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.
10
,
No
.
4
,
A
u
g
u
s
t
2
0
2
0
,
p
p
.
3
5
5
8
~
3
5
6
7
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
4
.
p
p
3
5
5
8
-
3
5
6
7
3558
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m/in
d
ex
.
p
h
p
/I
JE
C
E
Ex
tracted
f
ea
tur
e
s
b
a
sed
m
ulti
-
cla
ss
c
la
ss
ificatio
n
o
f
o
rthodo
ntic
i
m
a
g
es
H
icha
m
R
iri
1
,
M
o
ha
mm
ed
E
d
-
D
ha
hra
o
uy
2
,
Abdel
m
a
j
id
E
l
m
o
uta
o
ua
kk
il
3
,
A
bd
er
ra
hi
m
B
eni
-
H
s
s
a
ne
4
,
F
a
rid B
o
urzg
ui
5
1
,
2,
3,
4
L
A
ROSE
RI
L
a
b
o
ra
to
ry
,
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
s,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s,
Ch
o
u
a
i
b
Do
u
k
k
a
li
Un
iv
e
rsity
,
M
o
ro
c
c
o
5
De
p
a
rtme
n
t
o
f
Orth
o
d
o
n
ti
c
s,
F
a
c
u
lt
y
o
f
De
n
ti
str
y
,
Un
iv
e
rsit
y
o
f
H
a
ss
a
n
II
A
in
Ch
o
k
,
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
J
u
l 2
7
,
2
0
1
9
R
ev
i
s
ed
Dec
2
,
2
0
1
9
A
cc
ep
ted
Feb
1
,
2
0
2
0
T
h
e
p
u
rp
o
se
o
f
th
is
st
u
d
y
is
to
in
v
e
stig
a
te
c
o
m
p
u
ter
v
isio
n
a
n
d
m
a
c
h
in
e
lea
rn
in
g
m
e
th
o
d
s
f
o
r
c
las
si
f
ic
a
ti
o
n
o
f
o
rth
o
d
o
n
ti
c
im
a
g
e
s
in
o
rd
e
r
to
p
ro
v
id
e
o
rth
o
d
o
n
ti
sts
w
it
h
a
so
l
u
ti
o
n
f
o
r
m
u
lt
i
-
c
las
s
c
las
sif
i
c
a
ti
o
n
o
f
p
a
ti
e
n
ts’
im
a
g
e
s
to
e
v
a
lu
a
te
th
e
e
v
o
lu
ti
o
n
o
f
th
e
ir
trea
t
m
e
n
t.
Of
w
h
ich
,
w
e
p
ro
p
o
se
d
th
re
e
a
lg
o
rit
h
m
s
b
a
se
d
o
n
e
x
trac
ted
f
e
a
tu
re
s,
su
c
h
a
s
f
a
c
ial
fe
a
tu
re
s
a
n
d
sk
in
c
o
lo
u
r
u
sin
g
YCb
Crc
o
l
o
u
r
sp
a
c
e
,
a
ss
ig
n
e
d
to
n
o
d
e
s
o
f
a
d
e
c
isio
n
tree
to
c
las
si
fy
o
rth
o
d
o
n
ti
c
im
a
g
e
s:
a
n
a
l
g
o
rit
h
m
f
o
r
in
tra
-
o
ra
l
im
a
g
e
s,
a
n
a
lg
o
rit
h
m
f
o
r
m
o
u
ld
im
a
g
e
s
a
n
d
a
n
a
lg
o
rit
h
m
f
o
r
e
x
tra
-
o
ra
l
i
m
a
g
e
s.
T
h
e
n
,
w
e
c
o
m
p
a
re
d
o
u
r
m
e
th
o
d
b
y
i
m
p
lem
e
n
ti
n
g
th
e
L
o
c
a
l
Bin
a
r
y
P
a
tt
e
rn
(L
BP
)
a
lg
o
rit
h
m
to
e
x
trac
t
tex
tu
ra
l
f
e
a
tu
re
s
f
ro
m
i
m
a
g
e
s.
Af
ter
th
a
t,
w
e
a
p
p
li
e
d
th
e
p
rin
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
ly
sis
(P
CA
)
a
lg
o
rit
h
m
to
o
p
ti
m
ize
th
e
re
d
u
n
d
a
n
t
p
a
ra
m
e
ters
in
o
rd
e
r
to
c
las
sify
LBP
f
e
a
tu
re
s
w
it
h
six
c
las
sif
ier
s;
Qu
a
d
ra
ti
c
S
u
p
p
o
rt
V
e
c
to
r
M
a
c
h
in
e
(
S
V
M
),
Cu
b
ic
S
V
M
,
Ra
d
ial
Ba
sis
F
u
n
c
t
io
n
S
V
M
,
Co
sin
e
K
-
Ne
a
re
st
Ne
ig
h
b
o
u
rs
(KN
N),
Eu
c
li
d
ia
n
KN
N,
a
n
d
L
in
e
a
r
Disc
ri
m
in
a
n
t
A
n
a
l
y
sis
(
L
D
A
).
T
h
e
p
re
se
n
ted
a
lg
o
rit
h
m
s
h
a
v
e
b
e
e
n
e
v
a
lu
a
ted
o
n
a
d
a
tas
e
t
o
f
i
m
a
g
e
s
o
f
9
8
d
iff
e
re
n
t
p
a
ti
e
n
ts,
a
n
d
e
x
p
e
rim
e
n
tal
re
su
lt
s
d
e
m
o
n
stra
te
th
e
g
o
o
d
p
e
rf
o
r
m
a
n
c
e
s
o
f
o
u
r
p
ro
p
o
se
d
m
e
th
o
d
w
it
h
a
h
ig
h
a
c
c
u
ra
c
y
c
o
m
p
a
re
d
w
it
h
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
ri
t
h
m
s.
W
h
e
re
L
D
A
c
las
si
f
ie
r
a
c
h
iev
e
s a
n
a
c
c
u
ra
c
y
o
f
8
4
.
5
%
.
K
ey
w
o
r
d
s
:
Diag
n
o
s
is
aid
e
I
m
ag
e
a
n
al
y
s
i
s
K
NN
L
i
n
ea
r
d
is
cr
i
m
i
n
a
n
t a
n
a
l
y
s
is
SVM
Co
p
y
rig
h
t
©
2
0
2
0
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
:
Hich
a
m
R
ir
i
,
D
ep
ar
t
m
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
s
,
C
h
o
u
aib
Do
u
k
k
ali
U
n
i
v
er
s
it
y
,
J
ab
r
an
Kh
alil J
ab
r
an
av
en
u
e,
B
.
P
2
9
9
-
2
4
0
0
0
E
l J
ad
i
d
a,
Mo
r
o
cc
o
.
E
m
ail:
r
ir
i.h
@
u
cd
.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
I
m
ag
e
s
et
cla
s
s
i
f
icatio
n
p
r
o
b
le
m
ca
n
b
e
s
o
l
v
ed
b
y
co
m
b
i
n
in
g
d
i
f
f
er
en
t
f
ield
s
s
u
ch
as
co
m
p
u
ter
v
is
io
n
alg
o
r
it
h
m
s
,
m
ac
h
i
n
e
l
ea
r
n
in
g
clas
s
i
f
ier
s
an
d
i
m
ag
e
p
r
o
ce
s
s
in
g
[
1
-
6
]
.
W
ith
th
eir
th
r
ee
w
ell
-
k
n
o
w
n
s
tep
s
:
tr
an
s
f
o
r
m
a
n
d
p
r
esen
t
im
ag
e
s
ap
p
r
o
p
r
iately
,
ex
tr
ac
t
m
ea
n
in
g
f
u
l
in
f
o
r
m
atio
n
f
r
o
m
i
m
ag
e
s
,
an
d
u
s
i
n
g
th
ese
e
x
tr
ac
ted
f
ea
tu
r
e
s
to
b
u
il
d
a
m
o
d
el
f
o
r
class
i
f
icatio
n
[
7
-
12
]
.
I
t
ca
n
b
e
class
i
f
ied
i
n
to
t
w
o
ca
te
g
o
r
ies:
p
ar
am
etr
ic
a
n
d
n
o
n
-
p
ar
a
m
etr
i
c
m
et
h
o
d
s
.
A
n
d
i
t
ca
n
b
e
als
o
ca
teg
o
r
ized
b
y
t
y
p
e
o
f
t
h
e
tr
ea
ted
p
r
o
b
lem
o
f
class
i
f
icatio
n
: b
in
ar
y
c
lass
if
ic
atio
n
o
r
m
u
l
ti
-
c
lass
cla
s
s
i
f
icat
i
o
n
.
M
an
y
m
e
th
o
d
s
f
o
r
i
m
ag
e
s
et
class
i
f
icat
io
n
ar
e
p
r
o
p
o
s
ed
.
Me
f
r
az
et
al.
p
r
ese
n
ted
a
n
o
v
el
s
p
ar
s
e
n
o
n
p
ar
a
m
etr
ic
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
cla
s
s
i
f
ier
[
6
]
.
A
l
s
o
,
W
an
g
a
n
d
Z
h
a
n
g
[
1
3
]
p
r
esen
ted
a
m
et
h
o
d
f
o
r
i
m
a
g
e
r
ec
o
g
n
itio
n
an
d
c
lass
if
icatio
n
b
ased
o
n
i
m
p
r
o
v
ed
B
ag
o
f
Feat
u
r
es
(
B
OF)
.
I
n
ad
d
itio
n
to
th
a
t,
Af
f
o
n
s
o
et
al.
co
m
p
ar
e
d
ee
p
l
ea
r
n
in
g
ar
c
h
itect
u
r
e
w
i
th
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
e
s
:
Dec
is
io
n
tr
ee
i
n
d
u
ctio
n
alg
o
r
ith
m
s
,
Ne
u
r
al
Net
w
o
r
k
s
,
Nea
r
est n
ei
g
h
b
o
u
r
s
an
d
S
u
p
p
o
r
t v
ec
to
r
m
ac
h
i
n
e
s
[
1
4
].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
E
xtra
cted
fea
tu
r
es b
a
s
ed
mu
lti
-
cla
s
s
cla
s
s
ifica
tio
n
o
f o
r
th
o
d
o
n
tic
ima
g
es
(
Hich
a
m
R
ir
i
)
3559
Dah
m
an
e
et
a
l.
p
r
esen
ted
a
te
ch
n
iq
u
e
to
ca
r
r
y
o
u
t
h
ea
d
p
o
s
e
esti
m
atio
n
b
ased
o
n
t
h
e
o
r
ie
n
tatio
n
o
f
th
e
s
y
m
m
etr
y
a
x
is
t
h
at
in
d
icat
es
th
e
r
o
ll
an
g
le
o
f
th
e
h
ea
d
.
T
h
e
s
y
m
m
etr
ical
ar
ea
o
f
th
e
f
ac
e
w
it
h
r
esp
ec
t
to
th
is
o
r
ien
tatio
n
o
f
f
er
s
t
h
e
m
f
ea
tu
r
es
s
u
c
h
a
s
t
h
e
w
id
t
h
o
f
r
eg
io
n
w
h
ich
a
llo
w
t
h
e
m
to
class
i
f
y
a
n
d
t
h
en
,
to
p
r
ed
ict
y
a
w
an
g
le
s
[
1
5
]
.
Mo
r
eo
v
er
,
ac
co
r
d
in
g
to
th
e
e
x
h
au
s
ti
v
e
r
e
v
ie
w
[
1
6
]
,
m
o
s
t
f
ac
e
d
etec
tio
n
m
et
h
o
d
s
ca
n
b
e
ca
teg
o
r
ized
in
to
f
o
u
r
m
aj
o
r
class
es:
k
n
o
w
led
g
e
-
b
ased
m
et
h
o
d
s
,
f
ea
t
u
r
e
in
v
a
r
ian
t
ap
p
r
o
ac
h
es,
te
m
p
late
m
atc
h
i
n
g
m
et
h
o
d
s
a
n
d
ap
p
ea
r
an
ce
-
b
ased
m
et
h
o
d
s
.
Kn
o
w
led
g
e
-
b
ased
m
et
h
o
d
s
ar
e
b
ased
o
n
r
u
les
d
ed
u
cted
f
r
o
m
k
n
o
w
led
g
e
ab
o
u
t
th
e
co
m
p
o
n
en
ts
o
f
a
t
y
p
i
ca
l
f
ac
e
an
d
h
o
w
a
f
ac
e
ca
n
ap
p
ea
r
in
th
e
i
m
ag
e.
