I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
5
,
No
.
2
,
J
u
n
e
2016
,
p
p
.
45
~
54
I
SS
N:
2252
-
8938
45
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
J
A
I
Auto
m
a
tic
Ex
uda
tes De
tectio
n
i
n
D
ia
betic
R
etinopa
t
hy
I
m
a
g
es
H.
F
a
o
uzi
,
M.
F
a
k
ir
In
f
o
rm
a
ti
o
n
a
n
d
p
r
o
c
e
ss
in
g
tea
m
s
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s an
d
T
e
c
h
n
ics
,
S
u
lt
a
n
M
o
u
lay
S
li
m
a
n
e
Un
iv
e
r
sity
,
M
o
ro
c
c
o
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
2
,
2
0
1
6
R
ev
i
s
ed
Ma
y
6
,
2
0
1
6
A
cc
ep
ted
Ma
y
2
3
,
2
0
1
6
Dia
b
e
ti
c
Re
ti
n
o
p
a
t
h
y
(DR)
re
fe
r
s
to
th
e
p
re
se
n
c
e
o
f
t
y
p
ica
l
re
ti
n
a
l
m
icro
v
a
sc
u
lar
les
io
n
s
in
p
e
rso
n
s
w
it
h
d
iab
e
ti
c
s.
W
h
e
n
th
e
d
ise
a
se
is
a
t
th
e
e
a
rl
y
sta
te,
a
p
ro
m
p
t
d
iag
n
o
sis
m
a
y
h
e
lp
in
p
re
v
e
n
ti
n
g
irrev
e
rsib
le
d
a
m
a
g
e
s
to
th
e
d
iab
e
ti
c
e
y
e
.
I
f
th
e
e
x
u
d
a
te
s are
c
lo
se
r
to
m
a
c
u
la,
th
e
n
th
e
situ
a
ti
o
n
is
c
rit
ica
l.
Early
d
e
te
c
ti
o
n
c
a
n
p
o
ten
ti
a
ll
y
re
d
u
c
e
th
e
risk
o
f
b
li
n
d
.
T
h
is
p
a
p
e
r
p
ro
p
o
se
s
to
o
l
f
o
r
th
e
e
a
rly
d
e
tec
ti
o
n
o
f
Dia
b
e
ti
c
Re
ti
n
o
p
a
t
h
y
u
sin
g
e
d
g
e
d
e
tec
ti
o
n
,
a
lg
o
rit
h
m
k
m
e
a
n
s
in
se
g
m
e
n
tatio
n
p
h
a
se
,
i
n
v
a
ri
a
n
t
m
o
m
e
n
ts
(Hu
a
n
d
A
ff
in
e
)
a
n
d
d
e
sc
rip
to
r
G
IS
T
in
e
x
trac
ti
o
n
p
h
a
se
.
I
n
t
h
e
re
c
o
g
n
it
i
o
n
p
h
a
se
,
n
e
u
ra
l
n
e
tw
o
rk
is
a
d
o
p
ted
.
A
ll
tes
ts
a
re
a
p
p
li
e
d
o
n
d
a
tab
a
se
DIA
RE
T
DB1
.
K
ey
w
o
r
d
:
Dete
ctio
n
Diab
etic
R
eti
n
o
p
ath
y
Fu
zz
y
c
-
m
ea
n
s
Gip
t d
escr
ip
to
r
s
N
eu
r
al
n
et
w
o
r
k
.
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
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
H.
Fa
o
u
zi
,
I
n
f
o
r
m
a
tio
n
a
n
d
p
r
o
ce
s
s
in
g
te
a
m
s
,
Facu
lt
y
o
f
Sc
ien
ce
s
an
d
T
ec
h
n
ics,
Su
lta
n
Mo
u
la
y
Sl
i
m
a
n
e
U
n
i
v
e
r
s
it
y
,
E
m
ail:
i
n
f
o
.
d
ec
0
7
@
y
ah
o
o
.
f
r
1.
I
NT
RO
D
UCT
I
O
N
R
eti
n
al
i
m
a
g
e
an
a
l
y
s
is
is
a
n
ess
e
n
tial
s
tep
i
n
t
h
e
d
ia
g
n
o
s
is
o
f
v
ar
io
u
s
e
y
e
d
is
ea
s
es
.
Diab
etic
r
etin
o
p
ath
y
i
s
a
co
m
p
licatio
n
o
f
d
iab
etes
th
at
is
ca
u
s
ed
b
y
ch
an
g
es
in
t
h
e
b
lo
o
d
v
ess
el
s
o
f
th
e
r
etin
a.
T
h
e
s
y
m
p
to
m
s
ca
n
b
lu
r
o
r
d
is
to
r
t
th
e
p
atie
n
t
’
s
v
i
s
io
n
a
n
d
ar
e
a
m
ai
n
ca
u
s
e
o
f
b
li
n
d
n
es
s
.
E
x
u
d
ates
ar
e
o
n
e
o
f
t
h
e
p
r
im
ar
y
s
i
g
n
s
o
f
d
iab
etic
r
eti
n
o
p
ath
y
.
Me
d
ical
I
m
ag
e
Se
g
m
e
n
tat
io
n
is
th
e
p
r
o
ce
s
s
o
f
au
to
m
at
ic
o
r
s
e
m
i
-
au
to
m
at
ic
d
etec
tio
n
o
f
b
o
u
n
d
ar
ies
w
ith
in
a
2
D
o
r
3
D
i
m
a
g
e
.
I
m
a
g
e
Se
g
m
en
tatio
n
i
s
a
p
r
o
ce
s
s
f
o
r
d
iv
id
i
n
g
a
g
iv
e
n
i
m
a
g
e
in
to
m
ea
n
i
n
g
f
u
l
r
eg
io
n
s
w
it
h
h
o
m
o
g
e
n
eo
u
s
p
r
o
p
er
ties
.
