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.
9
,
No
.
1
,
Ma
r
ch
2
0
2
0
,
p
p
.
58
~
64
I
SS
N:
2252
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
9
.i
1
.
p
p
58
-
64
58
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Iris
seg
m
e
ntatio
n
using
a new
unsuperv
ised neu
ra
l appro
a
ch
H
icha
m
O
h
m
a
id
1
,
S.
E
dd
a
ro
uich
2
,
A.
B
o
uro
uh
o
u
3
,
M
.
T
i
m
o
uy
a
s
4
1
Na
ti
o
n
a
l
S
c
h
o
o
l
o
f
Co
m
p
u
ter S
c
ien
c
e
a
n
d
S
y
ste
m
s
A
n
a
l
y
sis
(ENS
IA
S
),
M
o
h
a
m
m
e
d
V
Un
iv
e
rsity
,
Ra
b
a
t/
1
0
1
1
2
,
M
o
ro
c
c
o
.
2
Re
g
io
n
a
l
Ed
u
c
a
ti
o
n
a
l
Ce
n
ter,
Ra
b
a
t/
1
0
1
1
2
,
M
o
r
o
c
c
o
3,
4
Hig
h
e
r
No
rm
a
l
S
c
h
o
o
l
o
f
T
e
c
h
n
ica
l
Ed
u
c
a
ti
o
n
(ENS
ET
),
M
o
h
a
m
m
e
d
V
Un
iv
e
rsity
,
ENS
ET
,
Ra
b
a
t/
1
0
1
1
2
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 7
,
2
0
1
9
R
ev
i
s
ed
Oct
18
,
2
0
1
9
A
cc
ep
ted
Dec
3
,
2
0
1
9
A
b
io
m
e
tri
c
s
y
ste
m
o
f
id
e
n
ti
f
ica
ti
o
n
a
n
d
a
u
t
h
e
n
ti
c
a
ti
o
n
p
ro
v
id
e
s
a
u
to
m
a
ti
c
re
c
o
g
n
it
io
n
o
f
a
n
in
d
iv
id
u
a
l
b
a
se
d
o
n
c
e
rtain
u
n
iq
u
e
f
e
a
tu
re
s
o
r
c
h
a
ra
c
teristic
p
o
ss
e
ss
e
d
b
y
a
n
in
d
iv
id
u
a
l.
Iris
re
c
o
g
n
it
io
n
is
a
b
io
m
e
tri
c
id
e
n
ti
f
ica
ti
o
n
m
e
th
o
d
t
h
a
t
u
se
s
p
a
tt
e
rn
re
c
o
g
n
it
io
n
o
n
t
h
e
im
a
g
e
s
o
f
th
e
iri
s.
O
win
g
to
th
e
u
n
i
q
u
e
e
p
ig
e
n
e
ti
c
p
a
tt
e
rn
s
o
f
th
e
iri
s,
Iris
re
c
o
g
n
it
io
n
is
c
o
n
sid
e
re
d
a
s
o
n
e
o
f
th
e
m
o
st
a
c
c
u
ra
te
m
e
th
o
d
s
in
th
e
f
ield
o
f
b
io
m
e
tri
c
id
e
n
ti
f
ica
ti
o
n
.
On
e
o
f
th
e
c
ru
c
ial
ste
p
s
in
th
e
iri
s
re
c
o
g
n
it
io
n
sy
ste
m
is
th
e
iri
s
se
g
m
e
n
tatio
n
b
e
c
a
u
se
it
sig
n
if
ica
n
tl
y
a
ff
e
c
ts
th
e
a
c
c
u
ra
c
y
o
f
th
e
f
e
a
tu
re
e
x
tra
c
ti
o
n
th
e
iri
s.
T
h
e
se
g
m
e
n
tatio
n
a
lg
o
rit
h
m
p
ro
p
o
se
d
in
th
is
a
rti
c
le
sta
rts
w
it
h
d
e
ter
m
in
in
g
th
e
r
e
g
io
n
s
o
f
th
e
e
y
e
u
sin
g
u
n
su
p
e
rv
ise
d
n
e
u
ra
l
a
p
p
ro
a
c
h
,
a
f
ter
th
e
o
u
tl
i
n
e
o
f
th
e
e
y
e
is
f
o
u
n
d
u
sin
g
t
h
e
Ca
n
n
y
e
d
g
e
,
T
h
e
Ho
u
g
h
T
ra
n
s
f
o
r
m
is
e
m
p
lo
y
e
d
to
d
e
term
in
e
th
e
c
e
n
ter an
d
ra
d
i
u
s o
f
th
e
p
u
p
il
a
n
d
t
h
e
iri
s
.
K
ey
w
o
r
d
s
:
B
io
m
etr
ic
Ho
u
g
h
tr
an
s
f
o
r
m
I
r
is
s
eg
m
e
n
tatio
n
Ma
h
ala
n
o
b
is
d
is
tan
ce
Un
s
u
p
er
v
i
s
ed
n
eu
r
al
ap
p
r
o
ac
h
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
:
Hich
a
m
O
h
m
aid
,
Natio
n
al
Sc
h
o
o
l o
f
C
o
m
p
u
ter
Scien
ce
a
n
d
S
y
s
te
m
s
An
al
y
s
i
s
(
E
NSI
A
S),
Mo
h
a
m
m
ed
V
Un
iv
er
s
it
y
,
R
ab
at/1
0
1
1
2
,
M
o
r
o
cc
o
.
E
m
ail:
h
ic
h
a
m
.
o
h
m
aid
@
g
m
a
i
l
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
No
w
ad
a
y
s
,
b
io
m
etr
ic
r
ec
o
g
n
it
io
n
h
as
b
ec
o
m
e
a
r
eliab
le
w
a
y
f
o
r
th
e
id
en
ti
f
icatio
n
a
n
d
r
ec
o
g
n
i
tio
n
o
f
in
d
iv
id
u
als,
b
ased
o
n
p
h
y
s
io
lo
g
ical
o
r
b
eh
av
io
r
al
ch
ar
ac
ter
is
tics
o
f
t
h
ese
p
eo
p
le.
C
o
n
v
en
t
io
n
al
m
eth
o
d
s
id
en
tit
y
c
h
ec
k
lik
e
p
as
s
w
o
r
d
an
d
id
en
ti
f
icatio
n
ca
r
d
ar
e
n
o
t
al
w
a
y
s
cr
ed
ib
le
b
ec
au
s
e
th
e
s
e
m
et
h
o
d
s
ca
n
b
e
f
o
r
g
o
tten
o
r
s
to
len
.
T
h
e
id
en
tif
icat
io
n
b
ased
o
n
b
io
m
etr
ic
s
h
as
b
ee
n
u
s
ed
f
o
r
s
ec
u
r
it
y
s
y
s
te
m
s
,
s
u
c
h
as
au
th
e
n
tica
tio
n
a
n
d
i
n
f
o
r
m
atio
n
p
r
o
tectio
n
.
T
h
e
b
io
m
e
tr
ic
te
ch
n
o
lo
g
ies
[
1
-
4
]
u
ti
lize
b
eh
a
v
io
r
al
ch
ar
ac
ter
is
tic
s
(
s
u
c
h
as
h
an
d
w
r
iti
n
g
)
o
r
p
h
y
s
io
lo
g
ical
ch
ar
ac
ter
i
s
tics
(
s
u
ch
a
s
f
in
g
er
p
r
in
t,
f
ac
e
an
d
ir
is
)
to
ac
cu
r
atel
y
au
t
h
e
n
tica
te
p
er
s
o
n
al
id
e
n
tit
y
.
Am
o
n
g
t
h
ese
b
io
m
etr
ic
t
ec
h
n
o
lo
g
ies,
ir
is
r
ec
o
g
n
i
tio
n
o
f
f
er
s
th
e
h
i
g
h
est
ac
cu
r
ac
y
i
n
id
e
n
ti
f
y
in
g
i
n
d
iv
i
d
u
als
as
co
m
p
ar
ed
to
an
y
o
th
er
b
io
m
etr
ic
ap
p
r
o
ac
h
es.