Gen
er
all
y
,
t
h
e
r
u
les
b
ased
o
n
th
e
r
elatio
n
s
h
ip
a
m
o
n
g
t
h
e
f
a
cial
s
tr
u
ct
u
r
es
[
1
7
]
.
Featu
r
e
in
v
ar
ian
t
ap
p
r
o
ac
h
es
ar
e
d
esig
n
ed
to
f
i
n
d
s
tr
u
ct
u
r
a
l
f
ea
t
u
r
es
t
h
at
ex
i
s
t
e
v
en
d
u
r
i
n
g
th
e
p
o
s
e,
th
e
v
ie
w
an
g
le
o
r
th
e
ill
u
m
in
a
tio
n
co
n
d
itio
n
s
ch
a
n
g
e,
a
n
d
t
h
en
th
e
y
u
s
e
t
h
ese
in
v
ar
ia
n
t
f
ea
tu
r
es
to
lo
ca
te
t
h
e
p
o
s
it
i
o
n
s
o
f
f
ac
es
[
1
8
]
.
T
em
p
l
ate
m
atch
in
g
m
et
h
o
d
s
;
th
e
m
ai
n
id
ea
o
f
t
h
ese
ap
p
r
o
ac
h
es
i
s
to
cr
ea
te
s
ta
n
d
ar
d
m
o
d
els
ab
le
to
d
escr
ib
e
a
f
ac
e
o
r
a
f
ac
e
p
o
r
tio
n
.
T
h
en
,
th
e
co
r
r
elatio
n
b
et
w
ee
n
t
h
e
i
n
p
u
t
i
m
ag
e
an
d
th
e
m
o
d
el
i
s
co
m
p
u
ted
.
T
h
an
k
s
to
th
is
,
f
ac
es
ar
e
id
en
ti
f
ied
in
th
e
i
m
ag
e
[
1
9
].
A
p
p
ea
r
an
ce
-
b
ase
d
m
et
h
o
d
s
[
2
0
]
;
th
e
p
r
in
cip
le
o
f
th
ese
m
et
h
o
d
s
is
to
co
n
s
id
er
th
e
p
r
o
b
lem
o
f
f
ac
e
d
etec
tio
n
as
a
class
i
f
icat
io
n
p
r
o
b
lem
.
I
t
is
to
class
i
f
y
a
ca
p
tu
r
ed
m
o
d
el
in
o
n
e
o
f
t
w
o
cla
s
s
e
s
: f
ac
e
cla
s
s
a
n
d
n
o
n
-
f
ac
e
class
.
T
h
e
ai
m
o
f
o
u
r
s
tu
d
y
i
s
to
d
es
ig
n
r
ec
o
g
n
itio
n
tec
h
n
iq
u
es
o
f
ex
tr
a
-
o
r
al,
m
o
u
ld
,
an
d
i
n
tr
a
-
o
r
al
i
m
ag
e
s
,
in
o
r
d
er
to
p
r
o
v
id
e
o
r
th
o
d
o
n
tis
ts
w
it
h
a
s
o
lu
tio
n
f
o
r
th
e
class
i
f
icatio
n
p
r
o
b
lem
b
ased
o
n
t
y
p
e
o
f
p
atien
t’
s
i
m
a
g
es
to
ev
alu
a
te
th
e
ev
o
l
u
t
io
n
o
f
p
atien
t
s
tat
u
s
u
s
i
n
g
i
m
ag
es
o
f
p
r
ev
io
u
s
v
is
its
.
T
h
is
w
il
l
g
i
v
e
ef
f
icie
n
c
y
an
d
s
a
v
e
t
h
e
ti
m
e
i
n
t
h
eir
m
an
ip
u
latio
n
s
.
T
h
ese
tec
h
n
iq
u
es
ar
e
b
ased
o
n
ex
tr
ac
ted
f
ea
tu
r
es
a
n
d
s
i
m
p
l
e
m
at
h
e
m
a
tical
m
e
th
o
d
s
.
T
h
er
ef
o
r
e,
in
th
is
p
ap
er
,
w
e
p
r
esen
t
t
h
r
ee
lo
ca
l
alg
o
r
ith
m
s
b
ased
o
n
ex
tr
ac
ted
f
ea
tu
r
e
s
o
r
g
an
ized
w
i
th
a
d
ec
is
io
n
tr
ee
r
ep
r
esen
tatio
n
to
class
if
y
an
d
r
ec
o
g
n
ize
s
i
x
-
teen
clas
s
es
o
f
o
r
th
o
d
o
n
tic
i
m
a
g
es:
an
al
g
o
r
ith
m
f
o
r
in
tr
a
-
o
r
al
i
m
ag
e
s
,
t
h
en
a
o
n
e
f
o
r
m
o
u
ld
i
m
a
g
es
a
n
d
a
n
al
g
o
r
ith
m
f
o
r
ex
tr
a
-
o
r
al
i
m
a
g
es.
T
h
en
w
e
co
m
b
in
e
t
h
o
s
e
alg
o
r
ith
m
s
to
h
av
e
o
n
e
s
tr
u
ct
u
r
e
f
o
r
all
class
es
o
f
o
r
th
o
d
o
n
tic
im
ag
e
s
an
d
ev
al
u
ate
th
e
co
n
f
u
s
io
n
o
f
t
h
e
d
ec
is
io
n
b
et
w
ee
n
all
cla
s
s
e
s
.
T
h
is
h
ier
ar
ch
ical
r
ep
r
esen
tatio
n
ca
n
b
e
in
ter
p
r
eted
as
a
s
e
t
o
f
h
ier
ar
ch
ical
t
y
p
es
s
to
r
ed
i
n
lea
v
es
tr
ee
s
tr
u
ct
u
r
e.
B
y
u
s
in
g
s
ev
er
al
ex
tr
ac
ted
f
ea
t
u
r
es
f
r
o
m
o
r
th
o
d
o
n
tic
i
m
a
g
es,
s
u
ch
a
s
in
v
ar
ia
n
t
f
ac
i
al
f
ea
tu
r
e
s
an
d
s
k
i
n
co
lo
u
r
u
s
i
n
g
Y
C
b
C
r
co
lo
u
r
s
p
ac
e.
T
h
en
,
w
e
co
m
p
ar
ed
o
u
r
m
et
h
o
d
b
y
i
m
p
le
m
e
n
tin
g
s
ix
class
i
f
ier
s
o
f
m
ac
h
i
n
e
lear
n
i
n
g
;
Qu
ad
r
atic
S
u
p
p
o
r
t
Vec
to
r
Ma
ch
in
e
(
S
VM
)
,
C
u
b
ic
SV
M,
R
ad
ial
B
asis
Fu
n
ct
io
n
(
R
B
F)
SVM,
C
o
s
i
n
e
K
-
Nea
r
est
Nei
g
h
b
o
u
r
s
,
E
u
cl
id
ian
KNN,
an
d
L
i
n
ea
r
Dis
cr
i
m
in
a
n
t
An
al
y
s
i
s
(
L
D
A
)
.
B
y
u
s
i
n
g
t
h
e
L
o
ca
l
B
in
ar
y
P
atter
n
(
L
B
P
)
to
ex
tr
ac
t
f
ea
tu
r
es
f
r
o
m
i
m
a
g
es.
Af
ter
t
h
e
f
ea
t
u
r
es
e
x
tr
ac
tio
n
s
tep
,
w
e
ap
p
lied
th
e
p
r
i
n
cip
al
co
m
p
o
n
e
n
t
an
al
y
s
is
(
P
C
A
)
a
l
g
o
r
ith
m
to
o
p
ti
m
ize
th
e
r
ed
u
n
d
a
n
t
p
ar
a
m
eter
s
in
o
r
d
er
to
class
if
y
L
B
P
f
ea
t
u
r
es
w
it
h
t
h
at
s
i
x
class
if
ier
s
.
T
h
e
p
r
esen
ted
alg
o
r
it
h
m
s
h
av
e
b
ee
n
e
v
al
u
ated
o
n
a
lar
g
e
d
ataset
o
f
i
m
a
g
es
o
f
d
if
f
er
e
n
t
p
atien
t
s
,
an
d
ex
p
er
i
m
en
tal
r
esu
lt
s
d
e
m
o
n
s
tr
ate
th
e
g
o
o
d
p
er
f
o
r
m
an
ce
s
o
f
o
u
r
p
r
o
p
o
s
ed
m
et
h
o
d
w
ith
a
h
ig
h
ac
cu
r
ac
y
.
T
h
e
r
est
o
f
th
is
p
ap
er
is
o
r
g
an
ized
as
f
o
llo
w
s
.
T
h
e
m
ater
ia
ls
u
s
ed
to
estab
lis
h
th
e
ex
p
er
i
m
en
ts
ar
e
p
r
esen
ted
i
n
s
ec
tio
n
2
.
I
n
S
ec
tio
n
3
,
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
o
f
clas
s
i
f
icatio
n
o
f
th
e
t
h
r
ee
m
ain
clas
s
es
(
m
o
u
ld
,
i
n
tr
a
-
o
r
al
a
n
d
e
x
tr
a
-
o
r
al)
o
f
o
r
th
o
d
o
n
tic
i
m
a
g
es
is
p
r
ese
n
ted
.
T
h
en
,
th
e
r
es
ea
r
ch
m
eth
o
d
s
ar
e
p
r
esen
ted
in
s
ec
tio
n
4
.
T
h
en
,
th
e
e
x
p
er
i
m
e
n
tal
r
es
u
lt
s
a
n
d
c
o
r
r
esp
o
n
d
in
g
a
n
al
y
s
es
ar
e
p
r
o
v
id
ed
in
Sectio
n
5
.
A
t la
s
t,
co
n
cl
u
d
in
g
r
e
m
ar
k
s
ar
e
p
r
esen
ted
i
n
Sectio
n
6
.
2.
M
AT
E
RIAL
2
.
1
.
Da
t
a
s
a
m
ple descr
iptio
n
A
ll
p
r
esen
ted
ap
p
r
o
ac
h
es a
r
e
t
r
ain
ed
u
s
in
g
1
2
0
7
i
m
ag
e
o
f
9
8
d
iffer
en
t p
atie
n
ts
a
s
cited
i
n
T
ab
le
1
f
o
r
ea
ch
m
ai
n
-
cla
s
s
.
t
h
e
i
m
a
g
es
a
r
e
ac
q
u
ir
ed
w
i
th
d
i
g
ital
ca
m
er
as
r
esp
ec
tin
g
t
h
e
p
r
o
to
co
l
d
escr
ib
ed
in
[
21
]
w
it
h
r
eso
lu
tio
n
s
v
ar
ied
b
et
w
ee
n
1
0
2
4
*
7
6
8
an
d
3
8
8
8
*
2
5
9
2
in
R
GB
co
lo
u
r
s
p
ac
e
i
n
J
P
E
G
f
o
r
m
at.
T
ab
les
an
d
f
i
g
u
r
es a
r
e
p
r
esen
ted
ce
n
ter
,
as sh
o
w
n
i
n
T
ab
le
1
an
d
Fig
u
r
e
1
,
an
d
cited
in
th
e
m
a
n
u
s
cr
ip
t
b
ef
o
r
e
ap
p
ea
r
e
d
.
T
ab
le
1
.
A
s
u
m
m
ar
y
o
f
t
h
e
o
r
th
o
d
o
n
tic
d
ataset: c
las
s
es a
n
d
n
u
m
b
er
o
f
s
a
m
p
les p
er
m
a
in
-
cl
ass
M
a
i
n
c
l
a
ss
C
l
a
ss l
a
b
e
l
N
°
o
f
i
mag
e
s
Ex
t
r
a
-
o
r
a
l
1
:
F
a
c
e
,
2
:
F
a
c
e
w
i
t
h
smil
e
,
3
:
T
h
r
e
e
-
q
u
a
r
t
e
r
,
4
:
T
h
r
e
e
-
q
u
a
r
t
e
r
w
i
t
h
s
mi
l
e
,
5
:
R
i
g
h
t
p
r
o
f
i
l
e
,
6
:
L
e
f
t
P
r
o
f
i
l
e
3
2
5
I
n
t
r
a
-
o
r
a
l
7
:
F
r
o
n
t
,
8
:
L
e
f
t
,
9
:
M
a
n
d
i
b
u
l
a
r
,
1
0
:
M
a
x
i
l
l
a
r
y
,
1
1
:
R
i
g
h
t
.
8
1
2
M
o
u
l
d
1
2
:
F
r
o
n
t
,
1
3
:
L
e
f
t
,
1
4
:
M
a
n
d
i
b
u
l
a
r
,
1
5
:
M
a
x
i
l
l
a
r
y
,
1
6
:
R
i
g
h
t
.
70
3.