A
m
aj
o
r
d
if
f
icu
lt
y
o
f
m
ed
ical
i
m
a
g
e
s
eg
m
e
n
tatio
n
i
s
t
h
e
h
i
g
h
v
ar
iab
ilit
y
i
n
m
ed
ical
i
m
a
g
e
s
.
T
h
e
tech
n
iq
u
e
s
p
r
esen
ted
i
n
th
is
p
ap
er
ca
n
b
e
clas
s
i
f
ied
in
to
th
r
ee
ca
teg
o
r
ies
:
ass
ess
m
en
t
an
d
i
m
p
r
o
v
e
m
e
n
t
o
f
th
e
i
m
a
g
e
q
u
alit
y
,
s
eg
m
e
n
tatio
n
o
f
e
x
u
d
ate
s
an
d
r
ec
o
g
n
itio
n
[
2
,
3
,
5
]
.
2.
P
RE
P
RO
CE
SS
I
NG
Af
ter
th
e
ac
q
u
is
i
tio
n
o
f
t
h
e
i
m
ag
e,
r
ec
o
g
n
it
io
n
s
y
s
te
m
b
eg
in
s
w
ith
t
h
e
p
r
e
-
p
r
o
ce
s
s
i
n
g
m
et
h
o
d
co
m
p
r
is
in
g
t
h
e
f
o
llo
w
i
n
g
f
u
n
c
tio
n
s
:
2
.
1
.
Cho
o
s
ing
a
Co
l
o
r
Sp
a
ce
T
h
e
co
lo
r
o
f
o
u
r
d
atab
ase
i
m
a
g
es
ar
e
ac
q
u
ir
ed
in
th
e
R
GB
s
p
ac
e
[
4
]
.
W
e
ca
n
p
er
f
o
r
m
th
e
tr
ea
t
m
e
n
t
in
t
h
is
s
p
ac
e.
T
h
is
ap
p
r
o
ac
h
h
as th
e
ad
v
a
n
ta
g
e
o
f
e
n
s
u
r
in
g
d
ata
in
te
g
r
it
y
,
b
u
t it
h
a
s
a
n
u
m
b
er
o
f
d
is
ad
v
an
ta
g
es
a
m
o
n
g
w
h
ic
h
t
h
e
s
a
m
e
co
lo
r
ca
n
b
e
m
ad
e
q
u
ite
d
i
f
f
er
e
n
tl
y
b
y
t
w
o
s
cr
ee
n
s
.
W
e
ca
n
t
h
en
lo
o
k
f
o
r
a
m
o
r
e
ap
p
r
o
p
r
iate
s
p
ac
e.
T
h
e
tr
ea
tm
en
t i
s
p
er
f
o
r
m
ed
i
n
t
h
e
n
e
w
s
p
ac
e
(
af
ter
a
tr
a
n
s
f
o
r
m
atio
n
T
)
,
an
d
t
h
e
ap
p
licatio
n
o
f
th
e
i
n
v
er
s
e
tr
an
s
f
o
r
m
a
tio
n
(
T
-
1
,
if
it e
x
is
t
s
)
is
u
s
ed
,
if
n
ec
ess
ar
y
,
f
o
r
r
etu
r
n
to
th
e
R
GB
s
p
ac
e
(
Fig
u
r
e
1
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
45
–
54
46
Fig
u
r
e
1
.
T
r
ea
tm
e
n
t a
f
ter
p
r
o
jectio
n
in
a
s
u
itab
le
w
o
r
k
s
p
ac
e
2
.
2
.
RG
B
t
o
H
SI
Sp
a
ce
HSI
C
o
lo
r
Mo
d
el
H
u
e,
Sa
tu
r
at
io
n
a
n
d
I
n
te
n
s
i
t
y
ar
e
t
h
r
ee
i
m
p
o
r
tan
t d
escr
ip
to
r
s
u
s
ed
b
y
h
u
m
an
b
ein
g
in
d
escr
ib
in
g
co
lo
r
s
.
Fo
r
ex
a
m
p
le
w
h
e
n
w
e
d
escr
ib
e
th
e
co
lo
r
o
f
r
etin
a
o
r
o
th
er
o
b
j
ec
t
w
e
n
ev
er
s
a
y
t
h
at
h
o
w
r
ed
,
g
r
ee
n
o
r
b
lu
e
th
e
r
etin
a
will
b
e.
I
n
s
tead
w
e
d
escr
ib
e
t
h
e
f
ac
e
in
ter
m
o
f
t
h
e
b
r
ig
h
t
n
ess
,
th
e
co
lo
r
,
an
d
also
h
o
w
d
ar
k
o
r
lig
h
t
t
h
e
co
lo
r
w
il
l
b
e.
T
h
is
is
ex
ac
tl
y
w
h
at
HSI
co
lo
r
s
p
ac
e
d
o
w
h
er
e
H
s
tan
d
s
f
o
r
h
u
e
w
i
ll
d
ef
in
e
t
h
e
co
lo
r
,
S
s
tan
d
s
f
o
r
s
atu
r
atio
n
w
ill
d
ef
in
e
h
o
w
s
tr
o
n
g
t
h
at
p
ar
ticu
lar
co
lo
r
w
ill
b
e
an
d
I
th
e
in
te
n
s
it
ies
w
h
ic
h
s
p
ec
i
f
ied
t
h
e
b
r
ig
h
t
n
es
s
.
T
o
s
ep
ar
ate
th
e
co
lo
r
co
m
p
o
n
e
n
ts
w
it
h
i
n
ten
s
it
y
an
d
b
r
ig
h
t
n
es
s
,
th
e
co
m
p
o
n
e
n
t r
eti
n
al
in
p
u
t i
m
ag
e
s
i
n
th
e
R
GB
co
lo
r
s
p
ac
e
ar
e
co
n
v
er
ted
to
th
e
HSI
co
lo
r
s
p
ac
e
(
Fig
u
r
e
2
)
.
Fig
u
r
e
2
.
R
GB
to
HSI
s
p
ac
e
2
.
3
.