T
h
e
ir
is
is
th
e
co
lo
r
ed
r
eg
io
n
o
f
t
h
e
e
y
e
b
o
u
n
d
ed
b
y
t
h
e
p
u
p
il
an
d
w
h
ite
s
cler
a,
it
is
s
o
u
n
iq
u
e
th
at
n
o
t
w
o
ir
is
es
ar
e
alik
e,
ev
e
n
a
m
o
n
g
id
e
n
tical
t
w
i
n
s
o
r
e
v
en
b
et
w
ee
n
t
h
e
le
f
t
a
n
d
r
i
g
h
t
e
y
e
o
f
t
h
e
s
a
m
e
p
er
s
o
n
[
5
]
.
T
h
ese
v
i
s
ib
le
ch
ar
ac
ter
is
tic
s
ar
e
t
h
o
u
g
h
t
to
b
e
h
i
g
h
l
y
d
i
s
cr
i
m
i
n
ato
r
y
an
d
u
n
iq
u
e
to
ea
c
h
e
y
e,
as
w
e
l
l
as
s
tab
le
o
v
er
a
n
in
d
iv
id
u
al
’
s
lif
e
ti
m
e.
T
h
is
m
a
k
es
t
h
e
ir
is
a
v
er
y
u
s
ef
u
l
b
io
m
etr
ic
id
e
n
ti
f
ier
w
h
en
it
i
s
p
o
s
s
ib
le
to
ca
p
t
u
r
e
ir
is
i
m
a
g
es o
f
r
ea
s
o
n
ab
le
q
u
a
lit
y
.
d
escr
ib
e
th
e
s
tep
o
f
r
esear
ch
an
d
u
s
ed
in
th
e
ch
ap
ter
"
R
es
u
lts
an
d
Dis
c
u
s
s
io
n
"
to
s
u
p
p
o
r
t
th
e
an
al
y
s
is
o
f
t
h
e
r
esu
lts
[
2
]
.
I
f
th
e
m
a
n
u
s
c
r
ip
t
w
as
w
r
it
ten
r
ea
ll
y
h
a
v
e
h
ig
h
o
r
ig
i
n
alit
y
,
w
h
ic
h
p
r
o
p
o
s
ed
a
n
e
w
m
et
h
o
d
o
r
alg
o
r
ith
m
,
t
h
e
ad
d
itio
n
a
l c
h
ap
ter
af
ter
th
e
"
I
n
tr
o
d
u
ctio
n
"
ch
ap
ter
an
d
b
e
f
o
r
e
th
e
"
R
e
s
ea
r
ch
Me
t
h
o
d
"
ch
ap
ter
ca
n
b
e
ad
d
ed
t
o
ex
p
lain
b
r
ief
l
y
t
h
e
th
eo
r
y
a
n
d
/o
r
th
e
p
r
o
p
o
s
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
I
r
is
s
eg
men
ta
tio
n
u
s
in
g
a
n
ew u
n
s
u
p
ervis
ed
n
eu
r
a
l a
p
p
r
o
a
c
h
(
Hich
a
m
Oh
ma
id
)
59
m
et
h
o
d
/
alg
o
r
ith
m
[
4
]
.
T
h
e
s
y
s
te
m
i
s
i
n
s
p
ir
ed
b
y
Dau
g
m
an
w
o
r
k
s
[
5
]
.
I
t
is
co
m
p
o
s
ed
o
f
a
n
u
m
b
er
o
f
s
u
b
s
y
s
te
m
s
s
h
o
w
n
i
n
Fi
g
u
r
e
1
th
a
t c
o
n
tain
s
s
ta
g
es o
f
ir
i
s
d
etec
ti
o
n
s
y
s
t
e
m
:
Fig
u
r
e
1
.
I
r
is
d
etec
tio
n
s
y
s
te
m
T
h
e
s
eg
m
en
tatio
n
s
tep
allo
w
s
y
o
u
lo
ca
lize
t
h
e
ir
i
s
r
eg
io
n
i
n
a
n
e
y
e
i
m
a
g
e,
th
is
co
n
s
i
s
ts
o
f
f
in
d
i
n
g
t
w
o
co
n
to
u
r
s
o
f
ir
is
(
p
u
p
il
an
d
ir
is
)
,
T
h
at
is
to
s
a
y
,
lo
ca
te
th
e
I
n
ter
n
al
co
n
to
u
r
(
p
u
p
il/i
r
is
)
a
n
d
o
u
ter
(
ir
is
/s
cler
a)
.
On
ce
th
e
ir
is
r
eg
io
n
is
s
eg
m
e
n
ted
,
t
h
e
n
e
x
t
s
t
ag
e
is
to
n
o
r
m
a
lize
t
h
is
p
ar
t.
T
h
e
n
o
r
m
aliza
tio
n
A
llo
w
s
to
tr
an
s
f
o
r
m
th
e
ir
i
s
t
ex
tu
r
e
f
r
o
m
C
ar
tesi
a
n
to
p
o
lar
co
o
r
d
in
ates,
I
n
o
r
d
er
to
li
m
it
th
e
p
r
o
b
le
m
s
o
f
v
ar
iatio
n
s
in
th
e
e
y
e
(
lik
e
th
e
o
p
tical
s
ize
o
f
th
e
ir
is
,
th
e
p
o
s
itio
n
o
f
a
p
u
p
il
in
t
h
e
ir
is
)
,
th
is
s
tep
is
b
ased
o
n
Dau
g
m
a
n
'
s
r
u
b
b
er
-
s
h
ee
t
m
o
d
e
l
[
5
]
.
Featu
r
e
ex
tr
ac
tio
n
i
s
a
te
ch
n
iq
u
e
u
s
ed
to
ex
tr
ac
t
t
h
e
i
n
f
o
r
m
at
io
n
f
r
o
m
t
h
e
ir
is
i
m
ag
e,
I
n
o
r
d
er
to
lo
w
er
t
h
e
s
ize
o
f
ir
is
m
o
d
els a
n
d
i
m
p
r
o
v
e
clas
s
if
i
er
ac
cu
r
ac
y
.
T
h
ese
f
ea
t
u
r
es c
a
n
n
o
t b
e
u
s
ed
f
o
r
r
ec
o
n
s
tr
u
ctio
n
o
f
i
m
a
g
es,
b
u
t
th
e
s
e
v
al
u
es
ar
e
u
s
ed
in
class
i
f
icatio
n
.
Fin
al
l
y
,
m
a
t
ch
in
g
is
p
er
f
o
r
m
ed
b
y
co
m
p
ar
in
g
th
e
f
ea
tu
r
e
o
f
te
m
p
late
ir
is
w
i
th
t
h
e
f
ea
tu
r
e
v
ec
to
r
s
o
f
te
m
p
lates
in
th
e
d
atab
ase
an
d
d
ec
is
io
n
is
f
o
r
m
u
lated
[
6
]
.
Fig
u
r
e
2
s
h
o
w
s
t
y
p
ical
s
ta
g
es o
f
i
r
is
r
ec
o
g
n
itio
n
.
Fig
u
r
e
2
.
T
y
p
ical
s
tag
e
s
o
f
ir
is
r
ec
o
g
n
itio
n
T
h
e
m
ai
n
m
o
ti
v
atio
n
in
t
h
is
r
esear
ch
is
to
p
r
o
p
o
s
e
a
n
ew
e
f
f
ec
ti
v
e
an
d
r
o
b
u
s
t
alg
o
r
it
h
m
t
o
s
eg
m
e
n
t
th
e
clea
r
o
r
n
o
is
y
ir
is
i
m
ag
e
s
[
7
]
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
ad
d
ed
a
n
e
w
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
u
s
in
g
a
n
e
w
u
n
s
u
p
er
v
is
ed
n
e
u
r
al
ap
p
r
o
ac
h
to
d
iv
id
e
th
e
ir
i
s
i
m
ag
e
i
n
t
w
o
r
eg
io
n
s
n
a
m
el
y
ir
is
a
n
d
e
y
e
la
s
h
e
s
r
eg
io
n
,
s
cler
a
an
d
s
k
i
n
,
in
o
r
d
er
to
f
ac
ilit
ate
th
e
d
eter
m
in
at
io
n
o
f
th
e
ir
i
s
co
n
to
u
r
in
t
h
e
p
h
ase
C
a
n
n
y
ed
g
e
d
etec
tio
n
[
6
]
.
