiv
e
w
ei
g
h
t
h
as
it
s
o
n
e
ca
teg
o
r
y
an
d
n
eg
at
iv
e
w
ei
g
h
t
h
a
s
a
n
o
th
er
ca
teg
o
r
y
.
F(x
)
=
(
9
)
A
b
o
v
e
eq
u
at
io
n
is
f
o
r
o
n
e
d
i
m
en
s
io
n
al
in
p
u
t
w
h
er
e
µ=
m
ea
n
=
s
tan
d
ar
d
d
ev
iatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
2
6
7
4
–
2
6
8
1
2680
x
=
in
p
u
t
Sli
g
h
t d
i
f
f
er
e
n
t f
u
n
ctio
n
i
s
u
s
e
d
in
R
B
F a
ctiv
a
tio
n
f
u
n
ctio
n
:
=
(
1
0
)
µ
r
ef
er
s
to
m
ea
n
in
t
h
e
g
au
s
s
ian
b
u
t
h
er
e
µ
i
s
t
h
e
p
r
o
to
t
y
p
e
v
ec
to
r
w
h
ic
h
i
s
b
elo
n
g
s
th
e
b
ell
cu
r
v
e
ce
n
ter
.
Her
e
is
u
s
ed
f
o
r
th
e
ac
ti
v
ati
o
n
f
u
ct
io
n
.
R
B
F
Neu
r
o
n
A
cti
v
atio
n
F
u
n
ctio
n
:
T
h
e
m
ea
s
u
r
e
m
en
t
o
f
eq
u
alit
y
b
et
w
ee
n
th
e
p
r
o
to
t
y
p
e
an
d
th
e
in
p
u
t
v
ec
to
r
is
ca
lcu
la
ted
b
y
R
B
F n
e
u
r
o
n
[
1
2
]
.
3.
RE
SU
L
T
S
W
e
p
er
f
o
r
m
ed
t
h
e
e
x
p
er
i
m
e
n
t
o
n
an
i
m
a
g
e
w
h
ic
h
i
s
o
b
tain
ed
f
r
o
m
Op
to
s
[
4
]
an
d
t
h
at
ar
e
ac
q
u
ir
ed
b
y
t
h
e
u
ltra
w
id
e
f
ield
SL
O.
Fi
eld
o
f
v
ie
w
(
FO
V)
o
f
ea
ch
r
etin
al
i
m
a
g
e
as
s
h
o
w
n
i
n
Fi
g
u
r
e
7
is
u
p
to
2
0
0
º
w
it
h
th
e
r
eso
l
u
tio
n
o
f
1
4
µ
m
.
T
h
e
r
etin
al
i
m
a
g
e
w
it
h
o
u
t
d
ilatio
n
i
s
ca
p
tu
r
ed
b
y
t
h
e
d
ev
ice,
o
v
er
th
e
p
u
p
il
o
f
2
m
m
th
at
i
s
v
er
y
s
m
all.
T
h
er
e
ar
e
t
w
o
c
h
an
n
el
s
i
n
t
h
e
i
m
a
g
e
i.e
,
r
ed
an
d
g
r
ee
n
.
T
h
e
r
ed
ch
a
n
n
e
l
w
h
o
s
e
w
av
e
len
g
t
h
is
6
3
3
n
m
d
is
p
la
y
s
th
e
d
ee
p
er
s
tr
u
ctu
r
es
o
f
r
eti
n
a
ag
a
i
n
s
t
c
h
o
r
o
id
w
h
er
ea
s
t
h
e
g
r
ee
n
ch
a
n
n
e
l
w
h
o
s
e
w
a
v
ele
n
g
t
h
is
532
n
m
g
i
v
e
s
in
f
o
r
m
a
tio
n
ab
o
u
t
t
h
e
r
etin
a
l
p
ig
m
e
n
t
ep
it
h
eli
u
m
to
th
e
s
en
s
o
r
y
r
eti
n
a.
T
h
e
d
i
m
en
s
io
n
o
f
ea
c
h
i
m
a
g
e
is
3
9
0
0
×3
0
7
2
an
d
ea
ch
p
ix
el
is
s
h
o
w
e
d
b
y
8
b
it
o
n
b
o
th
g
r
ee
n
a
n
d
r
ed
ch
a
n
n
els.
T
h
e
d
ataset
is
co
llected
b
y
d
is
ea
s
ed
an
d
h
ea
lth
y
r
eti
n
al
i
m
a
g
es.