M
edia
n F
ilte
ring
Me
d
ian
f
il
ter
in
g
is
a
n
o
n
li
n
ea
r
p
r
o
ce
s
s
u
s
ef
u
l
in
r
ed
u
ci
n
g
i
m
p
u
l
s
i
v
e
o
r
s
alt
-
a
n
d
-
p
ep
p
er
n
o
is
e
[
1
]
.
I
t
is
also
u
s
ef
u
l
in
p
r
eser
v
i
n
g
ed
g
es
in
a
n
i
m
a
g
e
w
h
i
le
r
ed
u
ci
n
g
r
an
d
o
m
n
o
i
s
e.
T
h
e
m
ed
ian
i
s
ca
lc
u
lated
b
y
f
ir
s
t
s
o
r
tin
g
all
t
h
e
p
ix
e
l
v
a
lu
e
s
f
r
o
m
th
e
s
u
r
r
o
u
n
d
i
n
g
n
ei
g
h
b
o
r
h
o
o
d
in
to
n
u
m
er
ical
o
r
d
er
an
d
th
en
r
ep
laci
n
g
th
e
p
ix
el
b
ein
g
co
n
s
id
er
ed
w
it
h
t
h
e
m
id
d
le
p
ix
el
v
alu
e.
T
o
ev
en
l
y
d
is
tr
ib
u
te
th
e
i
n
te
n
s
i
t
y
o
f
o
u
r
d
atab
ase
o
f
i
m
a
g
es,
th
e
co
m
p
o
n
en
t
I
o
f
th
e
HSI
co
lo
r
s
p
ac
e
is
ex
tr
ac
ted
an
d
f
ilter
ed
th
r
o
u
g
h
a
3
x
3
m
ed
ia
n
f
i
lter
(
Fig
u
r
e
3
)
.
Fig
u
r
e
3
.
(
A
)
Or
ig
i
n
al
i
m
ag
e
(
B
)
R
esu
lt o
f
m
ed
ia
n
f
il
ter
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
u
to
ma
tic
E
xu
d
a
tes De
tectio
n
in
Dia
b
etic
R
etin
o
p
a
th
y
I
m
a
g
es
(
H.
F
a
o
u
z
i
)
47
2
.
4
.
Co
ntr
a
s
t
-
L
i
m
it
ed
Ada
ptiv
e
H
is
t
o
g
ra
m
E
q
ua
liza
t
io
n (
CL
AH
E
)
W
e
u
s
ed
th
e
C
o
n
tr
ast
-
L
i
m
it
ed
A
d
ap
tiv
e
Hi
s
to
g
r
a
m
E
q
u
aliza
tio
n
(
C
L
A
HE
)
[
1
]
b
ec
au
s
e
it
ca
n
in
cr
ea
s
e
t
h
e
co
n
tr
ast
b
et
w
ee
n
co
n
to
u
r
s
.
T
h
e
C
o
n
tr
ast
-
L
i
m
ite
d
A
d
ap
tiv
e
His
to
g
r
a
m
E
q
u
ali
za
tio
n
is
ap
p
lied
to
co
m
p
o
n
e
n
t I
o
f
o
u
r
f
ilter
ed
p
ictu
r
e
(
F
ig
u
r
e
4
)
.
Fig
u
r
e
4
.
C
o
n
tr
ast
-
L
i
m
ited
A
d
ap
tiv
e
His
to
g
r
a
m
E
q
u
aliza
tio
n
Fig
u
r
e
5.
Fin
al
I
m
a
g
e
Af
ter
P
r
e
-
P
r
o
ce
s
s
in
g
3.
SE
G
M
E
NT
A
T
I
O
N
B
y
d
ef
i
n
itio
n
,
in
i
m
a
g
e
p
r
o
ce
s
s
in
g
,
s
eg
m
e
n
tatio
n
is
co
n
v
e
n
tio
n
al
l
y
d
ef
i
n
ed
as
th
e
s
tep
o
f
p
ar
titi
o
n
in
g
a
n
i
m
ag
e
i
n
to
h
o
m
o
g
e
n
eo
u
s
r
eg
io
n
s
o
r
zo
n
es.
A
r
e
g
io
n
i
s
t
h
en
d
e
f
i
n
ed
as
a
s
et
o
f
p
i
x
els
t
h
at
s
h
ar
e
a
co
m
m
o
n
c
h
ar
ac
ter
is
tic
s
u
c
h
as i
n
ten
s
it
y
,
co
lo
r
,
tex
t
u
r
e.
3
.
1
.
C
o
nto
ur
Seg
m
ent
a
t
io
n
An
E
d
g
e
in
a
n
i
m
a
g
e
is
a
s
i
g
n
i
f
ican
t
lo
ca
l
ch
an
g
e
in
t
h
e
im
ag
e
i
n
te
n
s
it
y
,
u
s
u
a
ll
y
as
s
o
ci
ated
w
it
h
a
d
is
co
n
ti
n
u
i
t
y
in
eit
h
er
th
e
i
m
ag
e
i
n
te
n
s
it
y
o
r
t
h
e
f
ir
s
t
d
er
iv
ati
v
e
o
f
th
e
i
m
ag
e
in
ten
s
it
y
.
T
h
e
d
if
f
er
e
n
t
ed
g
e
d
etec
tio
n
m
eth
o
d
s
u
s
ed
ar
e
S
o
b
el,
P
r
ew
itt,
R
o
b
er
ts
,
C
an
n
y
,
L
o
G,
E
M
al
g
o
r
ith
m
,
OST
U
alg
o
r
it
h
m
[
8
]
an
d
Gen
etic
Alg
o
r
it
h
m
.
T
h
e
p
r
e
w
i
tt
o
p
er
ato
r
is
an
ap
p
r
o
x
i
m
ate
w
a
y
to
es
ti
m
ate
t
h
e
m
a
g
n
i
tu
d
e
an
d
o
r
ien
tatio
n
o
f
th
e
ed
g
e.
