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
.
6
,
Dec
em
b
er
2
0
2
0
,
p
p
.
5
6
7
8
~
5
6
8
6
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
6
.
pp
5
6
7
8
-
5
6
8
6
5678
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
An ef
ficien
t
m
e
th
o
d t
o
cl
a
ss
ify G
I
t
ra
ct
i
m
a
g
es fro
m
WCE
using
vis
ua
l
w
o
rd
s
R.
P
o
nn
us
a
m
y
1
,
S.
Sa
t
hia
m
o
o
rt
hy
2
,
R.
Vis
a
la
k
s
h
i
3
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter an
d
I
n
f
o
rm
a
ti
o
n
S
c
ien
c
e
,
A
n
n
a
m
a
l
a
i
Un
iv
e
rsit
y
,
In
d
ia
2
Co
n
tr
o
ll
e
r
o
f
Ex
a
m
in
a
ti
o
n
s,
T
a
m
il
Virtu
a
l
A
c
a
d
e
m
y
,
In
d
ia
3
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
A
n
n
a
m
a
lai
Un
iv
e
rsit
y
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
2
0
,
2
0
1
9
R
ev
i
s
ed
Ma
y
9
,
2020
A
cc
ep
ted
Ma
y
2
7
,
2020
T
h
e
d
ig
it
a
l
ima
g
e
s
m
a
d
e
w
it
h
th
e
w
ir
e
les
s
c
a
p
su
le
e
n
d
o
sc
o
p
y
(
W
CE)
f
ro
m
th
e
p
a
ti
e
n
t'
s
g
a
stro
in
tes
ti
n
a
l
trac
t
a
re
u
se
d
to
f
o
re
c
a
st
a
b
n
o
rm
a
li
ti
e
s.
T
h
e
b
ig
a
m
o
u
n
t
o
f
in
f
o
rm
a
ti
o
n
f
ro
m
W
C
E
p
ictu
re
s
c
o
u
l
d
tak
e
2
h
o
u
rs
to
re
v
ie
w
GI
trac
t
il
ln
e
ss
e
s
p
e
r
p
a
ti
e
n
t
t
o
r
e
se
a
rc
h
th
e
d
ig
e
stiv
e
s
y
ste
m
a
n
d
e
v
a
lu
a
te
th
e
m
.
It
is
h
ig
h
ly
ti
m
e
c
o
n
su
m
in
g
a
n
d
i
n
c
re
a
se
s
h
e
a
lt
h
c
a
re
c
o
sts
c
o
n
sid
e
ra
b
ly
.
In
o
rd
e
r
t
o
o
v
e
rc
o
m
e
th
is
p
ro
b
le
m
,
th
e
c
e
n
ter
s
y
m
m
e
tri
c
lo
c
a
l
b
in
a
r
y
p
a
tt
e
rn
(CS
-
L
BP
)
a
n
d
th
e
a
u
to
c
o
l
o
r
c
o
rre
lo
g
ra
m
(
A
CC)
we
re
p
ro
p
o
s
e
d
to
u
se
a
n
o
v
e
l
m
e
th
o
d
b
a
se
d
o
n
a
v
isu
a
l
b
a
g
o
f
fe
a
tu
re
s
(V
BOF).
In
o
rd
e
r
to
so
lv
e
th
is
issu
e
,
w
e
su
g
g
e
ste
d
a
v
is
u
a
l
b
a
g
o
f
f
e
a
tu
re
s
(V
BOF)
m
e
th
o
d
b
y
in
c
o
rp
o
ra
ti
n
g
sc
a
le
in
v
a
rian
t
f
e
a
tu
re
tran
sf
o
rm
(S
IF
T
),
CS
-
L
BP
a
n
d
A
CC
.
T
h
is
c
o
m
b
in
a
ti
o
n
o
f
f
e
a
tu
re
s
is
a
b
le
to
d
e
tec
t
th
e
i
n
tere
st
p
o
in
t
,
te
x
tu
re
a
n
d
c
o
lo
r
in
f
o
rm
a
ti
o
n
in
a
n
im
a
g
e
.