P
RO
P
O
SE
D
M
E
T
H
O
D
I
n
th
i
s
s
ec
tio
n
,
w
e
d
escr
ib
e
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
to
ex
tr
ac
t
f
ea
tu
r
e
s
an
d
class
if
y
o
r
th
o
d
o
n
tic
i
m
a
g
es.
So
,
to
id
en
tify
a
n
i
m
ag
e
t
y
p
e
a
m
o
n
g
s
i
x
-
tee
n
d
if
f
er
en
t
t
y
p
e
s
o
f
o
r
th
o
d
o
n
tic
im
ag
e
s
ill
u
s
tr
ated
i
n
Fig
u
r
e
1
,
w
e
d
ef
i
n
e
a
h
ier
ar
ch
y
o
f
t
h
ese
t
y
p
es
as
ill
u
s
tr
ated
in
Fi
g
u
r
e
2
o
f
clas
s
i
f
ic
atio
n
tr
ee
b
a
s
ed
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
A
u
g
u
s
t 2
0
2
0
:
3
5
5
8
-
3567
3560
ex
tr
ac
ted
f
ea
t
u
r
es.
Firstl
y
,
w
e
p
r
o
p
o
s
e
f
o
r
ea
ch
m
ai
n
cla
s
s
an
al
g
o
r
it
h
m
to
e
x
tr
ac
t
f
ea
tu
r
es
an
d
cla
s
s
i
f
y
it
s
class
es
i
n
s
ec
tio
n
s
3
.
4
.
2
,
3
.
4
.
3
an
d
3
.
4
.
4
.
T
h
en
w
e
co
m
b
i
n
e
th
ese
al
g
o
r
ith
m
s
in
o
n
e
alg
o
r
i
th
m
to
class
i
f
y
all
class
es o
f
o
r
th
o
d
o
n
tic
i
m
ag
e
s
in
s
ec
tio
n
3
.
4
.
5
.
Fig
u
r
e
1
.
Sa
m
p
le
i
m
ag
e
s
o
f
ea
ch
class
i
n
o
u
r
d
ataset: e
x
tr
a
-
o
r
al:
1
: Fac
e,
2
: Fac
e
w
it
h
s
m
il
e,
3
:
T
h
r
ee
-
q
u
ar
ter
,
4
:
T
h
r
ee
-
q
u
ar
te
r
w
it
h
s
m
ile,
5
: Ri
g
h
t p
r
o
f
ile,
6
: L
ef
t P
r
o
f
ile; i
n
tr
a
-
o
r
al:
7
: Fr
o
n
t,
8
: L
ef
t,
9
:
Ma
n
d
ib
u
lar
,
1
0
: M
ax
illar
y
,
1
1
: Ri
g
h
t;
m
o
u
ld
: 1
2
: Fr
o
n
t,
1
3
:
L
e
f
t,
1
4
: M
an
d
ib
u
lar
,
1
5
: M
ax
illar
y
,
1
6
: Ri
g
h
t;
Fig
u
r
e
2
.
C
lass
if
ica
tio
n
tr
ee
o
f
o
r
th
o
d
o
n
tic
i
m
a
g
es
3
.
1
.
Sk
in co
lo
ur
det
ec
t
io
n us
ing
YCbCrc
o
lo
ur
s
pa
ce
T
h
e
o
r
th
o
g
o
n
al
co
lo
u
r
s
p
ac
es
r
ed
u
ce
th
e
r
ed
u
n
d
an
c
y
p
r
esen
t
in
R
GB
co
lo
u
r
ch
an
n
el
s
an
d
r
ep
r
esen
t
th
e
co
lo
u
r
w
it
h
s
ta
tis
tica
ll
y
i
n
d
ep
en
d
en
t
co
m
p
o
n
e
n
t
s
.
A
s
th
e
lu
m
i
n
an
ce
an
d
th
e
c
h
r
o
m
in
a
n
ce
co
m
p
o
n
e
n
t
s
ar
e
ex
p
licitl
y
s
ep
ar
ated
.
T
h
e
YC
b
C
r
co
lo
u
r
s
p
ac
e
is
o
n
e
o
f
th
e
m
o
s
t
w
id
el
y
u
s
ed
ch
o
ices
f
o
r
s
k
i
n
d
etec
tio
n
[2
2
,
2
3
]
. T
h
u
s
,
w
e
u
s
e
t
h
e
r
an
g
es
o
f
C
r
a
n
d
C
b
v
al
u
es
th
at
co
r
r
esp
o
n
d
to
s
k
in
-
co
lo
u
r
r
ef
er
en
ce
m
ap
in
t
h
e
C
b
C
r
p
lan
e
d
ef
i
n
ed
b
y
C
h
ai
an
d
Ng
a
n
[2
4
]
w
h
i
ch
ar
e
[
7
7
;
1
2
7
]
an
d
[
1
3
3
;
1
7
3
]
r
esp
ec
tiv
el
y
.
T
o
c
o
m
p
u
te
th
e
v
al
u
es
o
f
Y
C
b
C
r
co
m
p
o
n
e
n
ts
o
f
an
i
m
a
g
e
f
r
o
m
R
GB
co
m
p
o
n
e
n
t
s
,
w
e
u
s
e
t
h
e
f
o
llo
w
in
g
eq
u
atio
n
[2
5
,
2
6
]
:
[
]
=
[
16
128
128
]
+
1
256
[
65
.
7
3
8
129
.
057
25
.
0
6
4
−
37
.
9
4
5
−
74
.
4
9
4
112
.
4
3
9
112
.
439
−
94
.
1
5
4
−
18
.
2
8
5
]
[
]
(1
)
3
.
2
.
Cla
s
s
if
ica
t
io
n m
et
ho
d o
f
m
o
ul
d i
m
a
g
es
I
n
th
i
s
s
u
b
s
ec
tio
n
,
w
e
h
a
v
e
f
i
v
e
t
y
p
es
o
f
m
o
u
ld
in
g
i
m
a
g
es
to
class
i
f
y
:
m
a
x
illar
y
,
m
a
n
d
ib
u
lar
,
f
r
o
n
t
,
r
ig
h
t,
a
n
d
lef
t.
So
,
to
d
is
ti
n
g
u
i
s
h
t
h
ese
cla
s
s
es,
w
e
u
s
e
te
n
f
e
atu
r
es i
n
alg
o
r
it
h
m
1
as
f
o
llo
ws:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
E
xtra
cted
fea
tu
r
es b
a
s
ed
mu
lti
-
cla
s
s
cla
s
s
ifica
tio
n
o
f o
r
th
o
d
o
n
tic
ima
g
es
(
Hich
a
m
R
ir
i
)
3561
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
u
p
p
er
h
alf
o
f
t
h
e
i
m
a
g
e
µ
up
=
2
ℎ
∑
∑
(
,
)
<
=
0
<
ℎ
2
=
0
(
2
)
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
l
o
w
er
h
al
f
o
f
t
h
e
i
m
ag
e
µ
lo
=
2
ℎ
∑
∑
(
,
)
<
=
0
<
ℎ
=
ℎ
2
(
3
)
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
f
o
u
r
co
r
n
er
s
o
f
th
e
i
m
a
g
e
w
it
h
a
s
ize
eq
u
als to
s
²:
µ
lu
=
1
2
∑
∑
(
,
)
<
=
0
<
=
0
(4
)
µ
ru
=
1
2
∑
∑
(
,
)
<
=
−
<
=
0
(5
)
µ
ld
=
1
2
∑
∑
(
,
)
<
=
0
<
ℎ
=
ℎ
−
(6
)
µ
rd
=
1
2
∑
∑
(
,
)
<
=
−
<
ℎ
=
ℎ
−
(7
)
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
l
ef
t a
n
d
r
ig
h
t q
u
ar
ter
s
o
f
th
e
i
m
ag
e
µ
lq
=
4
ℎ
∑
∑
(
,
)
<
4
=
0
<
ℎ
=
0
(8
)
µ
rq
=
4
ℎ
∑
∑
(
,
)
<
=
3
4
<
ℎ
=
0
(9
)
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
l
ef
t h
a
lf
o
f
th
e
i
m
a
g
e
µ
l
=
2
ℎ
∑
∑
(
,
)
<
2
=
0
<
ℎ
=
0
(1
0
)
-
T
h
e
m
ea
n
o
f
t
h
e
p
ix
el
s
in
t
h
e
r
ig
h
t h
al
f
o
f
t
h
e
i
m
ag
e
µ
r
=
2
ℎ
∑
∑
(
,
)
<
=
2
<
ℎ
=
0
(1
1
)
W
ith
I
is
th
e
m
atr
i
x
o
f
p
ix
el
s
i
n
g
r
a
y
s
ca
le
co
lo
u
r
s
p
ac
e,
h
i
s
t
h
e
i
m
a
g
e
h
eig
h
t,
a
n
d
w
i
s
th
e
i
m
ag
e
w
id
t
h
.
3
.
3
.
Cla
s
s
if
ica
t
io
n m
et
ho
d o
f
intr
a
-
o
ra
l im
a
g
es
As
f
o
r
m
o
u
ld
i
m
a
g
es
in
t
h
is
s
u
b
s
ec
tio
n
w
e
h
a
v
e
f
i
v
e
cla
s
s
e
s
f
o
r
in
tr
a
-
o
r
al
i
m
a
g
es:
in
tr
a
m
ax
illar
y
,
in
tr
a
m
an
d
ib
u
lar
,
in
tr
a
r
ig
h
t,
in
tr
a
f
r
o
n
t
an
d
i
n
t
r
a
le
f
t.
W
e
p
r
o
p
o
s
e
an
ap
p
r
o
ac
h
b
a
s
ed
o
n
alg
o
r
it
h
m
2
to
class
i
f
y
th
e
s
e
clas
s
es.
W
e
d
if
f
er
e
n
tiate
b
et
w
ee
n
t
h
e
in
tr
a
m
ax
illar
y
a
n
d
th
e
in
tr
a
m
a
n
d
ib
u
lar
b
y
s
t
u
d
y
i
n
g
th
e
m
o
n
o
to
n
y
o
f
t
h
e
f
u
n
c
tio
n
ch
ar
ac
ter
ized
b
y
:
(
)
=
(1
2
)
W
ith
i
an
d
j
ar
e
th
e
p
ix
e
l
co
o
r
d
in
ates
o
f
teet
h
b
y
d
etec
ti
n
g
t
h
eir
ex
ter
io
r
ed
g
e
i
n
t
h
e
l
ef
t
h
alf
o
f
th
e
i
m
ag
e
b
y
u
s
i
n
g
m
o
r
p
h
o
lo
g
ical
f
ilter
o
n
s
eg
m
e
n
ted
teet
h
b
y
t
h
r
es
h
o
ld
in
g
.
I
f
t
h
is
f
u
n
ctio
n
is
i
n
cr
ea
s
i
n
g
,
it
is
an
i
m
a
g
e
o
f
in
tr
a
m
a
n
d
ib
u
lar
t
y
p
e.
E
ls
e,
i
f
t
h
is
f
u
n
c
tio
n
is
d
ec
r
ea
s
i
n
g
,
i
t
is
a
n
i
m
ag
e
o
f
in
tr
a
m
ax
illar
y
.
E
ls
e,
if
th
is
f
u
n
ct
io
n
i
s
i
n
cr
ea
s
in
g
o
n
a
p
ar
t
o
f
t
h
e
le
f
t
h
al
f
o
f
th
e
i
m
ag
e
a
n
d
d
ec
r
ea
s
i
n
g
o
n
t
h
e
o
th
er
,
it
is
a
n
i
m
a
g
e
class
o
f
th
e
FLR (
Face
L
e
f
t Ri
g
h
t)
n
o
d
e
o
f
th
e
tr
ee
ill
u
s
tr
ated
in
Fig
u
r
e
2
.
Fo
r
th
e
F
L
R
n
o
d
e,
w
e
u
s
e
th
e
lo
ca
tio
n
o
f
p
ix
e
ls
t
h
at
co
r
r
esp
o
n
d
to
th
e
s
k
i
n
co
lo
u
r
.
Fo
r
th
e
in
tr
a
r
ig
h
t
i
m
a
g
es,
t
h
e
d
en
tal
r
etr
ac
to
r
is
p
r
esen
t
i
n
th
e
le
f
t
s
id
e
o
f
th
e
i
m
a
g
e
s
.
W
h
ile
i
n
th
e
r
ig
h
t
s
id
e,
th
er
e
is
th
e
i
n
ter
io
r
p
ar
t
o
f
th
e
m
o
u
th
.