T
h
e
co
n
v
o
l
u
tio
n
m
a
s
k
o
f
p
r
e
w
itt o
p
er
ato
r
is
s
h
o
wn
in
f
i
g
u
r
e
6
.
Fig
u
r
e
6.
Ma
s
k
o
f
P
r
e
w
i
tt Op
er
ato
r
A
p
p
l
y
in
g
t
h
is
f
ilter
to
th
e
p
r
et
r
ea
ted
in
th
e
p
r
ep
r
o
ce
s
s
in
g
p
ictu
r
e
w
e
g
et
t
h
e
r
es
u
lt s
h
o
w
n
i
n
Fig
u
r
e
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
45
–
54
48
Fig
u
r
e
7
.
(
A
)
Gr
a
y
s
ca
le
i
m
a
g
e
(
B
)
R
esu
lt P
r
e
w
itt
f
ilter
L
ap
lacia
n
f
ilter
s
ar
e
d
er
iv
ativ
e
f
ilter
s
u
s
ed
to
f
in
d
ar
ea
s
o
f
r
ap
id
ch
an
g
e
(
ed
g
es)
in
i
m
a
g
es.
Si
n
ce
d
er
iv
ativ
e
f
ilter
s
ar
e
v
er
y
s
en
s
itiv
e
to
n
o
is
e,
it
is
co
m
m
o
n
to
s
m
o
o
t
h
th
e
i
m
ag
e
(
e.
g
.
,
u
s
i
n
g
a
Ga
u
s
s
ian
f
ilter
)
b
ef
o
r
e
ap
p
ly
i
n
g
t
h
e
L
ap
lacia
n
.
T
h
is
t
w
o
-
s
tep
p
r
o
ce
s
s
is
ca
lle
d
th
e
L
ap
lacia
n
o
f
Ga
u
s
s
ian
(
L
o
G)
o
p
er
ati
o
n
.
(
1
)
T
h
e
in
p
u
t
i
m
a
g
e
is
r
ep
r
esen
t
ed
as
a
s
et
o
f
d
is
cr
ete
p
ix
els;
w
e
n
ee
d
to
f
in
d
a
d
is
cr
ete
co
n
v
o
l
u
tio
n
k
er
n
el
t
h
at
ca
n
ap
p
r
o
x
i
m
ate
th
e
s
ec
o
n
d
d
er
iv
ativ
es
i
n
t
h
e
d
ef
in
itio
n
o
f
t
h
e
L
ap
laci
an
.
T
h
r
ee
k
er
n
els
co
m
m
o
n
l
y
u
s
ed
ar
e
s
h
o
w
n
i
n
t
h
e
F
i
g
u
r
e
8
:
Fig
u
r
e
8
.
Dis
cr
ete
C
o
n
v
o
lu
tio
n
Ker
n
el
A
p
p
r
o
x
i
m
ate
t
h
e
L
ap
lacia
n
Fig
u
r
e
9
.
L
ap
lacia
n
o
f
Gau
s
s
i
an
M
eth
o
d
3
.
2
.
Reg
i
o
n B
a
s
ed
Seg
m
ent
a
t
io
n
R
eg
io
n
g
r
o
w
i
n
g
is
a
s
i
m
p
le
r
eg
io
n
s
e
g
m
en
tatio
n
m
et
h
o
d
.
I
t
is
also
class
i
f
ied
as
a
p
ix
el
b
ased
i
m
ag
e
s
eg
m
e
n
tatio
n
m
et
h
o
d
s
in
ce
it
in
v
o
l
v
es
th
e
s
e
lectio
n
o
f
i
n
i
tial
s
ee
d
p
o
in
ts
.
T
h
is
ap
p
r
o
a
ch
to
s
eg
m
en
ta
tio
n
ex
a
m
in
e
s
n
ei
g
h
b
o
r
in
g
p
i
x
els
o
f
in
it
ial
“seed
p
o
in
ts
”
a
n
d
d
eter
m
i
n
es
w
h
eth
er
t
h
e
p
ix
el
n
eig
h
b
o
r
s
s
h
o
u
ld
b
e
ad
d
ed
to
th
e
r
eg
io
n
.
T
h
e
p
r
o
ce
s
s
is
iter
ated
o
n
,
i
n
th
e
s
a
m
e
m
an
n
er
as
g
en
er
al
d
ata
cl
u
s
ter
i
n
g
alg
o
r
it
h
m
s
.
I
n
th
is
s
ec
tio
n
,
w
e
p
r
esen
t
th
e
s
eg
m
e
n
tatio
n
b
ased
o
n
t
h
e
co
lo
r
ch
ar
ac
ter
is
tics
o
f
t
h
e
i
m
ag
e.
T
h
e
w
o
r
k
is
d
i
v
id
ed
in
to
t
w
o
s
ta
g
e
s
:
Firs
t,
w
e
w
ill
i
m
p
r
o
v
e
t
h
e
c
o
lo
r
o
f
th
e
i
m
ag
e
b
y
ex
tr
ac
tin
g
t
h
e
co
m
p
o
n
e
n
ts
a
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
u
to
ma
tic
E
xu
d
a
tes De
tectio
n
in
Dia
b
etic
R
etin
o
p
a
th
y
I
m
a
g
es
(
H.
F
a
o
u
z
i
)
49
*
b
*
o
f
s
p
ac
e
L
*
a
*
b
*
[
2
]
.
Seco
n
d
t
h
e
i
m
a
g
e
w
ill
b
e
g
r
o
u
p
in
g
i
n
to
a
s
et
o
f
f
iv
e
g
r
o
u
p
s
u
s
in
g
t
h
e
k
m
ea
n
s
alg
o
r
ith
m
.
T
h
e
w
h
o
le
p
r
o
ce
s
s
ca
n
b
e
s
u
m
m
ar
ized
in
t
h
e
f
o
ll
o
w
i
n
g
s
tep
s
:
3.
3
.