Un
li
k
e
o
th
er
ir
i
s
s
e
g
m
en
ta
tio
n
al
g
o
r
ith
m
s
,
t
h
at
p
r
o
ce
s
s
t
h
e
w
h
o
le
i
m
ag
e
o
f
t
h
e
e
y
e
(
w
h
ic
h
co
n
tai
n
s
th
e
n
o
n
-
ir
is
r
eg
io
n
s
th
at
g
e
n
er
ate
s
eg
m
en
tatio
n
er
r
o
r
s
)
s
u
c
h
as Da
u
g
m
an
al
g
o
r
it
h
m
s
[
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
58
–
64
60
2.
I
RIS
SE
G
M
E
NT
A
T
I
O
N
T
h
e
s
eg
m
e
n
tat
io
n
s
ta
g
e
is
cr
itical
to
th
e
s
u
cc
es
s
o
f
an
ir
i
s
r
ec
o
g
n
itio
n
s
y
s
te
m
s
in
ce
d
ata
th
at
is
f
alsel
y
r
ep
r
esen
ted
as
ir
is
p
atter
n
d
ata
w
ill
co
r
r
u
p
t
th
e
b
i
o
m
e
tr
ic
te
m
p
late
s
g
e
n
er
ated
,
r
esu
lti
n
g
in
p
o
o
r
r
ec
o
g
n
itio
n
r
ates.
T
h
e
m
ai
n
m
o
tiv
atio
n
i
n
th
i
s
p
ap
er
is
to
p
r
o
p
o
s
e
a
r
o
b
u
s
t
ir
is
s
eg
m
e
n
tati
o
n
alg
o
r
ith
m
ab
le
to
p
r
o
ce
s
s
ca
p
tu
r
ed
ir
is
i
m
ag
e
s
n
o
n
id
ea
l
e
n
v
ir
o
n
m
en
ts
co
n
d
itio
n
s
.
I
n
t
h
i
s
s
ec
tio
n
,
w
e
p
r
esen
t
a
n
e
w
ir
is
s
eg
m
e
n
tatio
n
ap
p
r
o
ac
h
u
s
ed
a
s
a
p
r
e
-
p
r
o
ce
s
s
i
n
g
s
tep
,
b
ased
o
n
n
e
u
r
al
co
m
p
eti
tiv
e
co
n
ce
p
t
s
.
T
h
is
ap
p
r
o
ac
h
is
d
iv
id
ed
in
to
f
o
u
r
p
h
a
s
es r
esp
e
ctiv
el
y
ca
lled
n
a
m
el
y
:
T
h
e
E
s
ti
m
atio
n
o
f
Un
d
er
l
y
i
n
g
P
r
o
b
ab
ilit
y
De
n
s
it
y
Fu
n
ctio
n
,
E
x
tr
ac
tio
n
o
f
lo
ca
l
m
ax
i
m
a,
T
h
e
m
o
d
es d
etec
tio
n
,
T
h
e
class
if
ica
tio
n
p
r
o
ce
d
u
r
e.
T
h
e
s
eg
m
en
tat
io
n
ap
p
r
o
ac
h
allo
w
s
p
ar
titi
o
n
i
n
g
th
e
ir
is
i
m
ag
e
in
t
w
o
r
e
g
io
n
s
n
a
m
el
y
ir
is
a
n
d
e
y
elas
h
es
r
e
g
io
n
,
s
cler
a
a
n
d
s
k
in
.
T
h
en
,
t
h
e
o
u
tl
in
e
o
f
th
e
e
y
e
is
f
o
u
n
d
u
s
in
g
t
h
e
C
a
n
n
y
ed
g
e,
T
h
e
Ho
u
g
h
T
r
an
s
f
o
r
m
is
e
m
p
lo
y
ed
to
d
et
er
m
in
e
th
e
ce
n
ter
a
n
d
r
ad
iu
s
o
f
t
h
e
p
u
p
il
a
n
d
th
e
ir
i
s
.
Fi
g
u
r
e
3
s
h
o
w
s
a
l
g
o
r
ith
m
u
s
ed
f
o
r
ir
is
r
ec
o
g
n
itio
n
.
Fig
u
r
e
3
.
A
l
g
o
r
ith
m
u
s
ed
f
o
r
i
r
is
r
ec
o
g
n
itio
n
2
.
1
.
I
ris s
eg
m
ent
a
t
io
n us
ing
t
he
un
s
u
perv
is
ed
neura
l a
pp
ro
a
ch
I
n
an
u
n
s
u
p
er
v
is
ed
co
n
tex
t,
th
at
m
ea
n
s
w
h
en
n
o
p
r
io
r
i
n
f
o
r
m
atio
n
ab
o
u
t
th
e
d
ata
s
a
m
p
le,
w
e
in
v
o
l
v
e
th
e
n
eu
r
al
co
m
p
etiti
v
e
c
lu
s
ter
in
g
p
r
o
ce
d
u
r
e
to
th
e
I
r
is
s
e
g
m
en
tatio
n
.
T
h
e
o
b
j
ec
tiv
e
o
f
t
h
ese
ap
p
r
o
ac
h
es
to
class
i
f
y
in
g
t
h
e
i
m
a
g
e
p
ix
e
ls
a
cc
o
r
d
in
g
to
th
eir
d
is
tr
ib
u
t
io
n
in
th
e
r
ep
r
ese
n
tatio
n
s
p
ac
e
an
d
to
ass
ig
n
t
h
e
m
a
lab
el.
I
t
s
tar
ts
f
ir
s
t,
b
y
t
h
e
o
r
g
an
izat
io
n
o
f
t
h
e
p
ix
el
s
o
f
I
r
is
i
m
a
g
e
in
a
n
o
b
s
er
v
atio
n
m
atr
ix
(
ea
c
h
r
o
w
r
ep
r
esen
ts
a
p
i
x
el
a
n
d
ea
ch
c
o
lu
m
n
r
ep
r
esen
ts
a
n
attr
ib
u
te
)
to
esti
m
a
te
th
e
u
n
d
er
l
y
i
n
g
p
r
o
b
a
b
ilit
y
d
en
s
it
y
f
u
n
ctio
n
(
p
d
f
)
o
f
th
e
p
ix
els
d
is
tr
ib
u
tio
n
u
s
i
n
g
a
n
o
n
-
p
ar
a
m
e
tr
ic
esti
m
ato
r
.
I
n
th
e
s
ec
o
n
d
s
tep
,
th
e
p
r
o
c
e
d
u
r
e
u
s
e
s
an
ar
t
if
ic
ial
n
eu
r
al
n
et
wo
r
k
w
ith
co
m
p
etiti
v
e
tr
ai
n
i
n
g
(
NNCT
)
to
ex
tr
ac
t
th
e
lo
ca
l
m
ax
i
m
a
o
f
t
h
e
p
d
f
.
Fo
llo
w
i
n
g
a
m
o
d
es
d
etec
tio
n
m
et
h
o
d
u
s
in
g
a
tech
n
iq
u
e
to
d
etec
t
th
e
ex
i
s
ti
n
g
i
n
ter
n
e
u
r
al
c
o
n
n
ec
tio
n
[
8
]
.
T
h
e
last
s
tep
is
f
o
r
af
f
ec
ti
n
g
t
h
e
r
em
a
i
n
in
g
p
ix
e
ls
to
th
e
ir
class
e
s
.
2
.
1
.
1
.