Ma
n
y
d
is
ea
s
ed
r
etin
al
i
m
a
g
es
ar
e
co
llected
f
r
o
m
D
iab
etic
R
eti
n
o
p
ath
y
p
atie
n
ts
[
3
]
.
T
h
e
r
esu
lt
s
an
d
th
e
ac
c
u
r
ac
ies
o
f
R
B
FN
b
y
d
ice
-
co
e
f
f
icien
t
lik
e
ev
alu
at
io
n
m
etr
ic.
T
h
e
a
m
o
u
n
t o
f
o
v
er
lap
b
et
w
ee
n
t
h
e
b
en
c
h
m
ar
k
ac
q
u
ir
ed
b
y
t
h
e
cli
n
icia
n
a
n
d
o
u
tp
u
t o
f
f
r
am
e
w
o
r
k
is
ca
l
led
as
Dice
co
ef
f
ic
ien
t.
Dice
co
e
f
f
ici
en
t c
an
b
e
w
r
itte
n
as
:
w
h
er
e,
A
=
i
m
a
g
e
o
b
tain
ed
b
y
f
r
a
m
e
wo
r
k
B
=
im
a
g
e
o
b
tain
ed
b
y
b
en
c
h
m
ar
k
in
ter
s
ec
tio
n
an
d
v
a
lu
e
s
v
ar
ie
s
in
0
an
d
1
w
h
er
e
1
is
th
e
h
ig
h
e
r
v
alu
e
a
n
d
0
is
th
e
lo
w
er
v
al
u
e
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
(
f
)
Fig
u
r
e
7
.
(
a)
an
d
(
d
)
R
ep
r
e
s
en
t th
e
Te
s
t (
Or
ig
i
n
al)
Im
a
g
e
f
r
o
m
S
L
O
I
m
a
g
es (
b
)
an
d
(
e
)
R
ep
r
esen
t th
e
Ga
mma
No
r
m
alize
d
o
n
t
h
e
Or
ig
i
n
al
I
m
a
g
es (
c
)
an
d
(
f
)
R
ep
r
esen
t t
h
e
S
u
p
er
p
ix
el
C
la
s
s
i
f
ic
atio
n
R
e
s
u
lt
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
R
etin
a
l A
r
ea
S
eg
men
t
a
tio
n
u
s
i
n
g
A
d
a
p
tive
S
u
p
erp
ixa
la
tio
n
a
n
d
its
cla
s
s
ifica
tio
n
u
s
in
g
R
B
F
N
(
N
imi
s
h
a
)
2681
4.
DIS
CU
SS
I
O
N
AND
CO
NC
L
US
I
O
N
E
x
tr
ac
tio
n
o
f
r
eti
n
al
ar
ea
f
r
o
m
th
e
SLO
i
m
a
g
es
ar
e
v
er
y
i
m
o
r
tan
t
b
u
t
d
if
f
ic
u
lt
tas
k
.
I
n
th
is
s
tu
d
y
,
w
e
p
r
o
p
o
s
ed
a
tech
n
iq
u
e
u
s
in
g
wh
ich
tr
u
e
r
eti
n
al
ar
ea
is
d
etec
t
ed
f
r
o
m
th
e
S
L
O
i
m
ag
e
s
.
W
e
h
av
e
p
r
ese
n
ted
th
e
p
r
ep
r
o
ce
s
s
in
g
,
s
u
p
er
p
ix
el
g
en
er
atio
n
,
f
ea
tu
r
e
g
e
n
er
atio
n
an
d
class
if
ier
co
n
s
tr
u
ctio
n
.
SL
I
C
O
is
m
o
r
e
f
a
s
ter
,
b
etter
p
er
f
o
r
m
an
ce
an
d
m
o
r
e
m
e
m
o
r
y
e
f
f
icie
n
t
f
o
r
g
e
n
er
atio
n
o
f
s
u
p
er
p
ix
el
.
Feat
u
r
e
g
en
er
atio
n
i
s
v
er
y
i
m
p
o
r
tan
t
s
tep
to
r
ed
u
ce
th
e
co
m
p
u
tatio
n
al
co
s
t.
T
h
e
n
cla
s
s
i
f
ier
co
n
s
tr
u
ct
io
n
i
s
u
s
ed
to
ex
tr
ac
t
t
h
e
r
eti
n
al
ar
ea
.
T
h
e
R
B
FN
class
i
f
ier
h
as
b
ee
n
ap
p
lied
t
o
ac
h
iev
e
g
r
ea
t
er
ac
cu
r
ac
y
.