Co
nv
er
t
ing
t
he
I
m
a
g
e
o
f
t
he
RG
B
Co
lo
r
Sp
a
ce
t
o
t
h
e
CIE
L
*
a
*
b
T
h
e
C
I
E
L
*
a
*
b
*
h
elp
s
u
s
to
class
i
f
y
t
h
e
co
lo
r
d
if
f
er
en
c
e
b
ec
au
s
e
t
h
e
d
i
f
f
er
e
n
ce
s
b
et
w
ee
n
t
h
es
e
co
lo
r
s
ar
e
clo
s
er
to
h
u
m
an
p
er
ce
p
tio
n
.
Fig
u
r
e
1
0
.
C
o
n
v
er
ti
n
g
t
h
e
i
m
a
g
e
o
f
t
h
e
R
GB
co
lo
r
s
p
ac
e
to
t
h
e
C
I
E
L
*
a
*
b
3
.
4
.
Co
lo
r
Seg
m
ent
a
t
io
n Usi
ng
K
-
m
ea
n
s
Alg
o
rit
h
m
T
h
e
K
-
m
ea
n
s
al
g
o
r
ith
m
[
6
]
p
r
o
ce
s
s
es
ea
ch
o
b
j
ec
t
as
h
a
v
in
g
a
lo
ca
tio
n
i
n
s
p
ac
e.
T
h
e
alg
o
r
it
h
m
r
eq
u
ir
es
th
at
w
e
s
p
ec
i
f
y
i
n
g
t
h
e
n
u
m
b
er
o
f
r
eg
io
n
s
to
b
e
p
ar
titi
o
n
ed
an
d
a
d
is
tan
ce
m
etr
ic
f
o
r
q
u
an
ti
f
y
i
n
g
t
h
e
s
i
m
ilar
it
y
b
et
w
ee
n
t
w
o
p
i
x
els.
I
n
th
is
p
h
a
s
e
w
e
u
s
ed
th
e
K
-
m
ea
n
s
alg
o
r
it
h
m
to
g
r
o
u
p
p
ix
els
in
to
f
i
v
e
g
r
o
u
p
s
b
y
u
s
i
n
g
th
e
E
u
cl
id
ea
n
d
is
ta
n
ce
as
th
e
m
etr
ic.
B
ased
o
n
th
e
r
esu
lt
o
f
th
i
s
alg
o
r
it
h
m
,
w
e
w
il
l
lab
el
ea
ch
p
ix
el
o
f
th
e
i
m
a
g
e
w
it
h
h
er
clu
s
ter
i
n
d
ex
.
3
.
5
.
Cre
a
t
ing
I
m
a
g
es F
ro
m
Seg
m
e
nts
I
n
th
is
s
tep
w
e
w
ill
cr
ea
te
i
m
ag
e
s
f
r
o
m
s
eg
m
e
n
ts
w
e
o
b
tain
ed
th
e
p
r
ev
io
u
s
s
tag
e.
T
h
e
r
esu
lt
th
e
f
o
llo
w
in
g
Fi
g
u
r
e
(
F
ig
u
r
e1
1
)
.
Fig
u
r
e
1
1
.
R
esu
lt
s
o
f
t
h
e
K
-
m
ea
n
s
al
g
o
r
ith
m
3
.
6
.
Select
io
n
a
nd
B
ina
riza
t
io
n o
f
Ca
nd
ida
t
e
I
m
a
g
e
I
n
th
i
s
s
ec
tio
n
,
w
e
w
ill
s
elec
t
an
d
b
in
ar
ized
th
e
i
m
a
g
e
ca
n
d
id
ate
k
n
o
w
in
g
t
h
at
th
e
o
p
tic
al
d
is
c
an
d
ex
u
d
ate
s
ar
e
h
o
m
o
g
e
n
eo
u
s
i
n
th
e
m
co
lo
r
p
r
o
p
er
ty
an
d
ar
e
ch
ar
ac
ter
ized
b
y
a
s
tr
o
n
g
an
d
b
r
ig
h
tn
e
s
s
co
n
tr
as
t
co
m
p
ar
ab
le
to
o
th
er
i
m
a
g
e.
T
h
e
r
esu
l
t is
i
n
d
icate
d
in
t
h
e
f
o
l
lo
w
i
n
g
F
i
g
u
r
e
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
45
–
54
50
Fig
u
r
e
1
2
.
Fin
al
Se
g
m
e
n
tat
io
n
R
es
u
lt o
f
Op
tical
Di
s
c
a
n
d
E
x
u
d
at
4.
RE
CO
G
NI
T
I
O
N
O
F
E
X
UD
AT
E
x
u
d
ates
is
f
ir
s
t
d
escr
ib
ed
b
y
a
s
et
o
f
i
n
v
ar
ian
t
m
o
m
e
n
ts
(
H
u
an
d
Af
f
i
n
e)
an
d
d
escr
ip
to
r
GI
ST
th
en
u
s
ed
i
n
t
h
e
s
ta
g
es
o
f
lear
n
in
g
an
d
r
ec
o
g
n
itio
n
.
T
h
e
r
ep
r
esen
tatio
n
o
f
i
m
a
g
es
i
s
a
n
i
m
p
o
r
tan
t
s
tep
in
th
e
r
ec
o
g
n
itio
n
p
h
a
s
e.
I
t
m
u
s
t
b
e
in
v
ar
ian
t
to
g
eo
m
etr
ic
tr
an
s
f
o
r
m
at
io
n
s
(
r
o
tatio
n
,
tr
an
s
latio
n
a
n
d
s
ca
le
f
ac
to
r
)
an
d
r
o
b
u
s
t
to
v
ar
io
u
s
d
is
t
u
r
b
an
ce
s
(
n
o
is
e,
d
i
m
m
i
n
g
,
etc)
.
T
h
e
r
ep
r
esen
tat
io
n
w
e
h
a
v
e
ad
o
p
ted
in
th
i
s
p
r
o
j
ec
t
is
b
ased
o
n
af
f
i
n
e
m
o
m
en
t,
H
u
m
o
m
e
n
t a
n
d
GI
ST
ap
p
lied
to
b
in
ar
y
i
m
ag
e
s
.