T
he
esti
m
a
t
io
n o
f
un
d
er
ly
ing
pro
ba
bil
it
y
dens
it
y
f
un
ct
io
n
Af
ter
co
n
s
tr
u
cti
n
g
t
h
e
o
b
s
er
v
a
tio
n
m
a
tr
ix
o
f
an
I
r
is
i
m
a
g
e
p
i
x
els
Γ
=
{
1
,
2
,
…
,
}
co
n
s
id
er
in
g
as
a
s
et
o
f
Q
N
-
d
i
m
e
n
s
io
n
a
l
o
b
s
er
v
atio
n
s
(
in
th
is
ca
s
e
N=
3
)
W
it
h
X
=
[
,
1
,
,
2
,
…
,
,
,
…
,
,
]
,
q
=1
,
2
,
.
.
.
,
Q
an
d
a
p
r
o
b
a
b
ilit
y
d
en
s
it
y
f
u
n
c
tio
n
P
(
X)
,
an
o
n
-
p
ar
a
m
etr
ic
m
eth
o
d
b
ased
o
n
T
h
e
P
ar
ze
n
[
9
]
w
in
d
o
w
u
s
i
n
g
to
esti
m
ate
t
h
is
u
n
d
er
l
y
in
g
d
e
n
s
i
t
y
f
u
n
ct
io
n
.
T
h
e
tec
h
n
iq
u
e
i
s
a
f
a
s
t
es
ti
m
atio
n
alg
o
r
it
h
m
t
h
at
i
s
p
r
o
p
o
s
ed
[
1
0
]
.
First,
t
h
e
r
a
n
g
e
o
f
v
ar
iatio
n
o
f
ea
c
h
co
m
p
o
n
en
t
o
f
t
h
ese
o
b
s
er
v
atio
n
s
is
n
o
r
m
alize
d
to
th
e
i
n
ter
v
al
[
0
,
R
]
,
w
h
er
e
R
is
a
n
i
n
teg
er
s
u
c
h
as
≥
2
,
b
y
m
ea
n
s
o
f
t
h
e
tr
an
s
f
o
r
m
atio
n
d
ef
i
n
ed
as:
,
=
(
,
−
,
)
(
,
−
,
)
∗
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
I
r
is
s
eg
men
ta
tio
n
u
s
in
g
a
n
ew u
n
s
u
p
ervis
ed
n
eu
r
a
l a
p
p
r
o
a
c
h
(
Hich
a
m
Oh
ma
id
)
61
E
ac
h
a
x
is
o
f
th
e
s
o
n
o
r
m
al
ize
d
d
ata
s
p
ac
e
is
t
h
en
p
ar
titi
o
n
ed
in
to
R
e
x
cl
u
s
i
v
e
in
ter
v
als
o
f
u
n
i
t
w
id
t
h
.
T
h
is
d
is
cr
etiza
tio
n
d
ef
in
e
s
a
s
e
t
o
f
h
y
p
er
cu
b
e
o
f
u
n
it
s
id
e
le
n
g
th
.
E
ac
h
h
y
p
er
cu
b
e
n
o
ted
(
)
,
is
a
s
ite
d
ef
i
n
ed
b
y
it
s
N
co
o
r
d
in
ates
1
,
2
,
…
,
,
…
,
w
h
ic
h
ar
e
th
e
in
teg
er
p
ar
ts
o
f
th
e
co
o
r
d
in
ates
o
f
its
ce
n
tr
e
X
.
T
o
b
e
m
o
r
e
s
p
ec
i
f
ic,
let
=
[
1
,
,
2
,
,
…
,
,
,
…
,
,
]
,
q
=1
,
2
,
Q
b
e
th
e
Q
o
b
s
er
v
atio
n
s
in
th
e
n
o
r
m
al
ized
s
p
ac
e.
E
ac
h
o
b
s
er
v
atio
n
Yq
is
f
o
u
n
d
in
s
id
e
a
n
o
n
-
e
m
p
t
y
h
y
p
er
cu
b
e
w
it
h
th
e
co
o
r
d
in
ates
=
(
,
)
,
n
=1
,
2
,
N,
w
h
er
e
(
,
)
d
esig
n
ate
s
th
e
in
te
g
er
p
ar
ts
o
f
,
.
I
f
s
e
v
er
al
o
b
s
er
v
ati
o
n
s
f
a
ll
i
n
th
e
s
a
m
e
h
y
p
er
c
u
b
e,
th
is
o
n
e
ap
p
ea
r
s
m
an
y
ti
m
es
o
n
th
e
lis
t
o
f
n
o
n
–
e
m
p
t
y
h
y
p
er
cu
b
es.
Fu
r
t
h
er
m
o
r
e,
th
e
n
u
m
b
er
o
f
t
i
m
e
s
t
h
e
h
y
p
er
cu
b
e
H(
X)
ap
p
ea
r
s
in
t
h
at
lis
t i
n
d
icate
s
t
h
e
n
u
m
b
er
o
f
d
ata
p
o
in
ts
q
[
H(
X)
]
w
h
ic
h
f
all
s
in
t
h
is
h
y
p
er
c
u
b
e.
Su
b
s
eq
u
e
n
tl
y
,
t
h
e
v
al
u
e
o
f
t
h
e
lo
ca
l d
en
s
it
y
e
s
ti
m
ated
is
:
(
)
=
[
(
)
]
(
2
)
Sin
ce
t
h
e
v
o
l
u
m
e
o
f
H(
X)
i
s
eq
u
al
to
u
n
it
y
.
So
,
th
i
s
f
a
s
t
p
r
o
ce
d
u
r
e
allo
w
s
o
n
l
y
t
h
e
esti
m
a
tio
n
o
f
t
h
e
u
n
d
er
l
y
in
g
p
r
o
b
ab
ilit
y
d
en
s
it
y
f
u
n
ct
io
n
at
t
h
e
ce
n
ter
s
o
f
th
e
n
o
n
-
e
m
p
t
y
h
y
p
er
cu
b
e
wh
o
s
e
n
u
m
b
er
n
e
v
er
ex
ce
ed
s
th
e
n
u
m
b
er
Q
o
f
a
v
ai
lab
le
o
b
s
er
v
atio
n
s
(
p
ix
el
s
)
.
A
t
th
e
ce
n
ter
s
o
f
t
h
e
h
y
p
er
cu
b
e
ce
lls
,
w
h
ich
ar
e
n
o
t
o
n
t
h
at
l
is
t,
th
e
d
en
s
it
y
est
i
m
ates
ar
e
k
n
o
w
n
to
b
e
n
u
l
l.
A
t
th
e
en
d
o
f
th
i
s
f
as
t
al
g
o
r
ith
m
,
all
t
h
e
a
v
ailab
l
e
in
f
o
r
m
atio
n
f
o
r
clu
s
ter
i
n
g
i
s
in
th
e
d
is
cr
ete
s
et
o
f
esti
m
ated
v
al
u
es
o
f
t
h
e
u
n
d
er
l
y
in
g
p
r
o
b
ab
ilit
y
d
en
s
it
y
f
u
n
ctio
n
(
)
.
Fig
u
r
e
4
s
h
o
w
s
a
r
c
h
itect
u
r
e
o
f
t
h
e
s
e
g
m
e
n
t
atio
n
p
r
o
ce
d
u
r
e
Fig
u
r
e
4
.
A
r
ch
itectu
r
e
o
f
t
h
e
s
eg
m
e
n
tat
io
n
p
r
o
ce
d
u
r
e
2
.
1
.
2
.
T
he
ex
t
ra
ct
io
n o
f
lo
ca
l
m
a
x
i
m
a
by
neura
l net
w
o
rk
Ass
i
m
ilat
in
g
th
e
m
o
d
es
to
t
h
e
lo
ca
l
m
ax
i
m
a
o
f
th
e
p
d
f
,
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
u
s
es
t
h
e
Ne
u
r
al
Net
w
o
r
k
s
w
it
h
C
o
m
p
etiti
v
e
T
r
ain
in
g
(
NN
C
T
)
[
1
1
]
.