T
h
er
ef
o
r
e,
w
e
h
a
v
e
i
n
t
h
e
l
ef
t
s
id
e
a
lack
o
f
p
ix
el
s
t
h
at
co
r
r
esp
o
n
d
to
s
k
in
co
lo
u
r
u
n
l
ik
e
t
h
e
o
t
h
er
s
id
e
th
e
p
ix
els
h
av
in
g
t
h
e
s
k
i
n
co
lo
u
r
ar
e
in
ab
u
n
d
a
n
ce
.
T
h
e
s
a
m
e
th
i
n
g
f
o
r
in
tr
a
lef
t
t
y
p
e,
ex
ce
p
t
th
at
i
n
t
h
is
ca
s
e
th
e
r
etr
ac
to
r
i
s
i
n
t
h
e
r
i
g
h
t
s
id
e,
w
h
ich
lead
s
to
a
p
r
esen
ce
o
f
p
i
x
el
s
t
h
a
t
co
r
r
esp
o
n
d
to
s
k
in
co
lo
u
r
in
a
b
u
n
d
an
ce
i
n
th
e
le
f
t
s
id
e.
I
n
ad
d
itio
n
,
f
o
r
th
e
in
tr
a
f
r
o
n
t
i
m
a
g
es,
w
e
d
is
tin
g
u
is
h
th
is
cla
s
s
b
y
u
s
i
n
g
t
h
e
f
ac
t
o
f
p
r
esen
ce
o
f
p
ix
els
t
h
at
co
r
r
e
s
p
o
n
d
to
s
k
in
-
co
lo
u
r
w
it
h
ap
p
r
o
x
i
m
atel
y
s
i
m
ilar
d
is
tr
ib
u
tio
n
i
n
t
h
e
b
o
th
s
id
es o
f
th
e
i
m
a
g
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
A
u
g
u
s
t 2
0
2
0
:
3
5
5
8
-
3567
3562
3
.
4
.
Cla
s
s
if
ica
t
io
n m
et
ho
d o
f
ex
t
ra
-
o
ra
l im
a
g
es
I
n
th
i
s
s
u
b
s
ec
tio
n
,
w
e
p
r
o
p
o
s
e
an
ap
p
r
o
ac
h
o
f
class
if
icat
io
n
d
etailed
in
alg
o
r
it
h
m
3
f
o
r
ex
tr
a
-
o
r
a
l
i
m
a
g
es.
W
e
h
a
v
e
s
i
x
cla
s
s
e
s
;
lef
t
p
o
r
tr
ait
p
r
o
f
ile,
r
i
g
h
t
p
o
r
tr
ait
p
r
o
f
ile,
f
ac
e
p
r
o
f
ile,
T
h
r
ee
-
Qu
ar
ter
p
r
o
f
i
le
(
4
5
°
o
r
ien
tatio
n
)
,
f
ac
e
p
r
o
f
ile
w
it
h
s
m
ile,
a
n
d
th
r
ee
-
q
u
ar
ter
w
it
h
s
m
ile.
W
e
d
is
ti
n
g
u
i
s
h
t
h
ese
cla
s
s
es
o
f
ex
tr
a
-
o
r
al
i
m
a
g
es
b
y
co
m
p
u
tin
g
t
h
e
n
u
m
b
er
o
f
p
i
x
els
co
r
r
esp
o
n
d
in
g
to
s
k
i
n
-
co
lo
u
r
i
n
th
e
le
f
t
u
p
p
er
q
u
ar
ter
o
f
th
e
i
m
a
g
e
(
1
)
an
d
th
e
n
u
m
b
e
r
o
f
p
ix
els
co
r
r
esp
o
n
d
in
g
to
s
k
in
-
co
lo
u
r
in
t
h
e
r
ig
h
t
u
p
p
er
q
u
ar
ter
(
2
)
.
I
f
th
e
q
u
o
tie
n
t
o
f
s
1
b
y
s
2
is
l
ess
t
h
an
a
ce
r
tai
n
th
r
e
s
h
o
ld
α
5
,
it
is
a
class
o
f
r
i
g
h
t
p
o
r
tr
ait
p
r
o
f
ile.
Oth
er
w
is
e,
if
t
h
is
q
u
o
tie
n
t
is
g
r
ea
ter
t
h
a
n
a
ce
r
tai
n
t
h
r
esh
o
ld
α
4
,
it
is
a
class
o
f
le
f
t
p
o
r
tr
ait
p
r
o
f
il
e.
Oth
er
w
is
e,
if
th
i
s
q
u
o
tien
t
is
b
et
w
ee
n
b
o
th
th
r
esh
o
ld
s
α
5
an
d
α
4
.
I
t
is
ab
o
u
t
FT
Q
(
Face
T
h
r
ee
-
Qu
ar
ter
)
n
o
d
e
o
f
th
e
tr
ee
illu
s
tr
ated
in
Fi
g
u
r
e
2
.
Fo
r
th
e
FT
Q
n
o
d
e,
w
e
d
is
t
in
g
u
is
h
t
h
e
p
o
r
tr
ait
f
ac
e
f
r
o
m
t
h
r
ee
-
q
u
ar
ter
b
y
p
er
f
o
r
m
i
n
g
a
m
eth
o
d
th
at
lo
ca
tes
th
e
p
ix
el
s
b
elo
n
g
i
n
g
to
th
e
lef
t
an
d
r
ig
h
t
e
y
e.
B
y
u
s
in
g
s
o
m
e
f
ea
t
u
r
es
s
u
ch
as
p
r
esen
ce
o
f
t
h
e
s
k
i
n
ar
o
u
n
d
th
e
e
y
e,
p
r
esen
ce
o
f
cr
y
s
tal
lin
e
a
n
d
w
h
ite
ar
ea
in
s
id
e
e
y
e’
s
ar
ea
.
T
h
en
w
e
b
ase
o
n
th
e
d
is
tr
ib
u
tio
n
o
f
p
ix
els
co
r
r
esp
o
n
d
i
n
g
to
s
k
i
n
co
lo
u
r
in
th
e
le
f
t
s
id
e
o
f
le
f
t
e
y
e
a
n
d
r
i
g
h
t
s
id
e
o
f
r
ig
h
t
e
y
e
to
d
i
f
f
er
en
tiat
e
p
o
r
tr
ait
f
ac
e
f
r
o
m
p
o
r
tr
ait
th
r
ee
-
q
u
ar
ter
.
T
h
en
w
e
lo
ca
te
t
h
e
m
o
u
t
h
b
y
u
s
in
g
t
h
e
m
et
h
o
d
o
f
s
k
in
d
etec
tio
n
b
ased
o
n
YC
b
C
r
co
lo
u
r
s
p
ac
e.
E
s
p
ec
iall
y
o
n
th
e
m
o
u
th
’
s
ar
ea
,
C
r
v
al
u
e
i
n
cr
ea
s
es,
n
ea
r
l
y
t
h
e
d
o
u
b
le,
f
r
o
m
o
t
h
er
p
ar
ts
o
f
f
ac
e.
T
h
en
w
e
lo
o
k
if
th
er
e
is
a
p
r
esen
ce
o
f
teet
h
in
th
is
ar
ea
to
k
n
o
w
i
f
it
is
ab
o
u
t
a
s
m
il
e
ca
s
e
o
r
n
o
t.
3
.
5
.
Cla
s
s
if
ica
t
io
n m
et
ho
d o
f
o
rt
ho
do
ntic
i
m
a
g
es
T
o
r
esu
m
e,
w
e
d
if
f
er
en
tiate
b
et
w
ee
n
t
h
ese
t
h
r
ee
m
ain
cla
s
s
es
(
ex
t
ra
-
o
r
al,
i
n
tr
a
-
o
r
al,
a
n
d
m
o
u
ld
)
b
y
u
s
i
n
g
alg
o
r
it
h
m
4
t
h
at
d
is
tin
g
u
i
s
h
es
ex
tr
a
-
o
r
al
i
m
a
g
e
s
f
r
o
m
o
t
h
er
s
b
y
u
s
i
n
g
t
h
e
f
ac
t
o
f
th
e
p
r
esen
ce
o
f
b
ac
k
g
r
o
u
n
d
a
n
d
s
k
i
n
co
lo
u
r
in
th
e
i
m
a
g
e.
U
n
li
k
e
e
x
tr
a
-
o
r
al
i
m
ag
e
s
,
i
n
i
n
tr
a
-
o
r
al
i
m
ag
e
s
t
h
er
e
i
s
n
o
b
ac
k
g
r
o
u
n
d
,
w
h
ile
t
h
er
e
is
a
m
as
s
iv
e
p
r
ese
n
ce
o
f
p
ix
e
l
s
th
at
co
r
r
esp
o
n
d
to
s
k
i
n
c
o
lo
u
r
.
I
n
co
n
tr
ast,
f
o
r
th
e
m
o
u
ld
i
m
a
g
es
t
h
er
e
ar
e
n
o
p
ix
els
t
h
at
co
r
r
esp
o
n
d
to
s
k
i
n
co
lo
u
r
.
So
in
ea
ch
ca
s
e
w
e
ca
l
l
th
e
ap
p
r
o
p
r
iate
alg
o
r
ith
m
to
wo
r
k
.
4.
RE
S
E
ARCH
M
E
T
H
O
D
4
.
1
.
L
inea
r
Dis
cr
i
m
ina
nt
a
n
a
ly
s
i
s
Dis
cr
i
m
in
a
n
t
an
a
l
y
s
is
i
s
a
class
i
f
icatio
n
m
et
h
o
d
.
T
h
e
m
o
d
el
f
o
r
d
is
cr
i
m
in
a
n
t
an
al
y
s
i
s
ass
u
m
e
s
s
a
m
p
le
s
ha
ve
a
Ga
u
s
s
ian
m
i
x
t
u
r
e
d
is
tr
ib
u
tio
n
.
F
o
r
lin
ea
r
d
is
cr
i
m
i
n
an
t
an
al
y
s
is
(
F
is
c
h
er
d
is
cr
i
m
i
n
an
t)
,
th
e
m
o
d
el
h
as t
h
e
s
a
m
e
co
v
ar
ian
ce
m
atr
i
x
f
o
r
ea
ch
class
,
o
n
l
y
t
h
e
m
ea
n
s
v
a
r
y
.
4
.
2
.
Su
pp
o
rt
v
ec
t
o
r
m
a
chine
s
Su
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
es
(
SV
M)
is
a
s
u
p
er
v
i
s
ed
m
ac
h
i
n
e
l
ea
r
n
in
g
cla
s
s
i
f
ier
t
h
at
ca
n
b
e
u
s
ed
f
o
r
class
i
f
icatio
n
an
d
r
eg
r
ess
io
n
.
SVM
u
tili
ze
s
a
p
r
io
r
i
in
f
o
r
m
atio
n
a
s
a
clu
s
ter
lab
els
f
o
r
th
e
tr
ain
in
g
s
tep
.
T
h
e
n
,
it
g
e
n
er
ate
b
y
u
s
i
n
g
t
h
e
s
u
p
p
o
r
t
v
ec
to
r
s
an
o
p
ti
m
al
h
y
p
er
p
la
n
e
s
o
th
at
th
e
e
x
p
ec
ted
class
i
f
icatio
n
er
r
o
r
f
o
r
u
n
lab
elled
test
s
a
m
p
les s
h
o
u
ld
b
e
m
i
n
i
m
ized
[
2
7
]
.
SVM
clas
s
i
f
ier
is
u
s
u
all
y
u
s
ed
f
o
r
b
in
ar
y
o
r
m
u
lti
-
clas
s
class
i
f
icatio
n
p
r
o
b
lem
s
.
I
n
t
h
is
p
ap
er
,
w
e
i
m
p
le
m
en
t t
h
r
ee
t
y
p
e
s
o
f
S
VM
class
i
f
ier
b
ased
o
n
th
e
f
o
l
lo
w
i
n
g
k
er
n
e
ls
:
-
P
o
ly
n
o
m
ial
Qu
ad
r
atic
K
(
X
,
Y
)
=
(
γ
X
i
T
Y
j
+
c
)
2
,
γ
>
0
(1
3
)
C
u
b
ic
K
(
X
,
Y
)
=
(
γ
X
i
T
Y
j
+
c
)
3
,
γ
>
0
(
14
)
-
Gau
s
s
ia
n
o
r
R
ad
ial
B
asis
Fu
n
c
tio
n
(
R
B
F)
K
(
X
,
Y
)
=
e
(
−
γ
|
|
X
i
−
Y
j
|
|
2
)
,
γ
>
0
(
1
5
)
W
ith
γ
an
d
c
ar
e
k
er
n
el
p
ar
a
m
eter
s
.
4
.
3
.