P
r
o
p
o
s
ed
w
o
r
k
h
as
b
ee
n
ap
p
lied
o
n
3
0
im
a
g
es
an
d
f
r
o
m
all
t
h
es
e
i
m
ag
e
s
it
h
as
b
ee
n
co
n
cl
u
d
ed
th
at
th
e
ef
f
ic
ien
c
y
ac
h
ie
v
ed
is
9
6
%
b
u
t
th
e
co
m
p
u
tatio
n
al
ti
m
e
i
s
litt
le
h
i
g
h
er
i
n
ca
s
e
o
f
R
B
FN
w
i
th
co
m
p
ar
ed
to
A
N
N.
T
h
is
co
m
p
u
t
atio
n
al
ti
m
e
ca
n
b
e
r
ed
u
ce
d
in
th
e
f
u
t
u
r
e.
RE
F
E
R
E
NC
E
S
[1
]
M
.
J.
A
li
g
h
o
li
z
a
d
e
h
,
S
.
Ja
v
a
d
i,
R
.
S
.
Na
d
o
o
s
h
a
n
,
a
n
d
K.
Ka
n
g
a
rlo
o
,
“
Ey
e
li
d
a
n
d
e
y
e
la
sh
se
g
m
e
n
tatio
n
b
a
se
d
o
n
w
a
v
e
let
tran
s
f
o
r
m
f
o
r
iri
s rec
o
g
n
it
io
n
”
,
in
Pro
c
.
4
th
In
t.
Co
n
g
r.
Ima
g
e
S
ig
n
a
l
Pro
c
e
ss
,
p
p
.
1
2
3
1
–
1
2
3
5
,
2
0
1
1
.
[2
]
T
h
o
m
a
s
M
.
De
se
rn
o
,
”
F
u
n
d
a
m
e
n
tals
o
f
Bio
m
e
d
ica
l
I
m
a
g
e
P
ro
c
e
ss
in
g
”
,
in
S
p
rin
g
e
r
-
Ver
la
g
Ber
li
n
He
id
e
lb
e
rg
o
f
Bi
o
me
d
ica
l
Ima
g
e
Pro
c
e
ss
in
g
,
2
0
1
1
.
[3
]
Ha
lee
m
,
M
.
S
.
,
Ha
n
,
L
.
,
v
a
n
He
m
e
rt,
J.,
L
i,
B.
a
n
d
F
lem
in
g
,
A
.
,
“
Re
ti
n
a
l
A
re
a
De
tec
to
r
f
ro
m
S
c
a
n
n
in
g
L
a
se
r
Op
h
th
a
lm
o
sc
o
p
e
(S
L
O)
Im
a
g
e
s
f
o
r
Dia
g
n
o
sin
g
Re
ti
n
a
l
Dise
a
se
s”
,
IEE
E
jo
u
rn
a
l
o
f
b
io
me
d
i
c
a
l
a
n
d
h
e
a
lt
h
in
fo
rm
a
ti
c
s
,
1
9
(4
)
,
p
p
.
1
4
7
2
-
1
4
8
2
,
2
0
1
5
.
[4
]
Op
to
s.
(
2
0
1
4
)
.
[
On
l
in
e
]
.
A
v
a
il
a
b
le:
ww
w
.
o
p
to
s.co
m
[5
]
S
e
e
m
a
Ra
n
i,
M
a
n
o
j
Ku
m
a
r,
“
Co
n
tras
t
En
h
a
n
c
e
m
e
n
t
u
sin
g
Im
p
ro
v
e
d
A
d
a
p
ti
v
e
Ga
m
m
a
Co
rre
c
ti
o
n
W
it
h
W
e
ig
h
ti
n
g
Distrib
u
ti
o
n
T
e
c
h
n
iq
u
e
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
Ap
p
li
c
a
ti
o
n
s,
1
0
1
(
1
1
)
,
p
p
.
0
9
7
5
–
8
8
8
7
,
2
0
1
4
.
[6
]
Ch
o
ll
e
tt
e
C
.
Ch
u
d
e
-
Olisa
h
l,
G
h
a
z
a
li
S
Ulo
n
g
,
Uc
h
e
A
.
K.
Ch
u
d
e
-
Ok
o
n
k
w
o
,
S
it
i
z
.
M
.