I
n
th
e
tr
ain
in
g
al
g
o
r
ith
m
,
w
e
w
o
r
k
o
n
l
y
o
n
t
h
e
p
d
f
b
y
p
r
esen
tin
g
s
eq
u
e
n
tial
l
y
,
th
e
ce
n
ter
s
o
f
t
h
e
n
o
n
-
e
m
p
t
y
h
y
p
er
c
u
b
e
o
f
th
e
s
et
X
to
t
h
e
n
et
w
o
r
k
,
i
n
s
tead
o
f
t
h
e
Q
o
b
s
er
v
atio
n
s
.
T
h
e
n
e
u
r
al
n
et
w
o
r
k
i
s
co
m
p
o
s
ed
o
f
t
w
o
la
y
er
s
:
th
e
i
n
p
u
t
la
y
er
an
d
th
e
o
u
tp
u
t
la
y
er
.
T
h
e
f
ir
s
t
o
n
e
is
m
ad
e
o
f
N
u
n
it
s
,
n
=1
,
2
,
.
.
.
,
N,
s
u
ch
th
at
u
n
it
is
s
o
li
cited
b
y
t
h
e
attr
ib
u
te
X
N
o
f
th
e
n
o
n
-
e
m
p
t
y
h
y
p
er
cu
b
e
H(
X)
w
h
e
n
th
i
s
o
n
e
is
p
r
esen
ted
to
th
e
n
et
w
o
r
k
.
Ho
w
e
v
er
,
ea
ch
o
u
tp
u
t
n
e
u
r
o
n
m
ater
ial
izes
a
h
y
p
er
cu
b
e
w
h
ic
h
r
ep
r
esen
t
s
t
h
e
s
ite
o
f
o
n
e
lo
ca
l
m
ax
i
m
u
m
o
f
t
h
e
p
d
f
i
n
t
h
e
s
et
X,
a
n
d
p
r
esen
t
s
its
w
ei
g
h
t
b
y
th
e
m
ea
n
v
ec
to
r
(
)
,
k
=1
,
2
,
…,
K
.
T
h
e
n
u
m
b
er
o
f
th
e
o
u
tp
u
t
n
eu
r
o
n
is
f
ir
s
t
in
itia
lized
ar
b
itr
ar
il
y
.
Du
r
in
g
th
e
tr
ain
i
n
g
p
h
a
s
e,
th
e
o
u
tp
u
t
n
eu
r
o
n
s
en
ter
in
to
co
m
p
etit
io
n
w
it
h
ea
ch
o
th
er
b
y
co
m
p
ar
in
g
th
e
d
is
ta
n
ce
[
(
)
,
(
)
]
,
k
=1
,
2
,
…,
K,
b
et
w
ee
n
t
h
e
in
p
u
t
h
y
p
er
c
u
b
e
H(
X)
an
d
ea
ch
o
u
tp
u
t
n
e
u
r
o
n
(
)
,
th
e
w
in
n
er
is
th
e
clo
s
es
t
o
n
e
to
th
e
h
y
p
e
r
cu
b
e,
th
en
w
e
co
m
p
ar
e
th
e
v
alu
es
o
f
t
h
e
p
d
f
ass
o
ciate
d
to
th
e
w
in
n
er
n
e
u
r
o
n
(
)
an
d
to
H(
X)
.
T
h
e
d
is
ta
n
ce
m
ea
s
u
r
e
u
s
ed
i
n
t
h
i
s
tr
ai
n
i
n
g
al
g
o
r
ith
m
i
s
Ma
h
ala
n
o
b
is
d
is
ta
n
ce
t
h
at
g
i
v
es
th
e
b
est r
es
u
lts
f
o
r
th
e
n
o
n
-
Ga
u
s
s
ian
d
is
tr
ib
u
tio
n
[
1
2
]
in
s
tea
d
o
f
E
u
clid
ian
d
is
ta
n
ce
as i
n
t
h
e
(
NNCT
)
.
2
.
1
.
3
.
Det
ec
t
io
n o
f
s
ig
nifica
nt
m
o
des
o
f
pd
f
T
h
e
ai
m
o
f
th
i
s
p
h
a
s
e
i
s
to
co
n
n
ec
t
ea
ch
g
r
o
u
p
o
f
th
e
clo
s
es
t
m
o
d
es
i
n
s
u
c
h
a
w
a
y
t
h
at
w
e
g
et
a
m
a
p
w
h
ic
h
p
r
eser
v
es
th
e
s
h
ap
e
an
d
s
tr
u
ctu
r
e
o
f
th
e
clas
s
es
ex
is
tin
g
in
th
e
i
m
a
g
e,
b
y
a
n
i
m
p
r
o
v
ed
C
o
m
p
eti
tiv
e
Heb
b
ian
L
ea
r
n
i
n
g
m
eth
o
d
(
C
HL
i
m
)
[
8
]
w
h
ich
allo
w
s
eli
m
i
n
ati
n
g
th
e
i
n
f
lu
e
n
ce
o
f
t
h
e
o
u
t
p
u
t
n
eu
r
al
n
u
m
b
er
.
T
o
g
en
er
ate
t
h
e
i
n
d
u
ce
d
Del
au
n
a
y
tr
ian
g
u
latio
n
,
th
e
C
H
L
i
m
,
g
i
v
en
t
h
e
K
m
o
d
es
d
e
te
cted
b
y
C
NN
as
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
58
–
64
62
p
r
o
to
ty
p
es
i
n
R
N,
s
u
cc
e
s
s
i
v
el
y
ad
d
s
co
n
n
ec
tio
n
s
a
m
o
n
g
th
e
m
.
T
h
e
m
eth
o
d
d
o
es
n
o
t
ch
an
g
e
th
e
w
ei
g
h
t
o
f
p
r
o
to
ty
p
es,
b
u
t
o
n
l
y
g
e
n
er
at
es
to
p
o
lo
g
y
ac
co
r
d
in
g
to
th
ese
p
r
o
to
ty
p
es.
Fo
r
ea
ch
m
o
d
e
H(
)
,
ca
n
b
e
co
n
n
ec
ted
to
th
e
t
w
o
clo
s
est
p
r
o
to
ty
p
es
b
y
a
n
ed
g
e
u
s
i
n
g
M
ah
alan
o
b
is
d
is
tan
ce
as
a
m
ea
s
u
r
e
o
f
r
ese
m
b
la
n
ce
in
s
tead
o
f
E
u
cl
id
ian
d
is
tan
ce
,
it
w
o
r
k
s
a
s
an
ac
ti
v
atio
n
f
u
n
c
t
io
n
f
o
r
co
m
p
eti
tio
n
b
et
w
ee
n
n
eu
r
o
n
s
.
T
h
is
lead
s
to
th
e
in
d
u
ce
d
Dela
u
n
a
y
tr
ia
n
g
u
la
tio
n
,
w
h
ich
i
s
li
m
ited
to
th
o
s
e
r
eg
io
n
s
o
f
th
e
i
n
p
u
t
s
p
ac
e
.
2
.
1
.
4
.
Cla
s
s
if
ica
t
io
n pro
ce
du
re
On
ce
t
h
e
d
if
f
er
e
n
t
m
o
d
e
s
ar
e
id
en
ti
f
ied
,
th
e
c
lass
if
ica
tio
n
m
et
h
o
d
[
1
3
]
th
at
w
e
u
s
e
i
n
t
h
e
p
r
esen
t
w
o
r
k
co
n
s
is
t
s
f
ir
s
t,
in
d
ef
i
n
i
n
g
t
h
e
p
i
x
els
f
a
lli
n
g
in
to
an
y
m
o
d
e
o
f
a
co
n
n
ec
ted
s
et
as
t
h
e
p
r
o
to
t
y
p
e
o
f
o
n
e
clu
s
ter
.
T
h
en
,
th
e
r
e
m
ai
n
i
n
g
p
ix
els,
w
h
ic
h
d
o
n
o
t
f
al
l
i
n
o
n
e
o
f
t
h
e
d
etec
ted
m
o
d
es,
a
r
e
ass
i
g
n
ed
to
th
e
clu
s
ter
s
attac
h
ed
to
th
eir
n
e
ar
est
n
eig
h
b
o
r
am
o
n
g
t
h
ese
p
r
o
to
ty
p
es
b
y
m
ea
n
s
o
f
Ma
h
alan
o
b
is
d
is
ta
n
ce
.