4
.
1
.
M
o
m
e
nt
I
nv
a
ria
nts
[
7
]
T
h
e
m
o
m
e
n
ts
ca
n
b
e
u
s
ed
to
d
escr
ib
e
a
f
o
r
m
g
lo
b
all
y
.
C
ar
t
esian
2
d
m
o
m
e
n
t
i
n
v
ar
ian
t
o
f
o
r
d
er
p
,
q
f
o
r
f
u
n
ctio
n
f
(
x
,
y
)
is
r
ep
r
ese
n
ted
b
y
:
(
2
)
f
o
r
a
d
is
cr
ete
i
m
ag
e
:
(
3
)
I
t is th
u
s
p
o
s
s
ib
le
to
ca
lcu
late
th
e
m
o
m
e
n
t o
r
d
er
0
r
e
p
r
esen
ts
th
e
m
a
s
s
o
r
ar
ea
o
f
a
f
u
n
ctio
n
:
(
4
)
T
h
e
t
w
o
m
o
m
e
n
t
s
o
f
o
r
d
er
1
ar
e
r
elate
d
to
th
e
co
r
r
esp
o
n
d
in
g
co
o
r
d
in
ates
o
f
t
h
e
ce
n
ter
o
f
t
h
e
f
u
n
ctio
n
:
(
5
)
(
6
)
Fro
m
th
e
t
w
o
m
o
m
e
n
ts
o
f
o
r
d
er
1
,
w
e
ca
n
f
i
n
d
th
e
ce
n
ter
o
f
m
as
s
o
f
t
h
e
f
o
r
m
(
7
)
T
h
e
ce
n
tr
al
m
o
m
e
n
t
s
ca
n
b
e
ex
p
r
ess
ed
b
y
:
(
8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
u
to
ma
tic
E
xu
d
a
tes De
tectio
n
in
Dia
b
etic
R
etin
o
p
a
th
y
I
m
a
g
es
(
H.
F
a
o
u
z
i
)
51
Fo
r
a
d
ig
ital i
m
a
g
e,
th
e
eq
u
a
ti
o
n
b
ec
o
m
e
s
:
(
9
)
4
.
2
.
Af
f
ine
M
o
m
e
nt
T
h
e
af
f
i
n
e
tr
an
s
f
o
r
m
is
g
e
n
er
a
l lin
ea
r
tr
an
s
f
o
r
m
atio
n
o
f
s
p
ac
e
co
o
r
d
in
ates o
f
th
e
i
m
ag
e:
(
1
0
)
(
1
1
)
T
h
e
af
f
in
e
m
o
m
e
n
t
i
n
v
ar
ian
t
s
ar
e
f
ea
t
u
r
es
f
o
r
p
atter
n
r
ec
o
g
n
itio
n
co
m
p
u
ted
f
r
o
m
m
o
m
e
n
t
s
o
f
o
b
j
ec
ts
o
n
i
m
a
g
es
t
h
at
d
o
n
o
t
ch
an
g
e
t
h
eir
v
al
u
e
i
n
a
f
f
in
e
tr
an
s
f
o
r
m
atio
n
.
I
n
th
is
w
o
r
k
w
e
w
ill
u
s
e
th
e
s
e
v
en
af
f
in
e
m
o
m
e
n
t
s
:
4
.
3
.
H
u
M
o
m
e
nt
B
ased
o
n
n
o
r
m
alize
d
ce
n
tr
al
m
o
m
e
n
t
s
,
Hu
i
n
tr
o
d
u
ce
d
s
ev
e
n
m
o
m
en
t in
v
ar
ia
n
ts
:
(
1
2
)
(
1
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
45
–
54
52
T
h
e
s
ev
en
m
o
m
e
n
t
in
v
ar
ia
n
t
s
ar
e
u
s
ef
u
l
p
r
o
p
er
ties
o
f
b
ein
g
u
n
ch
a
n
g
ed
u
n
d
er
i
m
a
g
e
s
ca
lin
g
,
tr
an
s
latio
n
an
d
r
o
tatio
n
.
4
.
4
.
G
I
ST
De
s
cr
ipto
rs
I
n
co
m
p
u
ter
v
is
io
n
,
GI
ST
d
escr
ip
to
r
s
ar
e
a
r
ep
r
esen
tatio
n
o
f
an
i
m
ag
e
in
lo
w
d
i
m
e
n
s
io
n
t
h
at
co
n
tain
s
en
o
u
g
h
in
f
o
r
m
atio
n
t
o
id
en
tify
t
h
e
s
ce
n
e.
A
c
tu
all
y
,
an
y
g
lo
b
al
d
escr
ip
to
r
m
u
s
t
ap
p
r
o
ac
h
th
e
GI
ST
t
o
b
e
u
s
e
f
u
l.
GI
ST
d
escr
ip
to
r
w
as
p
r
o
p
o
s
ed
b
y
Oli
v
a
a
n
d
m
o
r
e
p
r
ec
is
el
y
b
y
T
o
r
r
alb
a
[
9
]
.
T
h
e
y
tr
ied
to
ca
p
t
u
r
e
th
e
GI
ST
d
escr
ip
to
r
o
f
th
e
i
m
a
g
e
b
y
an
al
y
zi
n
g
th
e
s
p
atial
f
r
e
q
u
en
cie
s
an
d
o
r
ien
tatio
n
s
.
4
.
5
.
Neura
l net
w
o
rk
co
ns
t
ru
ct
io
n
Fig
u
r
e
1
3
illu
s
tr
ate
a
n
ex
a
m
p
le
o
f
n
eu
r
al
n
e
t
w
o
r
k
[
1
0
]
u
s
ed
f
o
r
th
e
a
f
f
in
e
m
o
m
e
n
t,
n
a
m
el
y
a
m
u
ltil
a
y
er
n
et
w
o
r
k
t
h
at
co
n
tai
n
s
a
h
id
d
en
la
y
er
.