Ha
sh
im
,
”
Ill
u
m
in
a
ti
o
n
No
rm
a
li
z
a
ti
o
n
f
o
r
Ed
g
e
-
Ba
se
d
F
a
c
e
Re
c
o
g
n
it
io
n
Us
in
g
th
e
F
u
sio
n
o
f
R
G
B
No
r
m
a
li
z
a
ti
o
n
a
n
d
G
a
m
m
a
Co
rre
c
ti
o
n
”
,
IEE
E
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
S
i
g
n
a
l
a
n
d
Irn
a
g
e
Pro
c
e
ss
in
g
A
p
p
li
c
a
ti
o
n
s
,
2
0
1
3
[7
]
A
.
K.
Bh
a
n
d
a
ria
,
A
.
Ku
m
a
r
a
,
G
.
K
.
S
in
g
h
,
”
Im
p
ro
v
e
d
k
n
e
e
tran
sf
e
r
f
u
n
c
ti
o
n
a
n
d
g
a
m
m
a
c
o
rre
c
ti
o
n
b
a
se
d
m
e
th
o
d
f
o
r
c
o
n
tras
t
a
n
d
b
r
ig
h
tn
e
ss
e
n
h
a
n
c
e
m
e
n
t
o
f
sa
telli
te
i
m
a
g
e
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
tro
n
ics
a
n
d
Co
mm
u
n
ic
a
ti
o
n
s
(
AE
Ü)
,
6
9
,
p
p
.
5
7
9
–
5
8
9
,
2
0
1
5
.
[8
]
h
tt
p
:
//
iv
rl.
e
p
f
l.
c
h
/res
e
a
rc
h
/su
p
e
rp
i
x
e
ls#
S
L
ICO
[9
]
Ra
d
h
a
k
rish
n
a
A
c
h
a
n
ta,
A
p
p
u
S
h
a
ji
,
Ke
v
in
S
m
it
h
,
A
u
re
li
e
n
L
u
c
c
h
i,
P
a
sc
a
l
F
u
a
,
a
n
d
S
a
b
i
n
e
S
u
,
“
S
L
IC
S
u
p
e
rp
ix
e
ls
Co
m
p
a
re
d
to
S
tate
-
of
-
th
e
-
A
rt
S
u
p
e
r
p
ix
e
l
M
e
th
o
d
s”
,
IEE
E
tr
a
n
sa
c
ti
o
n
s
o
n
p
a
tt
e
rn
a
n
a
lys
is
a
n
d
ma
c
h
in
e
in
telli
g
e
n
c
e
,
3
4
(1
1
),
2
2
7
4
-
2
2
8
1
,
2
0
1
2
[1
0
]
M
.
A
b
ra
`m
o
ff
,
W
.
A
l
w
a
rd
,
E.
G
re
e
n
lee
,
L
.
S
h
u
b
a
,
C.
Kim
,
J.
F
in
g
e
rt,
a
n
d
Y.
Kw
o
n
,
“
A
u
to
m
a
ted
se
g
m
e
n
tatio
n
o
f
th
e
o
p
ti
c
d
isc
f
ro
m
ste
r
e
o
c
o
lo
r
p
h
o
t
o
g
ra
p
h
s u
sin
g
p
h
y
sio
lo
g
ica
ll
y
p
lau
sib
le f
e
a
tu
re
s
”
,
In
v
e
st.
Op
h
th
a
lmo
l.
V
is.
S
c
i.
,
v
o
l.
4
8
,
p
p
.
1
6
6
5
–
1
6
7
3
,
2
0
0
7
[1
1
]
h
tt
p
:
//
in
.
m
a
th
w
o
rk
s.co
m
/h
e
lp
/i
m
a
g
e
s/re
f
/g
r
a
y
c
o
m
a
tri
x
.
h
tm
l#
re
sp
o
n
siv
e
_
o
ff
c
a
n
v
a
s
[1
2
]
h
tt
p
:
//
m
c
c
o
r
m
ic
k
m
l.
c
o
m
/2
0
1
3
/0
8
/1
5
/ra
d
ial
-
b
a
sis
-
f
u
n
c
ti
o
n
-
n
e
tw
o
rk
-
rb
f
n
-
tu
to
r
ial/
[1
3
]
M
.
S
.
Ha
lee
m
,
L
.
Ha
n
,
J.
v
a
n
He
m
e
rt,
a
n
d
B.
L
i,
“
A
u
to
m
a
ti
c
e
x
trac
ti
o
n
o
f
re
ti
n
a
l
f
e
a
tu
re
s
f
ro
m
c
o
lo
u
r
re
ti
n
a
l
im
a
g
e
s
f
o
r
g
lau
c
o
m
a
d
iag
n
o
sis: A
re
v
ie
w
”
,
Co
mp
u
t.
M
e
d
.