Fig
u
r
e
5
s
h
o
w
s
i
l
lu
s
tr
atio
n
o
f
th
e
I
r
is
s
e
g
m
e
n
tat
io
n
p
r
o
ce
d
u
r
e
.
Fig
u
r
e
5
.
I
llu
s
tr
atio
n
o
f
th
e
I
r
i
s
s
eg
m
e
n
tatio
n
p
r
o
ce
d
u
r
e
2
.
2
.
H
o
ug
h t
ra
ns
f
o
rm
Ho
u
g
h
tr
an
s
f
o
r
m
is
a
s
ta
n
d
ar
d
im
a
g
e
a
n
al
y
s
i
s
to
o
l
f
o
r
f
i
n
d
in
g
cu
r
v
es
t
h
at
ca
n
b
e
d
ef
in
ed
in
a
p
ar
am
etr
ical
f
o
r
m
s
u
c
h
as
li
n
es,
cir
cles,
p
ar
ab
o
las,
an
d
h
y
p
er
b
o
las
[
1
4
]
.
T
h
e
ed
g
e
m
a
p
is
th
en
u
s
ed
i
n
a
v
o
tin
g
p
r
o
ce
s
s
to
m
ax
i
m
ize
t
h
e
d
ef
i
n
ed
Ho
u
g
h
tr
an
s
f
o
r
m
f
o
r
th
e
d
esire
d
co
n
to
u
r
.
C
o
n
s
i
d
er
in
g
t
h
e
o
b
tain
ed
ed
g
e
p
o
in
ts
as (
x
j
;
y
j
)
,
j
=
1
;
2
; ……,
n
,
a
Ho
u
g
h
tr
an
s
f
o
r
m
ca
n
b
e
w
r
itte
n
as:
(
,
,
)
=
∑
ℎ
(
,
,
,
=
1
,
)
(
3
)
W
h
er
e
ℎ
(
,
,
,
,
)
=
{
1
(
,
,
,
,
)
=
0
0
ℎ
(
4
)
T
h
e
li
m
b
u
s
a
n
d
p
u
p
il a
r
e
b
o
th
m
o
d
eled
as c
ir
cle
s
an
d
th
e
p
ar
a
m
etr
ic
f
u
n
ctio
n
g
is
d
e
f
in
ed
a
s
:
(
,
,
,
,
)
=
(
−
)
2
+
(
−
)
2
−
2
(
5
)
Ass
u
m
in
g
a
cir
cle
w
it
h
th
e
ce
n
ter
(
x
c;
y
c)
an
d
r
ad
iu
s
r
,
t
h
e
ed
g
e
p
o
in
t
s
t
h
at
ar
e
lo
ca
te
d
o
v
er
th
e
cir
cle
r
esu
l
t
i
n
a
ze
r
o
v
al
u
e
o
f
th
e
f
u
n
ctio
n
.
T
h
e
v
a
lu
e
o
f
g
i
s
t
h
en
tr
an
s
f
o
r
m
ed
to
1
b
y
t
h
e
h
f
u
n
ctio
n
,
w
h
ich
r
ep
r
esen
ts
th
e
lo
ca
l
p
a
tter
n
o
f
th
e
co
n
to
u
r
.
T
h
e
lo
ca
l
p
atter
n
s
ar
e
th
e
n
u
s
ed
in
a
v
o
tin
g
p
r
o
ce
d
u
r
e
u
s
in
g
th
e
Ho
u
g
h
tr
an
s
f
o
r
m
,
H,
in
o
r
d
er
to
lo
ca
te
th
e
p
r
o
p
er
p
u
p
il a
n
d
lim
b
u
s
b
o
u
n
d
ar
ies.
I
n
o
r
d
er
to
d
etec
t li
m
b
u
s
,
o
n
l
y
v
er
tical
ed
g
e
in
f
o
r
m
at
io
n
is
u
s
ed
.
T
h
e
u
p
p
er
an
d
l
o
w
er
p
ar
ts
,
w
h
ich
h
av
e
t
h
e
h
o
r
izo
n
tal
ed
g
e
in
f
o
r
m
atio
n
,
ar
e
u
s
u
all
y
co
v
er
ed
b
y
t
h
e
t
w
o
e
y
elid
s
.
T
h
e
h
o
r
izo
n
tal
ed
g
e
in
f
o
r
m
atio
n
is
u
s
ed
f
o
r
d
etec
tin
g
t
h
e
u
p
p
er
an
d
lo
w
er
e
y
elid
s
,
w
h
ich
ar
e
m
o
d
eled
as
p
ar
ab
o
lic
ar
cs [
6
-
1
4
]
.
Fig
u
r
e
6
s
h
o
w
s
e
x
a
m
p
le
s
o
f
t
h
e
s
e
g
m
e
n
ted
ir
is
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
I
r
is
s
eg
men
ta
tio
n
u
s
in
g
a
n
ew u
n
s
u
p
ervis
ed
n
eu
r
a
l a
p
p
r
o
a
c
h
(
Hich
a
m
Oh
ma
id
)
63
Fig
u
r
e
6
.
E
x
a
m
p
les o
f
th
e
s
eg
m
en
ted
ir
is
,
(
a)
R
ea
l i
m
a
g
es,
(
b
)
Dete
ctio
n
o
f
ir
is
r
eg
io
n
s
,
(
c)
s
eg
m
e
n
ted
i
m
ag
e
s
f
r
o
m
U
B
I
R
I
S,
(
d
)
No
is
e
d
etec
tio
n
an
d
r
em
o
v
a
l u
s
in
g
li
n
ea
r
Ho
u
g
h
tr
an
s
f
o
r
m
3.
RE
SU
L
T
S
T
h
e
f
ir
s
t
s
ta
g
e
o
f
ir
is
r
ec
o
g
n
it
io
n
i
s
to
is
o
late
th
e
ac
t
u
al
ir
i
s
r
eg
io
n
i
n
a
d
ig
ita
l
e
y
e
i
m
a
g
e
.
T
h
e
ir
is
r
eg
io
n
ca
n
b
e
ap
p
r
o
x
im
ated
b
y
t
w
o
cir
cles,
o
n
e
f
o
r
th
e
ir
is
/
s
cler
a
b
o
u
n
d
ar
y
an
d
an
o
th
er
,
in
ter
io
r
to
th
e
f
ir
s
t,
f
o
r
th
e
ir
is
/p
u
p
il
b
o
u
n
d
ar
y
.
T
h
e
e
y
elid
s
a
n
d
e
y
elas
h
es
n
o
r
m
all
y
o
cc
l
u
d
e
th
e
u
p
p
er
an
d
l
o
w
er
p
ar
ts
o
f
th
e
ir
i
s
r
eg
io
n
.
T
h
e
s
u
cc
e
s
s
e
s
o
f
s
e
g
m
en
tatio
n
d
ep
en
d
o
n
t
h
e
i
m
a
g
in
g
q
u
alit
y
o
f
e
y
e
i
m
a
g
es.