I
n
th
e
f
o
llo
w
i
n
g
T
ab
le
w
e
s
u
m
m
ar
is
e
th
e
d
ec
r
ip
to
r
s
w
it
h
t
h
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
in
i
n
p
u
t la
y
er
,
h
id
d
en
la
y
er
an
d
o
u
tp
u
t la
y
er
.
Fig
u
r
e
1
3
.
Stru
ctu
r
e
o
f
t
h
e
n
e
u
r
al
n
et
w
o
r
k
u
s
ed
f
o
r
th
e
a
f
f
in
e
m
o
m
e
n
t
.
T
ab
le
1
.
C
o
n
s
tr
u
ctio
n
o
f
n
e
u
r
al
n
et
w
o
r
k
f
o
r
d
if
f
er
e
n
t d
escr
i
p
to
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
u
to
ma
tic
E
xu
d
a
tes De
tectio
n
in
Dia
b
etic
R
etin
o
p
a
th
y
I
m
a
g
es
(
H.
F
a
o
u
z
i
)
53
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
NS
I
n
o
r
d
er
to
in
ter
p
r
et
th
ese
r
es
u
lts
,
w
e
p
lo
tted
s
e
v
er
al
g
r
ap
h
s
(
Fig
u
r
e
1
4
,
1
5
,
1
6
)
.
T
h
ese
w
ill
allo
w
u
s
in
p
ar
ticu
lar
to
d
eter
m
i
n
e
t
h
e
b
est
s
eg
m
e
n
tatio
n
al
g
o
r
it
h
m
an
d
t
h
e
d
escr
ip
to
r
w
e
w
il
l
h
av
e
to
u
s
e.
Fro
m
g
r
ap
h
s
w
e
co
n
s
tate
t
h
at
k
m
ea
n
s
w
it
h
Hu
m
o
m
e
n
t
an
d
GI
ST
d
escr
ip
to
r
s
is
b
eth
er
th
an
o
th
er
d
ec
r
ip
to
r
s
in
r
ec
o
g
n
itio
n
r
ate.
Fig
u
r
e
1
4
.
R
esu
lt o
f
s
e
g
m
e
n
ta
tio
n
b
y
co
n
to
u
r
Fig
u
r
e
1
5
.
R
esu
lt o
f
s
e
g
m
e
n
ta
tio
n
b
y
k
m
e
n
s
Fig
u
r
e
1
6
.
C
o
m
p
ar
is
o
n
b
et
w
e
en
m
et
h
o
d
s
6.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
ex
p
lo
r
ed
th
e
is
s
u
e
o
f
s
eg
m
e
n
tatio
n
an
d
r
ec
o
g
n
iti
o
n
o
f
ex
u
d
ates
f
o
r
d
iab
etic
r
etin
o
p
ath
y
.
W
e
h
av
e
p
r
ese
n
ted
t
w
o
d
if
f
er
en
t
p
ar
ts
t
w
o
s
eg
m
e
n
tatio
n
al
g
o
r
ith
m
s
n
a
m
el
y
s
e
g
m
en
tatio
n
b
ased
o
n
co
n
to
u
r
an
d
k
-
m
ea
n
s
al
g
o
r
it
h
m
.
Fo
r
th
e
ex
tr
ac
tio
n
o
f
ch
ar
ac
te
r
is
tics
o
f
i
m
a
g
es
w
u
s
ed
H
u
Mo
m
e
n
t,
A
f
f
i
n
e
Mo
m
en
t,
GI
ST
an
d
d
if
f
er
en
t
co
m
b
i
n
ati
o
n
o
f
th
e
s
e
d
escr
ip
to
r
s
.
T
o
ev
alu
ate
t
h
e
p
er
f
o
r
m
an
ce
o
f
ea
ch
alg
o
r
it
h
m
o
n
th
e
d
etec
tio
n
o
f
ex
u
d
ates,
ea
c
h
d
escr
ip
to
r
w
as
ap
p
lied
to
th
e
s
a
m
e
s
et
o
f
i
m
ag
e
s
o
f
a
n
e
y
e
b
o
tto
m
an
d
t
h
e
r
es
u
lt
s
w
er
e
co
m
p
ar
ed
.
W
e
c
o
n
f
ir
m
e
d
in
th
is
s
t
u
d
y
t
h
at
k
m
ea
n
s
s
e
g
m
e
n
tat
io
n
alg
o
r
it
h
m
ap
p
ea
r
s
to
b
e
m
o
r
e
s
u
ited
to
th
e
d
etec
tio
n
o
f
e
x
u
d
ates u
s
i
n
g
Hu
m
o
m
en
t
s
an
d
GI
ST
as a
d
escr
ip
to
r
as a
d
escr
ip
to
r
.
RE
F
E
R
E
NC
E
S
[1
]
G
Eas
o
n
,
B
No
b
le,
IN
S
n
e
d
d
o
n
.
On
Ce
rta
in
In
teg
ra
ls
o
f
L
ip
sc
h
it
z
-
Ha
n
k
e
l
Ty
p
e
In
v
o
lv
in
g
P
r
o
d
u
c
ts
o
f
Be
ss
e
l
F
u
n
c
ti
o
n
s
.
Ph
i
l.
T
r
a
n
s.
Ro
y
.
S
o
c
.
L
o
n
d
o
n
.
1
9
5
5
;
A
2
4
7
:
5
29
-
5
5
1
.
[2
]
JP
C
o
c
q
u
e
re
z
,
S
P
h
il
ip
p
.
A
n
a
ly
se
D.Im
a
g
e
:
F
il
trag
e
Et
S
e
g
m
e
n
tati
o
n
.
e
d
.
M
a
ss
o
n
,
1
9
9
5
.
[3
]
Ak
a
ra
S
o
p
h
a
ra
k
,
Kh
in
e
T
h
e
t
N
w
e
,
Yin
Ay
e
M
o
e
,
M
a
tt
h
e
w
N.