Im
a
g
.
Gr
a
p
h
.
3
7
,
p
p
.
5
8
1
–
5
9
6
,
2
0
1
3
.
[1
4
]
R.
C.
G
o
n
z
a
lez
a
n
d
R
.
E.
W
o
o
d
s,
Ed
s.
,
Di
g
it
a
l
Ima
g
e
Pr
o
c
e
ss
in
g
,
3
r
d
e
d
.
E
n
g
lew
o
o
d
Cli
f
f
s,
NJ
,
USA
:
P
re
n
ti
c
e
-
Ha
ll
,
2
0
0
6
.
[1
5
]
D.
Zh
a
n
g
,
D.
M
o
n
ro
,
a
n
d
S
.
Ra
k
sh
it
,
“
Ey
e
las
h
re
m
o
v
a
l
m
e
th
o
d
f
o
r
h
u
m
a
n
iri
s
re
c
o
g
n
it
io
n
”
,
i
n
Pro
c
.
IEE
E
In
t
.
Co
n
f.
Ima
g
e
Pro
c
e
ss
.
,
p
p
.
2
8
5
–
2
8
8
,
2
0
0
6
.
[1
6
]
A
.
V
.
M
ire
a
n
d
B.
L
.
Dh
o
te,
“
Iris
re
c
o
g
n
it
io
n
sy
ste
m
w
it
h
a
c
c
u
ra
te
e
y
e
las
h
se
g
m
e
n
tatio
n
a
n
d
im
p
ro
v
e
d
F
A
R,
F
RR
u
sin
g
tex
tu
ra
l
a
n
d
to
p
o
lo
g
ica
l
f
e
a
tu
re
s”
,
In
t.
J
.
Co
m
p
u
t.
Ap
p
l.
,
7
,
p
p
.
0
9
7
5
–
8
8
8
7
,
2
0
1
0
.
[1
7
]
Y.H.
L
i,
M
.
S
a
v
v
id
e
s,
a
n
d
T
.
Ch
e
n
,
“
In
v
e
stig
a
ti
n
g
u
se
f
u
l
a
n
d
d
isti
n
g
u
ish
in
g
f
e
a
tu
re
s
a
ro
u
n
d
th
e
e
y
e
l
a
sh
re
g
io
n
”
,
in
Pro
c
.
3
7
th
IEE
E
W
o
rk
sh
o
p
Ap
p
l.
Ima
g
.
Pa
tt
e
rn
Rec
o
g
.
,
p
p
.
1
–
6
,
2
0
0
8
[1
8
]
B.
J.
Ka
n
g
a
n
d
K.R.
P
a
rk
,
“
A
ro
b
u
st
e
y
e
las
h
d
e
tec
ti
o
n
b
a
se
d
o
n
ir
i
s
f
o
c
u
s
a
s
se
ss
m
e
n
t
”
,
Pa
tt
e
rn
Rec
o
g
.
L
e
tt
.
,
2
8
,
p
p
.
1
6
3
0
–
1
6
3
9
,
2
0
0
7
.
[1
9
]
T
.
H.
M
in
a
n
d
R.
H.
P
a
rk
,
“
Ey
e
li
d
a
n
d
e
y
e
la
sh
d
e
tec
ti
o
n
m
e
th
o
d
in
t
h
e
n
o
rm
a
li
z
e
d
iri
s
ima
g
e
u
si
n
g
th
e
p
a
ra
b
o
li
c
Ho
u
g
h
m
o
d
e
l
a
n
d
Otsu
s t
h
re
sh
o
l
d
in
g
m
e
th
o
d
”
,
P
a
tt
e
rn
Rec
o
g
.
L
e
t
t.
3
0
,
p
p
.
1
1
38
–
1
1
4
3
,
2
0
0
9
.
[2
0
]
H.
Da
v
is,
S
.
Ru
ss
e
ll
,
E
.
Ba
rrig
a
,
M.
A
b
ra
m
o
ff
,
a
n
d
P
.
S
o
l
iz,
“
V
isio
n
-
b
a
se
d
,
re
a
l
-
ti
m
e
r
e
ti
n
a
l
im
a
g
e
q
u
a
li
t
y
a
ss
e
ss
m
e
n
t
”
,
in
Pro
c
.
2
2
n
d
IEE
E
In
t.
S
y
mp
.
C
o
mp
u
t.
-
B
a
se
d
M
e
d
.
S
y
st.
,
p
p
.
1
–
6
,
2
0
Evaluation Warning : The document was created with Spire.PDF for Python.