T
h
e
d
atab
ase
o
f
th
e
e
y
e
i
m
a
g
es
f
o
r
th
is
p
r
o
j
ec
t
h
as
b
ee
n
tak
e
n
f
r
o
m
th
e
UB
I
R
I
S
d
atab
ase
to
ev
alu
ate
t
h
e
e
f
f
ec
tiv
e
n
e
s
s
o
f
t
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
.
in
o
r
d
er
to
m
ak
e
t
h
e
cir
cle
d
etec
tio
n
s
tep
m
o
r
e
ac
cu
r
ate
a
n
d
ef
f
ic
ien
t,
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
tar
ts
w
it
h
d
eter
m
i
n
in
g
th
e
r
e
g
io
n
s
o
f
t
h
e
e
y
e
u
s
i
n
g
u
n
s
u
p
er
v
is
ed
n
eu
r
al
ap
p
r
o
ac
h
,
th
e
Ho
u
g
h
tr
an
s
f
o
r
m
f
o
r
th
e
ir
is
/s
cler
a
b
o
u
n
d
ar
y
w
a
s
p
er
f
o
r
m
ed
,
th
e
n
th
e
Ho
u
g
h
tr
an
s
f
o
r
m
f
o
r
t
h
e
ir
is
/p
u
p
il
b
o
u
n
d
ar
y
w
a
s
p
er
f
o
r
m
ed
w
it
h
i
n
th
e
ir
i
s
r
eg
io
n
,
i
n
s
tead
o
f
t
h
e
w
h
o
le
ey
e
r
eg
io
n
,
s
i
n
ce
th
e
p
u
p
il
is
al
w
a
y
s
w
i
th
in
t
h
e
ir
i
s
r
eg
io
n
.
A
f
ter
th
is
p
r
o
ce
s
s
w
as
co
m
p
lete,
s
i
x
p
ar
a
m
eter
s
ar
e
s
to
r
ed
,
th
e
r
ad
iu
s
an
d
x
an
d
y
ce
n
ter
co
o
r
d
in
ates
fo
r
b
o
th
cir
cles.
E
y
elid
s
w
er
e
is
o
lated
b
y
f
ir
s
t
f
itt
in
g
a
lin
e
to
th
e
u
p
p
er
an
d
lo
w
er
e
y
e
lid
u
s
i
n
g
t
h
e
lin
ea
r
Ho
u
g
h
tr
a
n
s
f
o
r
m
.
A
s
ec
o
n
d
h
o
r
izo
n
tal
lin
e
is
th
e
n
d
r
a
w
n
,
w
h
ic
h
i
n
ter
s
ec
t
s
w
it
h
t
h
e
f
ir
s
t
lin
e
a
t
t
h
e
ir
is
ed
g
e
th
at
i
s
clo
s
es
t
to
th
e
p
u
p
il.
T
h
is
p
r
o
ce
s
s
is
d
o
n
e
f
o
r
b
o
th
t
h
e
to
p
an
d
b
o
tto
m
e
y
elid
s
.
T
h
e
s
ec
o
n
d
h
o
r
izo
n
tal
lin
e
allo
w
s
m
a
x
i
m
u
m
i
s
o
latio
n
o
f
e
y
e
lid
r
eg
io
n
s
.
Fig
u
r
e
7
s
h
o
w
s
e
x
a
m
p
le
s
o
f
co
r
r
ec
t seg
m
en
ted
ir
is
es
.
Fig
u
r
e
7
.
E
x
a
m
p
les o
f
co
r
r
ec
t seg
m
e
n
ted
ir
is
es
UB
I
R
I
S
ir
is
d
atab
ase
i
s
u
s
ed
to
ev
al
u
ate
t
h
e
p
r
o
p
o
s
ed
s
eg
m
e
n
tat
io
n
al
g
o
r
ith
m
.
T
h
is
d
atab
ase
is
co
m
p
o
s
ed
o
f
1
2
1
0
m
a
g
es
co
ll
ec
ted
f
r
o
m
2
4
2
e
y
es.
T
ab
le
1
p
r
esen
ts
a
co
m
p
ar
is
o
n
b
et
w
e
en
th
e
ac
c
u
r
ac
y
o
f
th
e
s
e
g
m
en
ta
tio
n
f
o
r
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
an
d
s
o
m
e
p
r
ev
i
o
u
s
alg
o
r
it
h
m
s
[
1
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
58
–
64
64
Ta
b
le
1
.
C
o
m
p
ar
is
o
n
b
et
w
ee
n
th
e
ac
cu
r
ac
y
o
f
p
r
o
p
o
s
ed
alg
o
r
ith
m
a
n
d
s
o
m
e
p
r
ev
io
u
s
al
g
o
r
ith
m
s
M
e
t
h
o
d
A
c
c
u
r
a
c
y
(
%)
D
a
u
g
ma
n
9
5
.
2
2
C
a
m
u
s
a
n
d
W
i
l
d
e
s
9
6
.
7
8
M
a
r
t
i
n
-
R
o
c
h
e
P
r
o
p
o
se
d
7
7
.
1
8
9
7
.
8
3
4.
CO
NCLU
SI
O
NS
I
r
is
p
atter
n
s
h
a
v
e
b
ee
n
p
r
o
v
en
to
b
e
u
s
e
f
u
l
a
n
d
ac
cu
r
ate
in
r
ec
o
g
n
it
io
n
o
f
th
e
p
er
s
o
n
s
[
6
]
.
Au
to
m
a
tin
g
th
e
p
r
o
ce
s
s
r
eq
u
ir
es
r
o
b
u
s
t
i
m
p
le
m
en
tatio
n
o
f
t
h
e
ir
is
s
eg
m
e
n
tatio
n
,
n
o
r
m
aliza
t
io
n
,
a
n
d
r
ec
o
g
n
itio
n
p
h
ases
.
I
f
o
n
e
p
h
ase
is
lack
i
n
g
i
n
ac
cu
r
ac
y
o
r
p
r
ec
is
io
n
,
th
e
ef
f
ec
t
iv
e
n
es
s
o
f
th
e
s
y
s
te
m
a
s
a
w
h
o
le
is
d
ec
r
ea
s
ed
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
o
f
s
e
g
m
en
tatio
n
ad
d
ed
a
n
e
w
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
to
s
eg
m
e
n
tatio
n
s
tep
s
w
h
ic
h
is
clu
s
ter
i
n
g
th
e
ir
is
i
m
a
g
e
u
s
i
n
g
u
n
s
u
p
er
v
is
ed
n
e
u
r
al
ap
p
r
o
ac
h
.
A
f
ter
,
T
h
e
Ho
u
g
h
T
r
an
s
f
o
r
m
i
s
e
m
p
lo
y
ed
to
d
et
er
m
in
e
t
h
e
ce
n
ter
an
d
r
ad
iu
s
o
f
t
h
e
p
u
p
il
an
d
t
h
e
ir
is
.
Fin
a
ll
y
,
E
y
elid
s
(
u
p
p
er
an
d
lo
w
er
)
w
er
e
is
o
lated
u
s
i
n
g
th
e
li
n
ea
r
Ho
u
g
h
tr
an
s
f
o
r
m
.
T
h
e
alg
o
r
ith
m
o
f
s
eg
m
e
n
tati
o
n
p
r
esen
ted
in
t
h
i
s
p
ap
er
p
e
r
f
o
r
m
s
w
ell
o
n
m
o
s
t i
r
is
i
m
a
g
es,
t
h
e
s
e
g
m
e
n
tat
io
n
s
y
s
te
m
f
ails
w
h
e
n
i
m
ag
e
s
h
a
v
e
s
ev
er
e
o
cc
lu
s
io
n
.
RE
F
E
R
E
NC
E
S
[1
]
F
A
N
G
,
Bin
e
t
T
A
N
G
,
Yu
a
n
Ya
n
.
Im
p
ro
v
e
d
c
las
s
sta
ti
stics
e
stim
a
t
io
n
f
o
r
sp
a
rse
d
a
ta
p
r
o
b
lem
s
in
o
ff
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
.
IEE
E
T
r
a
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
,
a
n
d
Cy
b
e
rn
e
ti
c
s,
Pa
rt
C
(
Ap
p
li
c
a
ti
o
n
s
a
n
d
Rev
i
e
ws)
,
2
0
0
5
,
v
o
l.
3
5
,
n
o
3
,
p
.
2
7
6
-
2
8
6
.
[2
]
GA
O,
X
in
b
o
,
ZHON
G
,
Ju
a
n
ju
a
n
,
L
I,
Jie
,
e
t
a
l.
F
a
c
e
s
k
e
tch
s
y
n
t
h
e
sis
a
lg
o
rit
h
m
b
a
s
e
d
o
n
E
-
HM
M
a
n
d
se
lec
ti
v
e
e
n
se
m
b
le.