K
-
Nea
re
s
t
Neig
hb
o
urs
K
-
Nea
r
est
Nei
g
h
b
o
u
r
s
is
o
n
e
o
f
t
h
e
o
ld
est
n
o
n
-
p
ar
a
m
e
t
r
ic
lear
n
i
n
g
alg
o
r
it
h
m
.
KNN
is
b
ased
on
s
i
m
ilar
it
y
ca
lc
u
latio
n
b
et
wee
n
ex
a
m
p
les.
KNN
cla
s
s
i
f
ier
class
i
f
y
an
in
s
ta
n
ce
b
y
u
s
i
n
g
th
e
cla
s
s
es
o
f
t
h
e
k
clo
s
est
tr
ain
i
n
g
in
s
ta
n
ce
s
in
f
e
atu
r
e
s
p
ac
e
o
f
th
e
d
atas
et
[1
4
]
.
I
n
th
is
p
ap
er
,
w
e
i
m
p
le
m
e
n
t
t
w
o
t
y
p
e
s
o
f
KNN
class
i
f
ier
b
ased
o
n
th
e
f
o
llo
w
i
n
g
d
is
tan
ce
s
:
-
E
u
clid
ian
d
is
tan
ce
w
i
th
n
u
m
b
er
o
f
n
eig
h
b
o
u
r
s
eq
u
al
s
to
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
E
xtra
cted
fea
tu
r
es b
a
s
ed
mu
lti
-
cla
s
s
cla
s
s
ifica
tio
n
o
f o
r
th
o
d
o
n
tic
ima
g
es
(
Hich
a
m
R
ir
i
)
3563
-
C
o
s
i
n
e
d
is
tan
ce
:
D
(
U
,
V
)
=
1
−
U
⃗
⃗
.
V
⃗
⃗
‖
U
‖
‖
V
⃗
⃗
‖
(
1
6
)
W
ith
U
an
d
V
t
w
o
n
-
d
i
m
en
s
io
n
al
v
ec
to
r
s
.
5.
RE
SU
L
T
AND
DI
SCUS
SI
O
NS
I
n
t
h
is
s
ec
tio
n
,
w
e
ev
al
u
ate
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
p
r
ese
n
te
d
m
et
h
o
d
s
.
A
t
fi
r
s
t,
ex
p
er
i
m
en
tal
s
ett
in
g
s
ar
e
p
r
esen
ted
.
T
h
en
th
ec
lass
i
f
icatio
n
r
esu
lt
s
w
i
ll
b
e
p
r
ese
n
ted
.
An
d
f
in
a
ll
y
,
a
n
ev
al
u
ati
o
n
o
f
all
p
r
esen
ted
m
et
h
o
d
s
w
ill
b
e
s
h
o
w
n
.
Fo
r
ev
alu
a
tio
n
o
f
th
e
clas
s
if
icatio
n
r
es
u
lts
,
d
i
f
f
er
en
t
m
ea
s
u
r
es
ar
e
u
s
ed
,
s
u
ch
a
s
A
cc
u
r
ac
y
,
R
ec
all,
P
r
ec
is
io
n
,
F1
-
m
ea
s
u
r
e,
Se
n
s
it
iv
i
t
y
,
S
p
ec
if
icit
y
,
an
d
C
o
h
e
n
’
s
Ka
p
p
a.
A
cc
u
r
ac
y
is
t
h
e
n
u
m
b
er
o
f
co
r
r
ec
tl
y
class
i
fi
ed
i
m
a
g
es
r
elativ
e
to
th
e
to
tal
n
u
m
b
er
o
f
class
i
fi
ed
i
m
a
g
es.
R
ec
all(
ak
a
Sen
s
iti
v
it
y
)
is
th
e
r
ate
o
f
r
ea
l
p
o
s
i
tiv
e
ca
s
e
s
t
h
a
t
ar
e
co
r
r
ec
tl
y
p
r
ed
icted
p
o
s
itiv
e.
C
o
n
tr
ar
i
w
i
s
e,
p
r
ec
is
io
n
is
t
h
e
r
ate
o
f
p
r
ed
icted
p
o
s
itiv
e
ca
s
es
t
h
at
ar
e
co
r
r
ec
tl
y
r
ea
l
p
o
s
itiv
e
s
.
F1
-
m
ea
s
u
r
e,
is
th
e
w
eig
h
ted
h
ar
m
o
n
ic
m
ea
n
o
f
P
r
ec
is
io
n
an
d
R
ec
al
l.
I
tco
m
p
u
tes
th
e
q
u
alit
y
o
f
cla
s
s
i
f
icatio
n
.
T
h
es
e
m
ea
s
u
r
e
s
ca
n
b
e
ex
p
r
ess
ed
as f
o
l
lo
w
s
:
=
=
+
(
1
7
)
=
+
(
1
8
)
1
−
=
2
∗
∗
+
(
1
9
)
=
+
(
2
0
)
W
h
er
e
T
P
is
th
e
to
tal
n
u
m
b
er
o
f
p
o
s
itiv
e
i
m
a
g
es,
FN
is
t
h
e
to
tal
n
u
m
b
er
o
f
f
al
s
e
n
e
g
ati
v
e
i
m
ag
e
s
an
d
FP
is
th
e
to
tal
n
u
m
b
er
o
f
f
al
s
e
p
o
s
itiv
e
i
m
a
g
es.
C
o
h
e
n
’
s
Kap
p
a
is
co
m
p
u
ted
f
r
o
m
a
co
n
f
u
s
io
n
m
atr
ix
as f
o
llo
w
s
[2
8
]
:
=
∑
ℎ
−
∑
=
1
=
1
2
−
∑
=
1
(
2
1
)
W
h
er
e
h
ii
is
th
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
itiv
es
f
o
r
ea
ch
class
in
t
h
e
d
iag
o
n
alo
f
th
e
co
n
f
u
s
io
n
m
atr
ix
,
n
is
th
e
n
u
m
b
er
o
f
ex
a
m
p
le
s
,
m
i
s
th
e
n
u
m
b
er
o
f
clas
s
es
a
n
d
T
r
i
an
d
T
ci
ar
e
th
e
r
o
w
s
’
an
d
co
lu
m
n
s
’
to
tal
co
u
n
t
s
,
r
esp
ec
tiv
el
y
.
5
.
1
.
X
peri
m
ent
a
l set
t
ing
Ou
r
p
r
o
p
o
s
ed
m
et
h
o
d
is
i
m
p
le
m
e
n
ted
u
s
in
g
C
p
r
o
g
r
am
i
n
g
la
n
g
u
a
g
e
a
n
d
L
ib
j
p
eg
lib
r
ar
y
to
d
ec
o
m
p
r
ess
j
p
eg
i
m
ag
e
s
a
n
d
w
e
u
s
e
t
h
e
J
P
E
G/YC
b
C
r
co
l
o
u
r
s
p
ac
e
f
o
r
s
k
in
d
etec
tio
n
.
F
o
r
th
e
al
g
o
r
ith
m
1
,
w
e
s
e
tα
1
as
6
0
to
ex
a
m
in
e
t
h
e
m
ea
n
s
o
f
i
n
te
n
s
itie
s
i
n
th
e
f
o
u
r
co
r
n
er
s
o
f
t
h
e
i
m
a
g
e
w
h
en
th
e
y
ar
e
lo
w
er
th
a
n
α
1
,
an
d
w
e
s
et
α
2
as
5
0
to
ex
a
m
in
e
th
e
d
i
f
f
er
en
ce
b
et
w
ee
n
t
h
e
le
f
t
a
n
d
th
e
r
i
g
h
t
q
u
ar
t
er
o
f
th
e
i
m
a
g
e
t
o
d
if
f
er
e
n
tiate
b
et
w
ee
n
th
e
le
f
t
an
d
r
ig
h
t
m
o
u
ld
class
e
s
.
Fo
r
th
e
alg
o
r
it
h
m
2
,
w
e
s
et
α
3
to
3
5
b
ec
au
s
e
o
f
th
e
d
en
tal
r
etr
ac
to
r
is
p
r
ese
n
t
i
n
th
e
lef
t
s
id
e
f
o
r
t
h
e
i
n
tr
a
r
ig
h
t
clas
s
an
d
it
is
i
n
t
h
e
r
ig
h
t
s
i
d
e
f
o
r
th
e
in
tr
a
lef
t
clas
s
w
h
ile
f
o
r
i
n
tr
a
f
r
o
n
t
clas
s
it
is
p
r
esen
t
i
n
t
h
e
b
o
th
s
id
e
.
W
e
u
s
e
α
3
to
lo
ca
te
t
h
e
r
etr
ac
to
r
’
s
p
o
s
itio
n
b
y
co
m
p
u
ti
n
g
th
e
m
ea
n
o
f
p
i
x
el
’
s
in
te
n
s
ities
o
f
C
b
co
m
p
o
n
en
t
in
t
h
e
b
o
th
s
id
e
o
f
t
h
e
i
m
ag
e
.
Fo
r
alg
o
r
ith
m
3
,
w
e
s
et
α
4
as 1
.
2
5
w
h
ic
h
m
ea
n
s
th
at
in
t
h
e
le
f
t s
id
e
o
f
th
e
i
m
ag
e,
th
e
p
i
x
els co
r
r
esp
o
n
d
in
g
t
o
s
k
i
n
co
lo
u
r
ar
e
i
n
ab
u
n
d
an
ce
co
m
p
ar
ed
to
r
ig
h
t
s
id
e.
A
n
d
w
e
s
et
α
5
as0
.
7
5
w
h
ich
co
r
r
esp
o
n
d
to
th
e
o
p
p
o
s
ite
ca
s
e
o
f
α
4
.
Fo
r
alg
o
r
ith
m
4
,
w
e
s
et
θ
as
1
0
0
b
ec
au
s
e
o
f
in
r
ea
l
f
o
r
m
o
u
l
d
im
a
g
e
s
N
s
h
o
u
ld
b
e
eq
u
al
ze
r
o
b
u
t
s
k
in
co
lo
u
r
d
etec
tio
n
m
et
h
o
d
h
a
s
it
s
in
ce
r
titu
d
e
[2
6
,
2
9
]
,
th
at
w
h
y
w
e
u
s
e
θ
to
en
g
lo
b
e
th
e
f
al
s
e
d
etec
tio
n
er
r
o
r
.
W
e
s
et
θ1
as
6
0
0
to
v
er
if
y
i
f
p
ix
el
s
th
at
co
r
r
esp
o
n
d
to
s
k
in
co
lo
u
r
ar
e
in
ab
u
n
d
an
ce
o
r
n
o
t
f
o
r
in
tr
a
-
o
r
al
an
d
ex
tr
a
-
o
r
al
i
m
ag
e
s
.
5
.
2
.
Cla
s
s
if
ica
t
io
n r
esu
lt
s
I
n
th
i
s
s
u
b
s
ec
tio
n
,
w
e
m
er
g
e
th
e
th
r
ee
s
a
m
p
les
o
f
i
m
a
g
e
s
(
m
o
u
ld
,
in
tr
a
-
o
r
al,
an
d
ex
tr
a
-
o
r
al)
in
o
n
e
s
a
m
p
le
to
e
x
tr
ac
t
f
ea
t
u
r
es
b
y
u
s
i
n
g
t
h
e
L
B
P
alg
o
r
ith
m
.
Af
t
er
th
at,
t
h
e
p
r
in
cip
al
co
m
p
o
n
en
t
a
n
al
y
s
is
(
P
C
A
)
alg
o
r
ith
m
w
as
ap
p
lied
in
th
is
p
ap
er
to
o
p
tim
ize
th
e
r
ed
u
n
d
an
t
p
ar
a
m
eter
s
o
b
tain
ed
w
it
h
th
e
L
B
P
.
T
h
en
w
e
tr
ain
t
h
eg
e
n
er
ated
d
ataset
w
it
h
t
h
e
S
VM
,
KNN
an
d
L
D
A
class
i
f
ier
s
a
s
s
h
o
w
n
o
n
Fi
g
u
r
e
3
.
So
to
ev
al
u
ate
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
A
u
g
u
s
t 2
0
2
0
:
3
5
5
8
-
3567
3564
th
e
p
er
f
o
r
m
a
n
ce
o
f
cla
s
s
if
ica
t
io
n
,
w
e
co
n
s
tr
u
ct
a
co
n
f
u
s
io
n
m
atr
i
x
o
f
t
h
e
s
i
x
tee
n
clas
s
o
f
o
r
th
o
d
o
n
tic
i
m
a
g
es
f
o
r
ea
ch
cla
s
s
i
f
ier
.