Da
il
e
y
e
t
Bu
n
y
a
rit
U
y
y
a
n
o
n
v
a
ra
A
u
to
m
a
ti
c
Ex
u
d
a
te
De
tec
ti
o
n
with
a
Na
iv
e
Ba
y
e
s
Cla
ss
i
f
ier.
S
iri
n
d
h
o
r
n
In
tern
a
ti
o
n
a
l
In
sti
tu
te
o
f
T
e
c
h
n
o
lo
g
y
,
T
h
a
m
m
a
sa
t
Un
iv
e
rsit
y
&
Co
m
p
u
ter S
c
ien
c
e
a
n
d
In
f
o
rm
a
ti
o
n
M
a
n
a
g
e
m
e
n
t,
As
ian
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
y
.
[4
]
A
li
re
z
a
Os
a
re
h
,
M
a
ji
d
M
irm
e
h
d
i,
Ba
rry
T
h
o
m
a
s,
Rich
a
rd
M
a
rk
h
a
m
.
Cla
ss
if
ica
ti
o
n
a
n
d
lo
c
a
li
za
t
i
o
n
o
f
d
i
a
b
e
ti
c
-
re
la
ted
e
y
e
d
ise
a
se
.
In
P
ro
c
.
o
f
7
t
h
Eu
r
o
p
e
a
n
Co
n
f
e
re
n
c
e
o
n
Co
m
p
u
ter Visio
n
(ECC
V
).
2
0
0
2
:
5
0
2
–
5
1
6
.
[5
]
Nic
h
o
a
ls
P
,
W
a
rd
,
S
tep
h
e
n
T
o
m
li
n
so
n
,
e
t,
C
h
ri
sto
p
h
e
r
J
,
T
a
y
lo
r
.
Im
a
g
e
A
n
a
l
y
sis
o
f
F
u
n
d
u
s
P
h
o
to
g
ra
p
h
s:
T
h
e
d
e
tec
ti
o
n
a
n
d
M
e
a
su
re
m
e
n
t
o
f
Ex
u
d
a
tes
A
ss
o
c
iate
d
w
it
h
Dia
b
e
ti
c
Re
ti
n
o
p
a
t
h
y
.
Op
h
th
a
lm
o
lo
g
y
1
9
8
9
;
9
6
:
80
-
86
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
45
–
54
54
[6
]
A
li
re
z
a
Os
a
re
h
M
a
ji
d
M
irm
e
h
d
i,
Ba
rry
T
h
o
m
a
s
e
t
Rich
a
rd
M
a
rk
h
a
m
.
Au
to
ma
ti
c
Rec
o
g
n
it
io
n
o
f
Exu
d
a
ti
v
e
M
a
c
u
lo
p
a
t
h
y
u
si
n
g
Fu
zz
y
C
-
M
e
a
n
s
Cl
u
ste
rin
g
a
n
d
Ne
u
ra
l
Ne
t
wo
rk
.
P
r
o
c
.
M
e
d
ica
l
Im
a
g
e
Un
d
e
rs
tan
d
i
n
g
a
n
d
A
n
a
l
y
si
s Co
n
f
e
re
n
c
e
.
Ju
il
let
2
0
0
1
:
49
-
52
.
[7
]
M
-
K Hu
.
V
is
u
a
l
P
a
tt
e
rn
Re
c
o
g
n
i
t
io
n
b
y
M
o
m
e
n
t
In
v
a
rian
t
.
IRE
T
ra
n
s In
f
o
rm
a
ti
o
n
T
h
e
o
ry
.
1
9
6
2
;
8
:
1
7
9
–
1
8
7
.
[8
]
N
Otsu
.
A
T
h
re
sh
o
ld
i
n
g
S
e
lec
ti
o
n
M
e
th
o
d
f
ro
m
G
r
a
y
L
e
v
e
l
Histo
g
ra
m
s
.
IEE
E
T
ra
n
s
.
O
n
s
y
st,
m
a
n
a
n
d
c
y
b
e
rn
e
ti
c
s,
1
9
7
9
,
9
(1
),
6
2
-
66.
[9
]
A
T
o
rra
lb
a
,
R
F
e
rg
u
s,
Y
W
e
iss.
S
m
a
ll
c
o
d
e
s an
d
larg
e
d
a
tab
a
se
s f
o
r
re
c
o
g
n
it
io
n
.
I
n
C
VP
R,
2
0
0
8
.
[1
0
]
G
Ga
rd
n
e
r,
D
Ke
a
ti
n
g
,
T
Wi
ll
iam
so
n
,
e
t
a
l.
A
u
to
m
a
ti
c
D
e
tec
ti
o
n
o
f
Dia
b
e
ti
c
Re
ti
n
o
p
a
t
h
y
U
sin
g
a
n
A
rti
f
icia
l
Ne
u
ra
l
Ne
t
w
o
rk
:
A
S
c
re
e
n
in
g
T
o
o
l.
Brit
ish
J
o
u
rn
a
l
o
f
O
p
h
t
h
a
lm
o
lo
g
y
.
1
9
9
6
;
8
0
:
9
4
0
-
9
4
4
.
[1
1
]
A
F
ra
m
e
,
P
Un
d
ril
l,
M
Cre
e
,
e
t
a
l.
A
Co
m
p
a
riso
n
o
f
Co
m
p
u
ter
Clas
sif
ic
a
ti
o
n
M
e
th
o
d
s
A
p
p
li
e
d
t
o
De
tec
ti
o
n
o
f
M
icro
a
n
e
u
ry
s
m
s
in
Op
h
th
a
lm
i
c
F
lu
o
re
sc
e
in
A
n
g
io
g
ra
m
s.
Co
m
p
u
ters
in
Bio
lo
g
y
a
n
d
M
e
d
icin
e
.
1
9
9
8
;
2
8
:
2
2
5
-
2
3
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.