IEE
E
T
r
a
n
sa
c
ti
o
n
s
o
n
Circ
u
it
s a
n
d
S
y
ste
ms
fo
r V
i
d
e
o
T
e
c
h
n
o
l
o
g
y
,
2
0
0
8
,
v
o
l.
1
8
,
n
o
4
,
p
.
4
8
7
-
4
9
6
.
[3
]
YU
,
L
i,
ZH
AN
G
,
D
a
v
id
,
e
t
WANG
,
Ku
a
n
q
u
a
n
.
T
h
e
re
lativ
e
d
istan
c
e
o
f
k
e
y
p
o
in
t
b
a
se
d
iri
s
re
c
o
g
n
it
io
n
.
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
2
0
0
7
,
v
o
l.
4
0
,
n
o
2
,
p
.
4
2
3
-
4
3
0
.
[4
]
ZHA
N
G
,
T
a
ip
in
g
,
F
A
N
G
,
Bin
,
L
IU,
W
e
in
in
g
,
e
t
a
l.
T
o
tal
v
a
riatio
n
n
o
rm
-
b
a
se
d
n
o
n
n
e
g
a
ti
v
e
m
a
t
rix
f
a
c
to
riz
a
ti
o
n
f
o
r
id
e
n
ti
fy
in
g
d
isc
rim
in
a
n
t
re
p
re
se
n
tatio
n
o
f
im
a
g
e
p
a
tt
e
rn
s.
Ne
u
ro
c
o
mp
u
ti
n
g
,
2
0
0
8
,
v
o
l.
7
1
,
n
o
1
0
,
p
.
1
8
2
4
-
1
8
3
1
.
[5
]
DA
U
G
M
A
N,
J.
Bio
m
e
tri
c
p
e
rso
n
a
l
id
e
n
ti
f
ica
ti
o
n
sy
ste
m
b
a
se
d
o
n
iri
s
a
n
a
ly
sis,
Un
it
e
d
S
tate
s
P
a
ten
t
,
US
5
2
9
1
5
6
0
.
1
9
9
4
.
[6
]
M
A
S
EK,
L
ib
o
r,
e
t
a
l.
Re
c
o
g
n
it
io
n
o
f
h
u
m
a
n
iri
s p
a
tt
e
rn
s f
o
r
b
i
o
m
e
tri
c
id
e
n
ti
f
ica
ti
o
n
.
2
0
0
3
.
[7
]
S
A
HMOUD
,
S
h
a
a
b
a
n
A
.
e
t
ABU
H
A
IB
A
,
Ib
ra
h
im
S
.
E
ff
icie
n
t
iri
s
se
g
m
e
n
tatio
n
m
e
th
o
d
in
u
n
c
o
n
stra
i
n
e
d
e
n
v
iro
n
m
e
n
ts.
Pa
tt
e
rn
Rec
o
g
n
it
i
o
n
,
2
0
1
3
,
v
o
l.
4
6
,
n
o
1
2
,
p
.
3
1
7
4
-
3
1
8
5
.
[8
]
T
i
m
o
u
y
a
s
M
.
,
Ed
d
a
ro
u
ich
S
.
a
n
d
Ha
m
m
o
u
c
h
A
.
(2
0
1
7
).
“
M
o
d
e
re
g
io
n
d
e
tec
ti
o
n
u
sin
g
im
p
ro
v
e
d
Co
m
p
e
ti
ti
v
e
He
b
b
ian
L
e
a
rn
in
g
f
o
r
u
n
su
p
e
rv
ise
d
c
lu
ste
rin
g
”
,
En
g
in
e
e
rin
g
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
:
An
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
(
ES
T
IJ
)
,
IS
S
N:
2
2
5
0
-
3
4
9
8
Vo
l.
7
,
No
.
4
,
p
p
.
2
6
-
35.
[9
]
P
a
rz
e
n
,
E.
(
1
9
6
2
)
‘A
n
Esti
m
a
ti
o
n
o
f
a
P
r
o
b
a
b
il
it
y
De
n
sity
F
u
n
c
ti
o
n
a
n
d
M
o
d
e
’,
An
n
.
M
a
t
h
.
S
t
a
t
.
,
v
o
l.
3
3
,
p
p
.
1
0
6
5
-
1
0
7
6
.
[1
0
]
P
o
sta
ire,
J.
-
G
.
,
&
V
a
ss
e
u
r,
C.
P
.
A
.
(1
9
8
2
)
‘A
f
a
st
A
l
g
o
rit
h
m
f
o
r
n
o
n
P
a
ra
m
e
tri
c
P
ro
b
a
b
il
it
y
De
n
sity
Esti
m
a
ti
o
n
’,
IEE
E,
T
r
a
n
s.
o
n
Pa
tt
e
rn
A
n
a
l
.
a
n
d
M
a
c
h
i
n
e
In
tel
.
P
A
M
I
-
4
,
n
°
6
,
p
p
.
6
6
3
-
6
6
6
.
[1
1
]
Ed
d
a
ro
u
ich
,
S
.
,
&
S
b
ih
i,
A
.
(2
0
0
7
)
‘Ne
u
ra
l
Ne
tw
o
rk
f
o
r
M
o
d
e
s
De
tec
ti
o
n
in
P
a
tt
e
r
n
Clas
sif
ic
a
ti
o
n
’.
ICT
IS
’0
7
,
M
o
ro
c
c
o
,
F
e
z
,
3
-
5
p
p
.
3
0
0
-
3
0
3
.
[1
2
]
T
i
m
o
u
y
a
s,
M
.
,
Ed
d
a
ro
u
ic
h
,
S
.
,
&
Ha
m
m
o
u
c
h
,
A
.
(2
0
1
2
).
A
n
e
w
a
p
p
ro
a
c
h
o
f
c
las
si
f
ica
ti
o
n
f
o
r
n
o
n
-
G
a
u
ss
ian
d
istri
b
u
ti
o
n
u
p
o
n
c
o
m
p
e
ti
ti
v
e
train
in
g
,
(ICCS
’
1
2
)
,
A
g
a
d
ir,
M
o
r
o
c
c
o
,
p
p
.
1
-
6.
[1
3
]
Ed
d
a
ro
u
ich
,
S
.
,
&
S
b
i
h
i,
A
.
(2
0
0
2
)
‘Dé
tec
ti
o
n
d
e
s
M
o
d
e
s
p
a
r
A
p
p
ro
c
h
e
Ne
u
ro
n
a
le
p
o
u
r
la
Cl
a
ss
if
i
c
a
ti
o
n
d
e
s
Do
n
n
é
e
s
d
’u
n
M
é
lan
g
e
d
e
s
Dist
rib
u
ti
o
n
s
No
rm
a
les
’
,
P
ro
c
e
e
d
in
g
o
f
th
e
6
th
Af
rica
n
Co
n
f
e
re
n
c
e
o
n
Re
se
a
rc
h
in
Co
m
p
u
ter S
c
ien
c
e
(CA
RI’0
2
),
Ya
o
u
n
d
é
,
Ca
m
e
ro
o
n
,
p
p
.
6
1
-
6
8
[1
4
]
W
I
L
DES
,
Rich
a
rd
P
.
Iris
re
c
o
g
n
i
ti
o
n
:
a
n
e
m
e
rg
in
g
b
io
m
e
tri
c
te
c
h
n
o
l
o
g
y
.
P
ro
c
e
e
d
in
g
s
o
f
th
e
IEE
E,
1
9
9
7
,
v
o
l.
8
5
,
n
o
9
,
p
.
1
3
4
8
-
1
3
6
3
.
[1
5
]
Hu
g
o
P
r
o
e
n
c
a
,
T
o
wa
rd
s
n
o
n
-
c
o
o
p
e
ra
ti
v
e
b
io
m
e
tri
c
iri
s
re
c
o
g
n
it
io
n
(
P
h
D
th
e
sis),
Un
iv
e
rsit
y
o
f
Be
ira
In
terio
r,
Oc
to
b
e
r
2
0
0
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.