Fi
g
u
r
e
4
s
h
o
w
s
th
e
co
n
f
u
s
io
n
m
atr
ice
s
o
f
th
e
e
x
p
er
i
m
e
n
t
r
es
u
lt
s
f
o
r
KN
N,
L
D
A
,
a
n
d
S
VM
class
i
f
ier
s
.
I
n
t
h
e
co
n
f
u
s
io
n
m
atr
ix
,
th
e
r
o
w
s
r
ep
r
esen
t
t
h
e
m
o
d
el
s
f
o
r
ea
ch
i
m
a
g
e
class
w
h
ile
t
h
e
co
lu
m
n
s
r
ep
r
esen
t
th
e
g
r
o
u
n
d
tr
u
t
h
class
es
o
f
i
m
a
g
es.
A
f
ter
w
e
ap
p
ly
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
d
esc
r
ib
ed
in
s
ec
tio
n
3
.
4
.
W
e
co
n
s
tr
u
ct
a
co
n
f
u
s
io
n
m
atr
ix
o
f
s
i
x
tee
n
class
o
f
o
r
th
o
d
o
n
tic
i
m
ag
e
s
to
ev
alu
a
te
th
e
p
er
f
o
r
m
an
ce
o
f
class
i
f
icatio
n
.
T
ab
le
2
s
h
o
w
s
a
co
n
f
u
s
io
n
m
atr
i
x
b
et
w
ee
n
t
h
e
s
ix
tee
n
clas
s
es o
f
o
r
th
o
d
o
n
tic
v
ie
w
s
.
Fig
u
r
e
3
.
C
lass
if
ica
tio
n
tec
h
n
i
q
u
es
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
m
atr
ices
o
f
KNN,
L
D
A
,
an
d
SVM
cla
s
s
if
ier
s
: (
a)
C
o
s
i
n
e
KNN,
(
b
)
E
u
clid
ian
K
NN,
(
c)
L
D
A
,
(
d
)
C
u
b
ic
SVM,
(
e)
R
B
F SVM,
(
f
)
Q
u
ad
r
atic
SV
M
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
E
xtra
cted
fea
tu
r
es b
a
s
ed
mu
lti
-
cla
s
s
cla
s
s
ifica
tio
n
o
f o
r
th
o
d
o
n
tic
ima
g
es
(
Hich
a
m
R
ir
i
)
3565
T
ab
le
2
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
r
ep
r
esen
ti
n
g
th
e
p
r
ed
icted
cla
s
s
ac
co
r
d
in
g
t
o
th
e
tr
u
e
clas
s
c
l
a
ss
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
88
2
2
3
0
0
3
0
0
0
0
0
0
0
0
0
2
1
26
1
2
0
0
0
0
0
0
0
0
0
0
0
0
3
1
0
7
0
0
0
0
0
0
0
0
0
0
0
0
0
4
1
3
2
74
0
0
1
0
0
0
0
0
0
0
0
0
5
0
0
0
0
1
0
3
0
0
0
0
0
0
0
0
0
0
0
6
0
0
0
0
0
5
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0
3
2
2
1
7
3
3
4
2
0
0
0
0
0
8
0
0
0
0
0
0
2
1
4
6
1
2
4
0
0
0
0
0
9
0
0
0
0
0
0
1
0
35
2
0
0
0
0
0
0
10
0
0
0
0
1
1
2
1
4
2
3
4
2
0
0
0
0
0
11
0
0
0
0
0
0
2
5
0
0
1
3
3
0
0
0
0
0
12
0
0
0
0
0
0
0
0
0
0
0
14
0
0
0
0
13
0
0
0
0
0
0
0
0
0
0
0
0
14
0
0
0
14
0
0
0
0
0
0
0
0
0
0
0
0
0
14
0
0
15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
14
0
16
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
14
5
.
3
.
E
v
a
lua
t
i
o
n o
f
cla
s
s
if
ica
t
io
n
A
cc
o
r
d
in
g
to
T
a
b
le
2
,
w
e
s
u
c
ce
ed
ed
to
class
if
y
th
e
i
m
a
g
es
w
it
h
an
av
er
a
g
e
r
a
te
v
alu
e
o
f
94.
28
%.
T
h
er
ef
o
r
e,
w
e
co
m
p
u
te
t
h
e
ac
u
u
r
ac
y
,
k
ap
p
a,
s
en
s
i
ti
v
it
y
,
s
p
ec
if
ic
it
y
,
a
n
d
F1
-
m
esu
r
e
to
ev
al
u
ate
th
e
clas
s
i
f
icatio
n
m
et
h
o
d
s
p
r
esen
ted
i
n
t
h
is
p
ap
er
.
T
ab
le
3
s
h
o
w
s
p
er
f
o
r
m
an
ce
m
ea
s
u
r
e
s
o
f
class
if
ica
tio
n
o
f
all
p
r
esen
ted
m
et
h
o
d
s
.
O
v
er
al
l
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
h
as
led
to
th
e
b
est
s
co
r
es
o
f
p
er
f
o
r
m
an
ce
m
ea
s
u
r
e
s
u
s
ed
in
th
i
s
p
ap
er
.
Desp
ite
th
e
E
u
clid
ian
KNN
h
a
s
led
to
th
e
w
o
r
s
t
r
esu
lt
o
f
cla
s
s
i
f
icatio
n
,
it
h
a
s
led
to
th
e
h
ieg
es
t
v
alu
e
o
f
s
p
ec
i
f
it
y
.
T
ab
le
3
.
C
o
m
p
ar
is
o
n
o
f
class
i
f
icatio
n
p
er
f
o
r
m
an
ce
o
f
p
r
esen
ted
ap
p
r
o
ac
h
es
M
e
t
h
o
d
A
c
c
u
r
a
c
y
(
%)
K
a
p
p
a
(
%)
S
e
n
si
t
i
v
i
t
y
(
%)
S
p
e
c
i
f
i
c
i
t
y
(
%)
F1
-
me
su
r
e
(
%)
R
B
F
S
V
M
5
7
.
5
5
0
.
1
0
6
0
.
0
5
9
6
.
7
8
7
8
.
4
5
C
u
b
i
c
S
V
M
8
0
.
3
7
7
.
0
8
7
1
.
1
1
9
8
.
5
5
8
2
.
0
6
Q
u
a
d
r
a
t
i
c
S
V
M
8
1
.
9
7
8
.
9
3
7
0
.
6
4
9
8
.
6
7
8
1
.
3
9
Eu
c
l
i
d
i
a
n
K
N
N
2
0
.
2
1
8
.
5
1
4
6
.
9
8
9
9
.
6
5
5
5
.
4
6
C
o
si
n
e
K
N
N
8
1
.
9
7
9
.
0
6
7
0
.
7
0
9
8
.
7
8
0
.
7
2
L
D
A
8
4
.
5
8
2
.
1
5
7
5
.
5
9
9
8
.
9
3
8
2
.
3
6
P
r
o
p
o
se
d
9
4
.
2
8
9
3
.
4
4
9
5
.
3
5
9
9
.
6
0
9
2
.
6
6.
CO
NCLU
SI
O
N
I
n
th
is
p
ap
er
,
w
e
p
r
o
p
o
s
ed
an
au
to
m
a
tic
ap
p
r
o
ac
h
to
class
i
f
y
o
r
th
o
d
o
n
tic
i
m
a
g
es
(
m
o
u
ld
,
in
tr
a
-
o
r
al
an
d
ex
tr
a
-
o
r
al)
w
it
h
o
u
t
i
n
ter
v
en
t
io
n
w
it
h
o
p
er
ato
r
s
b
y
p
r
esen
ti
n
g
a
n
al
g
o
r
ith
m
f
o
r
e
ac
h
m
ai
n
cla
s
s
o
f
o
r
th
o
d
o
n
tic
i
m
a
g
es
b
ased
o
n
a
d
ec
is
io
n
tr
ee
s
tr
u
ctu
r
e
b
y
ex
tr
ac
ti
n
g
a
n
d
s
elec
ti
n
g
f
ea
tu
r
es
to
class
i
f
y
t
h
es
e
i
m
a
g
es.
T
h
e
n
,
to
co
m
p
ar
e
o
u
r
r
esu
lt,
w
e
i
m
p
le
m
e
n
t
s
ix
al
g
o
r
ith
m
s
o
f
m
ac
h
i
n
e
lear
n
i
n
g
(
Qu
ad
r
atic
SVM,
C
u
b
ic
SVM,
R
B
F
SV
M,
C
o
s
i
n
e
KNN,
E
u
clid
ia
n
KNN,
an
d
L
D
A
)
b
ased
o
n
tex
tu
r
al
f
ea
tu
r
es
ex
tr
ac
tio
n
u
s
i
n
g
L
B
P
d
escr
ip
to
r
an
d
th
e
P
C
A
al
g
o
r
ith
m
to
r
ed
u
ce
th
e
r
ed
u
n
d
an
t
p
ar
a
m
eter
s
in
th
e
ex
tr
ac
ted
f
ea
tu
r
e
s
.
W
h
ich
d
e
m
o
n
s
tr
ate
a
n
ex
ce
lle
n
t
p
er
f
o
r
m
a
n
ce
o
f
clas
s
i
f
icati
o
n
w
it
h
h
ig
h
ac
c
u
r
ac
y
an
d
K
ap
p
a
v
alu
es
as
w
e
ll
as
th
e
e
x
p
er
i
m
e
n
tal
r
es
u
lt
o
f
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ex
ce
p
t
th
e
E
u
cl
id
i
an
KN
N
cl
ass
i
f
ier
th
a
t
g
i
v
e
s
a
w
o
r
s
t
r
esu
lt
s
.
I
n
f
u
tu
r
e
w
o
r
k
,
w
e
w
ill
i
m
p
r
o
v
e
th
i
s
ap
p
r
o
ac
h
to
lead
th
e
d
ev
elo
p
m
e
n
t
o
f
a
f
r
a
m
e
w
o
r
k
th
a
t
ca
n
p
r
o
v
id
e
t
h
e
e
v
o
lu
tio
n
o
f
t
h
e
p
atie
n
t
’
s
co
n
d
itio
n
u
s
in
g
i
m
ag
e
an
a
l
y
s
is
o
f
p
r
ev
io
u
s
v
i
s
i
ts
i
n
t
h
e
s
a
m
e
clas
s
,
esp
ec
ia
ll
y
f
o
r
th
e
ca
s
e
o
f
g
i
v
i
n
g
a
p
atien
t
h
is
b
est
s
m
ile.
AP
P
E
NDI
X
A
l
g
o
r
i
t
h
m
1
:
c
l
a
ssi
f
i
c
a
t
i
o
n
m
e
t
h
o
d
o
f
m
o
u
l
d
c
l
a
sses
In
p
u
t
:
i
m
a
g
e
I i
n
g
r
a
y
sca
l
e
c
o
l
o
u
r
sp
a
c
e
Ou
t
p
u
t
:
d
e
c
i
si
o
n
d
B
e
g
i
n
C
a
l
c
u
l
a
t
e
µ
up
,
µ
lo
,µ
lu
, µ
ru
, µ
ld
, µ
rd
, µ
lq
, µ
rq
, µ
l
,
a
n
d
µ
r
C
a
l
c
u
l
a
t
e
=
a
n
d
′
=
I
f
(
<
1
&
mi
n
(
µ
lu
, µ
ru
, µ
ld
, µ
rd
)
>
α1
&
(
µ
rq
-
µ
lq
)
>
α2
)
t
h
e
n
d
=
”
mo
u
l
d
l
e
f
t
”
El
se
i
f
(
>
=
1
&
mi
n
(
µ
lu
, µ
ru
, µ
ld
, µ
rd
)
>
α1
&
(
µ
lq
-
µ
rq
)
>
α2
)
T
h
e
n
d
=
”
mo
u
l
d
r
i
g
h
t
”
El
se
i
f
(
′
<
1
&
max
(
µ
lu
, µ
ru
)
<
α1
)
T
h
e
n
d
=
”
mo
u
l
d
max
i
l
l
a
r
y
”
El
se
i
f
(
′
>
=
1
&
max
(
µ
ld
, µ
rd
)
<
α1
)
T
h
e
n
d
=
”
mo
u
l
d
man
d
i
b
u
l
a
r
”
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
A
u
g
u
s
t 2
0
2
0
:
3
5
5
8
-
3567
3566
El
se
d
=
”
mo
u
l
d
f
r
o
n
t
”
En
d
i
f
E
n
d
A
l
g
o
r
i
t
h
m
2
:
C
l
a
ssi
f
i
c
a
t
i
o
n
m
e
t
h
o
d
o
f
i
n
t
r
a
-
o
r
a
l
c
l
a
sses
In
p
u
t
:
i
m
a
g
e
I i
n
R
GB
c
o
l
o
u
r
sp
a
c
e
Ou
t
p
u
t
:
d
e
c
i
si
o
n
d
B
e
g
i
n
T
r
a
n
sf
o
r
m I
t
o
I
gr
a
y
i
n
g
r
a
y
sca
l
e
c
o
l
o
u
r
sp
a
c
e
S
e
g
me
n
t
t
h
e
i
mag
e
I
gr
a
y
b
y
t
h
r
e
sh
o
l
d
i
n
g
D
e
t
e
c
t
t
h
e
e
x
t
e
r
i
o
r
e
d
g
e
E
x
I
f
(
(
)
Eq
.
1
2
i
n
c
r
e
a
si
n
g
)
t
h
e
n
d
=
”
man
d
i
b
u
l
a
r
”
El
se
(
(
)
Eq
.
1
2
d
e
c
r
e
a
si
n
g
)
t
h
e
n
d
=
”
max
i
l
l
a
r
y
”
El
se
o
T
r
a
n
sf
o
r
m I
t
o
I
Y
Cb
Cr
i
n
Y
C
b
C
r
c
o
l
o
u
r
sp
a
c
e
Eq
.
1
o
C
a
l
c
u
l
a
t
e
µ
=
∗
∑
(
,
)
≤
<
≤
<
/
a
n
d
µ
=
∗
∑
(
,
)
≤
<
∗
/
≤
<
o
f
C
b
c
o
mp
o
n
e
n
t
o
I
f
(
µ
l
<
α3
a
n
d
µ
2
>
α3
)
t
h
e
n
d
=
”
r
i
g
h
t
”
o
El
se
i
f
(
µ
l
>
α3
a
n
d
µ
2
<
α3
)
t
h
e
n
d
=
”
l
e
f
t
”
o
El
se
i
f
(
µ
l
>
α3
a
n
d
µ
2
>
α3
)
t
h
e
n
d
=
”
f
r
o
n
t
”
o
En
d
i
f
En
d
i
f
E
n
d
A
l
g
o
r
i
t
h
m
3
:
c
l
a
ssi
f
i
c
a
t
i
o
n
m
e
t
h
o
d
o
f
e
x
t
r
a
-
o
r
a
l
c
l
a
sses
In
p
u
t
:
i
m
a
g
e
I i
n
R
GB
c
o
l
o
u
r
sp
a
c
e
Ou
t
p
u
t
:
d
e
c
i
si
o
n
d
B
e
g
i
n
T
r
a
n
sf
o
r
m I
f
r
o
m R
G
B
t
o
Y
C
b
C
r
c
o
l
o
u
r
sp
a
c
e
C
a
l
c
u
l
a
t
e
a
n
d
I
f
(
<
α5
)
T
h
e
n
d
=
”
r
i
g
h
t
”
El
se
i
f
(
>
α4
)
T
h
e
n
d
=
”
l
e
f
t
”
/
/
p
i
x
e
l
s o
f
sk
i
n
c
o
l
o
u
r
i
n
l
e
f
t
El
se
i
f
(
>
α5
)
a
n
d
(
<
α4
)
T
h
e
n
o
L
o
c
a
t
e
l
e
f
t
e
y
e
o
L
o
c
a
t
e
r
i
g
h
t
e
y
e
o
C
a
l
c
u
l
a
t
e
n
u
mb
e
r
o
f
p
i
x
e
l
s o
f
sk
i
n
i
n
l
e
f
t
si
d
e
o
f
t
h
e
l
e
f
t
e
y
e
o
C
a
l
c
u
l
a
t
e
n
u
mb
e
r
o
f
p
i
x
e
l
s o
f
sk
i
n
i
n
r
i
g
h
t
si
d
e
o
f
t
h
e
r
i
g
h
t
e
y
e
s
4
o
I
f
(
α5
<
<
α4
)
T
h
e
n
I
f
e
x
i
st
i
n
g
p
i
x
e
l
s b
e
l
o
n
g
t
o
t
o
o
t
h
,
t
h
e
n
d
=
”
f
a
c
e
w
i
t
h
smil
e
”
El
se
d
=
”
f
a
c
e
”
o
El
se
I
f
e
x
i
st
i
n
g
p
i
x
e
l
s b
e
l
o
n
g
t
o
t
o
o
t
h
,
T
h
e
n
d
=
”
t
h
r
e
e
-
q
u
a
r
t
e
r
w
i
t
h
smil
e
”
El
se
d
=
”
t
h
r
e
e
-
q
u
a
r
t
e
r
”
o
En
d
i
f
E
n
d
A
l
g
o
r
i
t
h
m
4
:
C
l
a
ssi
f
i
c
a
t
i
o
n
m
e
t
h
o
d
o
f
o
r
t
h
o
d
o
n
t
i
c
c
l
a
sses
In
p
u
t
:
i
m
a
g
e
I i
n
R
GB
c
o
l
o
u
r
sp
a
c
e
Ou
t
p
u
t
:
d
e
c
i
si
o
n
d
B
e
g
i
n
T
r
a
n
sf
o
r
m I
t
o
I
Y
Cb
Cr
c
o
l
o
u
r
sp
a
c
e
Ex
t
r
a
c
t
f
e
a
t
u
r
e
s
C
a
l
c
u
l
a
t
e
N
:
n
u
mb
e
r
o
f
p
i
x
e
l
s b
e
l
o
n
g
t
o
sk
i
n
c
o
l
o
u
r
i
n
I
Y
Cb
Cr
I
f
(
N
<
θ)
t
h
e
n
A
l
g
o
r
i
t
h
m 1
El
se
o
C
a
l
c
u
l
a
t
e
N
1
:
n
u
mb
e
r
o
f
p
i
x
e
l
s b
e
l
o
n
g
t
o
sk
i
n
c
o
l
o
u
r
i
n
t
h
e
u
p
p
e
r
si
d
e
o
f
I
Y
Cb
Cr
o
I
f
(
N
1
>
θ1
)
t
h
e
n
A
l
g
o
r
i
t
h
m 2
o
El
se
A
l
g
o
r
i
t
h
m
3
o
En
d
i
f
En
d
i
f
E
n
d
ACK
NO
WL
E
D
G
E
M
E
NT
S
W
e
ar
e
g
r
atef
u
l
to
Fac
u
lt
y
o
f
s
cie
n
ce
s
,
C
h
o
u
aib
Do
u
k
k
al
i
Un
i
v
er
s
it
y
,
an
d
Fac
u
lt
y
o
f
Den
tis
tr
y
,
Un
i
v
er
s
it
y
o
f
Ha
s
s
a
n
I
I
A
i
n
C
h
o
k
,
f
o
r
t
h
e
i
m
m
e
n
s
e
s
u
p
p
o
r
t in
p
ilo
tin
g
th
is
w
o
r
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
E
xtra
cted
fea
tu
r
es b
a
s
ed
mu
lti
-
cla
s
s
cla
s
s
ifica
tio
n
o
f o
r
th
o
d
o
n
tic
ima
g
es
(
Hich
a
m
R
ir
i
)
3567
RE
F
E
R
E
NC
E
S
[1
]
B.
He
u
n
g
,
H.
Ch
a
k
,
J.
Zh
a
n
g
,
A.
Kn
u
d
b
y
,
C.
E.
Bu
lme
r,
a
n
d
M
.
G
.
S
c
h
m
id
t,
“
A
n
o
v
e
rv
i
e
w
a
n
d
c
o
m
p
a
riso
n
o
f
m
a
c
h
in
e
-
lea
rn
in
g
tec
h
n
i
q
u
e
s
f
o
r
c
las
sif
ic
a
ti
o
n
p
u
rp
o
se
s
in
d
ig
it
a
l
so
il
m
a
p
p
in
g
,
”
Ge
o
d
e
r
ma
,
v
o
l.
2
6
5
,
p
p
.
6
2
–
7
7
,
2
0
1
6
.
[2
]
M
.
Zan
g
,
D.
W
e
n
,
K.
W
a
n
g
,
T
.
L
iu
,
a
n
d
W
.
S
o
n
g
,
“
A
n
o
v
e
l
to
p
ic
f
e
a
tu
re
f
o
r
i
m
a
g
e
sc
e
n
e
c
las
si
f
ica
ti
o
n
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
1
4
8
,
p
p
.
4
6
7
–
4
7
6
,
2
0
1
5
.
[3
]
A
.
M
.
As
h
ir
a
n
d
A
.
El
e
y
a
n
,
“
F
a
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
b
a
se
d
o
n
im
a
g
e
p
y
ra
m
id
a
n
d
sin
g
le
-
b
ra
n
c
h
d
e
c
isio
n
tree
,
”
S
ig
n
a
l,
Ima
g
e
Vi
d
e
o
Pro
c
e
ss
.
,
v
o
l.
1
1
,
n
o
.
6
,
p
p
.
1
0
1
7
–
1
0
2
4
,
2
0
1
7
.
[4
]
S
.
M
.
L
a
jev
a
rd
i,
“
S
tru
c
tu
ra
l
sim
il
a
rit
y
c
las
si
f
ier
f
o
r
fa
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
i
o
n
,
”
S
ig
n
a
l,
Im
a
g
e
Vi
d
e
o
Pro
c
e
ss
.
,
v
o
l.
8
,
n
o
.
6
,
p
p
.
1
1
0
3
–
1
1
1
0
,
2
0
1
4
.
[5
]
G
.
G
il
a
n
ie,
U.
I.
Ba
j
wa
,
M
.
M
.
W
a
ra
ich
,
Z.
H
a
b
ib
,
H.
Ull
a
h
,
a
n
d
M
.
Na
sir,
“
Cl
a
ss
i
f
ica
ti
o
n
o
f
n
o
rm
a
l
a
n
d
a
b
n
o
rm
a
l
b
ra
in
M
RI
slice
s u
sin
g
G
a
b
o
r
tex
tu
re
a
n
d
s
u
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
s,”
S
ig
n
a
l,
Im
a
g
e
V
id
e
o
P
ro
c
e
ss
.
,
2
0
1
7
.
[6
]
N.
M
.
Kh
a
n
,
R.
Ks
a
n
ti
n
i,
I.
S
.
A
h
m
a
d
,
a
n
d
L
.
G
u
a
n
,
“
S
N
-
S
V
M
:
a
sp
a
rse
n
o
n
p
a
ra
m
e
tri
c
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
c
las
si
f
ier
,
”
S
ig
n
a
l
,
Ima
g
e
Vi
d
e
o
Pro
c
e
ss
.
,
v
o
l.
8
,
n
o
.
8
,
p
p
.
1
6
2
5
–
1
6
3
7
,
2
0
1
2
.
[7
]
S
.
Ra
it
h
e
t
a
l.
,
“
A
rti
f
icia
l
Ne
u
ra
l
Ne
tw
o
rk
s
a
s
a
p
o
w
e
r
f
u
l
n
u
m
e
r
ica
l
to
o
l
t
o
c
las
sify
sp
e
c
i
f
ic
fe
a
t
u
re
s
o
f
a
to
o
t
h
b
a
se
d
o
n
3
D sc
a
n
d
a
ta,”
Co
mp
u
t.
Bi
o
l.
M
e
d
.
,
v
o
l.
8
0
,
p
p
.
6
5
–
7
6
,
2
0
1
7
.
[8
]
M
.
Ka
w
u
lo
k
,
J.
Ka
w
u
lo
k
,
a
n
d
J.
Na
lep
a
,
“
S
p
a
ti
a
l
-
b
a
se
d
sk
in
d
e
tec
ti
o
n
u
sin
g
d
isc
rim
in
a
ti
v
e
s
k
in
-
p
re
se
n
c
e
fe
a
tu
re
s,”
Pa
tt
e
rn
Rec
o
g
n
it
.
L
e
tt
.
,
v
o
l
.
4
1
,
n
o
.
1
,
p
p
.
3
–
1
3
,
2
0
1
4
.
[9
]
L
.
L
iu
,
P
.
F
ieg
u
th
,
Y.
G
u
o
,
X.
W
a
n
g
,
a
n
d
M
.
P
ietik
ä
in
e
n
,
“
L
o
c
a
l
b
i
n
a
ry
f
e
a
tu
re
s
f
o
r
tex
tu
re
c
las
sif
ic
a
ti
o
n
:
T
a
x
o
n
o
m
y
a
n
d
e
x
p
e
ri
m
e
n
tal
stu
d
y
,
”
Pa
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l.
6
2
,
p
p
.
1
3
5
–
1
6
0
,
2
0
1
7
.
[1
0
]
S
.
Uy
u
n
a
n
d
L
.
Ch
o
rid
a
h
,
“
F
e
a
tu
r
e
S
e
lec
ti
o
n
M
a
m
m
o
g
r
a
m
b
a
se
d
o
n
Bre
a
st Can
c
e
r
M
in
in
g
,
”
In
t.
J
.
El
e
c
tr.
Co
mp
u
t
.
En
g
.
,
v
o
l.
8
,
n
o
.
1
,
p
p
.
6
0
–
6
9
,
2
0
1
8
.
[1
1
]
A
.
K.
Nik
h
a
th
a
n
d
K.
S
u
b
ra
h
m
a
n
y
a
m
,
“
F
e
a
tu
re
se
le
c
ti
o
n
,
o
p
ti
m
iza
ti
o
n
a
n
d
c
l
u
ste
rin
g
stra
teg
ies
o
f
te
x
t
d
o
c
u
m
e
n
ts,”
In
t.
J
.
El
e
c
tr.
Co
mp
u
t.
En
g
.
,
v
o
l.
9
,
n
o
.
2
,
p
.
1
3
1
3
,
2
0
1
9
.
[1
2
]
H.
Ka
rim
,
S
.
R.
Nia
k
a
n
,
a
n
d
R.
S
a
f
d
a
ri,
“
Co
m
p
a
riso
n
o
f
n
e
u
ra
l
n
e
tw
o
rk
train
in
g
a
lg
o
rit
h
m
s
f
o
r
c
las
si
f
ica
ti
o
n
o
f
h
e
a
rt
d
ise
a
se
s
,
”
IAE
S
In
t.
J
.
Arti
f
.
In
tell
.
,
v
o
l.
7
,
n
o
.
4
,
p
p
.
1
8
5
–
1
8
9
,
2
0
1
8
.
[1
3
]
K.
L
i,
F
.
W
a
n
g
,
a
n
d
L
.
Zh
a
n
g
,
“
A
n
e
w
a
lg
o
rit
h
m
f
o
r
i
m
a
g
e
re
c
o
g
n
it
io
n
a
n
d
c
las
sif
ica
ti
o
n
b
a
se
d
o
n
im
p
ro
v
e
d
Ba
g
o
f
F
e
a
tu
re
s a
lg
o
rit
h
m
,
”
Op
t.
-
In
t.
J
.
L
ig
h
t
El
e
c
tro
n
Op
t
.
,
v
o
l
.
1
2
7
,
n
o
.
1
1
,
p
p
.
4
7
3
6
–
4
7
4
0
,
2
0
1
6
.
[1
4
]
C.
Aff
o
n
so
,
A
.
L
.
D.
Ro
ss
i,
F
.
H.
A
.
V
ieira
,
a
n
d
A
.
C.
P
.
d
e
L
.
F
.
d
e
Ca
rv
a
lh
o
,
“
De
e
p
lea
rn
in
g
f
o
r
b
io
l
o
g
ica
l
ima
g
e
c
las
si
f
ica
ti
o
n
,
”
Exp
e
rt S
y
st.
Ap
p
l.
,
v
o
l.
8
5
,
p
p
.
1
1
4
–
1
2
2
,
2
0
1
7
.
[1
5
]
A
.
Da
h
m
a
n
e
,
S
.
L
a
ra
b
i,
I.
M
.
Bil
a
sc
o
,
a
n
d
C
.
Dje
ra
b
a
,
“
He
a
d
p
o
s
e
e
stim
a
ti
o
n
b
a
se
d
o
n
f
a
c
e
s
y
m
m
e
tr
y
a
n
a
l
y
sis,”
S
ig
n
a
l
,
Ima
g
e
Vi
d
e
o
Pro
c
e
ss
.
,
v
o
l
.
9
,
n
o
.
8
,
p
p
.
1
8
7
1
–
1
8
8
0
,
2
0
1
5
.
[1
6
]
M
.
H.
Ya
n
g
,
D.
J.
Krie
g
m
a
n
,
a
n
d
N.
A
h
u
ja,
“
De
tec
ti
n
g
fa
c
e
s
in
im
a
g
e
s:
A
su
rv
e
y
,
”
IEE
E
T
ra
n
s.
Pa
tt
e
rn
A
n
a
l
.
M
a
c
h
.
In
tell
.
,
v
o
l
.
2
4
,
n
o
.
1
,
p
p
.
3
4
–
5
8
,
2
0
0
2
.
[1
7
]
G
.
Ya
n
g
a
n
d
T
.
S
.
Hu
a
n
g
,
“
Hu
m
a
n
f
a
c
e
d
e
tec
ti
o
n
in
a
c
o
m
p
lex
b
a
c
k
g
ro
u
n
d
,
”
Pa
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l.
2
7
,
n
o
.
1
,
p
p
.
5
3
–
6
3
,
1
9
9
4
.
[1
8
]
K.
Yo
w
,
“
F
e
a
tu
re
-
b
a
se
d
h
u
m
a
n
fa
c
e
d
e
tec
ti
o
n
,
”
Ima
g
e
Vi
s.
C
o
mp
u
t
.
,
v
o
l.
1
5
,
n
o
.
9
,
p
p
.
7
1
3
–
7
3
5
,
1
9
9
7
.
[1
9
]
I.
Cra
w
,
D.
T
o
c
k
,
a
n
d
A
.
Be
n
n
e
tt
,
“
F
in
d
in
g
F
a
c
e
F
e
a
tu
re
s,”
Co
mp
u
t.
Vi
sio
n
—
ECC
V’9
2
,
p
p
.
1
–
2
3
,
1
9
9
2
.
[2
0
]
H.
a
Ro
w
le
y
,
S
.
Ba
lu
ja,
a
n
d
T
.
K
a
n
a
d
e
,
“
Ne
u
ra
l
n
e
tw
o
rk
-
b
a
se
d
f
a
c
e
d
e
tec
ti
o
n
,
”
IEE
E
T
r
a
n
s.
P
a
tt
e
rn
An
a
l.
M
a
c
h
.
In
tell.
,
v
o
l.
2
0
,
n
o
.
1
,
p
p
.
2
3
–
3
8
,
1
9
9
8
.
[2
1
]
H.
Riri
,
A
.
El
m
o
u
tao
u
a
k
k
il
,
A
.
Be
n
i
-
Hs
sa
n
e
,
a
n
d
F
.
B
o
u
rz
g
u
i,
“
Clas
sif
ic
a
ti
o
n
a
n
d
Re
c
o
g
n
it
io
n
o
f
De
n
tal
I
m
a
g
e
s
Us
in
g
a
D
e
c
isio
n
a
l
T
re
e
,
”
in
In
te
rn
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Co
mp
u
t
e
r
Gr
a
p
h
ics
,
Ima
g
in
g
a
n
d
Vi
su
a
l
iza
ti
o
n
(
C
GiV),
p
p
.
3
9
0
–
3
9
3
,
2
0
1
6
.
[2
2
]
P
.
Ka
k
u
m
a
n
u
,
S
.
M
a
k
ro
g
ian
n
is,
a
n
d
N.
B
o
u
r
b
a
k
is,
“
A
su
rv
e
y
o
f
sk
in
-
c
o
lo
r
m
o
d
e
li
n
g
a
n
d
d
e
te
c
ti
o
n
m
e
th
o
d
s,
”
Pa
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l.
4
0
,
n
o
.
3
,
p
p
.
1
1
0
6
–
1
1
2
2
,
2
0
0
7
.
[2
3
]
K.
Ba
sh
a
S
h
a
ik
a
n
d
Jm
.
M
a
ry
Je
n
it
h
a
,
“
C
o
m
p
a
ra
ti
v
e
S
tu
d
y
o
f
S
k
i
n
Co
l
o
r
De
tec
ti
o
n
a
n
d
S
e
g
m
e
n
tatio
n
in
HS
V
a
n
d
YCb
Cr
Co
l
o
r
S
p
a
c
e
,
”
Pro
c
e
d
ia
C
o
mp
u
t
.
S
c
i
.
,
v
o
l.
5
7
,
p
p
.
4
1
–
4
8
,
2
0
1
5
.
[2
4
]
D.
Ch
a
i
a
n
d
K.
N.
Ng
a
n
,
“
F
a
c
e
se
g
m
e
n
tatio
n
u
sin
g
sk
in
-
c
o
lo
r
m
a
p
in
v
id
e
o
p
h
o
n
e
a
p
p
l
ica
ti
o
n
s,
”
IEE
E
T
ra
n
s
.
Circ
u
it
s S
y
st.
V
id
e
o
T
e
c
h
n
o
l.
,
v
o
l
.
9
,
n
o
.
4
,
p
p
.
5
5
1
–
5
6
4
,
1
9
9
9
.
[2
5
]
C.
L
in
,
“
F
a
c
e
d
e
tec
ti
o
n
in
c
o
m
p
li
c
a
ted
b
a
c
k
g
ro
u
n
d
s
a
n
d
d
if
f
e
re
n
t
il
lu
m
in
a
ti
o
n
c
o
n
d
it
io
n
s
b
y
u
sin
g
YCb
Cr
c
o
lo
r
sp
a
c
e
a
n
d
n
e
u
ra
l
n
e
tw
o
rk
,
”
Pa
tt
e
rn
Rec
o
g
n
it
.
L
e
tt
.
,
v
o
l.
2
8
,
n
o
.
1
6
,
p
p
.
2
1
9
0
–
2
2
0
0
,
2
0
0
7
.
[2
6
]
J.
M
.
Ch
a
v
e
s
-
G
o
n
z
á
lez
,
M
.
A
.
V
e
g
a
-
Ro
d
ríg
u
e
z
,
J.
A
.
G
ó
m
e
z
-
P
u
li
d
o
,
a
n
d
J.
M
.
S
á
n
c
h
e
z
-
P
é
re
z
,
“
De
tec
ti
n
g
sk
in
in
f
a
c
e
re
c
o
g
n
it
io
n
sy
ste
m
s:
A
c
o
lo
u
r
sp
a
c
e
s stu
d
y
,
”
Dig
it
a
l
S
ig
n
a
l
Pr
o
c
e
ss
in
g
,
v
o
l
.
2
0
,
n
o
.
3
,
p
p
.
8
0
6
–
8
2
3
,
2
0
1
0
.
[2
7
]
D.
Ca
n
to
r
-
r
iv
e
ra
,
A
.
R.
Kh
a
n
,
M
.
G
o
u
b
ra
n
,
S
.
M
.
M
i
rsa
tt
a
ri,
a
n
d
T
.
M
.
P
e
ters
,
“
Co
m
p
u
teriz
e
d
M
e
d
ica
l
Im
a
g
in
g
a
n
d
G
ra
p
h
ics
De
t
e
c
ti
o
n
o
f
te
m
p
o
ra
l
lo
b
e
e
p
il
e
p
sy
u
sin
g
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
s
in
m
u
lt
i
-
p
a
ra
m
e
tr
ic
q
u
a
n
ti
tativ
e
M
R
im
a
g
in
g
,
”
Co
mp
u
t.
M
e
d
.
Ima
g
in
g
Gr
a
p
h
.
,
v
o
l.
4
1
,
p
p
.
1
4
–
2
8
,
2
0
1
5
.
[2
8
]
M
.
G
a
lar,
A
.
F
e
rn
a
n
d
e
z
,
E.
Ba
rre
n
e
c
h
e
a
,
H.
Bu
sti
n
c
e
,
a
n
d
F
.
H
e
rre
ra
,
“
A
n
o
v
e
rv
ie
w
o
f
e
n
se
m
b
le
m
e
th
o
d
s
f
o
r
b
in
a
ry
c
las
si
f
iers
in
m
u
lt
i
-
c
las
s
p
ro
b
lem
s:
Ex
p
e
ri
m
e
n
tal
stu
d
y
o
n
o
n
e
-
vs
-
o
n
e
a
n
d
o
n
e
-
vs
-
a
ll
sc
h
e
m
e
s,”
Pa
tt
e
rn
Rec
o
g
n
it
.,
v
o
l
.
4
4
,
n
o
.
8
,
p
p
.
1
7
6
1
–
1
7
7
6
,
2
0
1
1
.
[2
9
]
R.
Kh
a
n
,
A
.
H
a
n
b
u
ry
,
J.
S
tö
tt
in
g
e
r,
a
n
d
A
.
B
a
is,
“
Co
lo
r
b
a
se
d
sk
i
n
c
las
si
f
ica
ti
o
n
,
”
Pa
tt
e
rn
Rec
o
g
n
i
t.
L
e
tt
.
,
v
o
l.
3
3
,
n
o
.
2
,
p
p
.
1
5
7
–
1
6
3
,
2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.