F
e
a
tu
re
s
f
o
r
e
a
c
h
i
m
a
g
e
a
r
e
c
a
l
c
u
late
d
to
c
re
a
te
a
d
e
sc
rip
to
r
w
it
h
a
lar
g
e
d
im
e
n
sio
n
.
T
h
e
p
ro
p
o
se
d
f
e
a
tu
re
d
e
sc
rip
to
rs
a
re
c
lu
ste
re
d
b
y
K
-
m
e
a
n
s
re
f
e
rre
d
to
a
s
v
isu
a
l
w
o
rd
s,
a
n
d
th
e
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
(S
V
M
)
m
e
th
o
d
is
u
se
d
to
a
u
to
m
a
ti
c
a
ll
y
c
las
si
fy
m
u
lt
ip
le
d
ise
a
se
a
b
n
o
rm
a
li
ti
e
s
f
ro
m
th
e
G
I
trac
t.
F
in
a
ll
y
,
p
o
st
-
p
ro
c
e
ss
in
g
sc
h
e
m
e
is
a
p
p
li
e
d
to
d
e
a
l
w
it
h
f
in
a
l
c
las
sif
ica
ti
o
n
re
su
lt
s
i.
e
.
v
a
li
d
a
ted
t
h
e
p
e
rf
o
rm
a
n
c
e
o
f
m
u
lt
i
-
a
b
n
o
rm
a
l
d
ise
a
se
f
ra
m
e
d
e
tec
ti
o
n
.
K
ey
w
o
r
d
s
:
A
u
to
co
lo
r
co
r
r
elo
g
r
am
C
en
ter
s
y
m
m
etr
ic
F
ea
tu
r
e
tr
an
s
f
o
r
m
L
o
ca
l b
in
ar
y
p
atter
n
Su
p
p
o
r
t v
ec
to
r
m
ac
h
i
n
e
W
ir
eless
ca
p
s
u
le
e
n
d
o
s
co
p
y
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
:
R
.
P
o
n
n
u
s
a
m
y
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
an
d
I
n
f
o
r
m
a
tio
n
Scie
n
ce
,
An
n
a
m
ala
i U
n
i
v
er
s
i
t
y
,
An
n
a
m
ala
in
a
g
ar
,
C
h
id
a
m
b
ar
a
m
,
T
am
i
ln
ad
u
,
I
n
d
ia
.
E
m
ail:
p
o
v
i2
0
0
6
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
W
ir
eless
ca
p
s
u
le
e
n
d
o
s
co
p
e
(
W
C
E
)
[
1
]
h
as
co
n
s
id
er
ed
as
d
iag
n
o
s
t
ic
to
o
l
f
o
r
g
as
tr
o
in
tes
tin
al
(
GI
)
tr
ac
t
[
1
]
.
I
t
o
f
f
er
s
a
n
o
n
-
i
n
v
as
iv
e
d
ir
ec
t
o
p
tio
n
i
n
co
n
tr
ast
t
o
co
n
v
e
n
tio
n
a
l
e
n
d
o
s
co
p
e
an
d
e
m
p
o
w
er
s
d
o
cto
r
s
to
in
v
e
s
ti
g
ate
t
h
e
GI
tr
ac
t
w
h
i
ch
is
g
e
n
er
all
y
n
o
t
o
p
en
[
2
]
.
C
o
m
p
ar
ed
w
ith
co
n
v
e
n
tio
n
al
en
d
o
s
co
p
y
s
y
s
te
m
s
,
W
C
E
n
o
t
j
u
s
t
g
ets
f
u
l
l
ac
ce
s
s
to
t
h
e
s
m
a
ll
d
i
g
esti
v
e
s
y
s
te
m
[
3
,
4
]
,
y
et
in
ad
d
itio
n
o
f
f
er
s
th
e
p
atien
ts
a
n
ef
f
o
r
tle
s
s
s
tr
ateg
y
.
W
C
E
is
a
p
ill
s
h
ap
ed
g
ad
g
e
t
co
n
s
i
s
ti
n
g
o
f
a
b
r
ief
co
r
e
len
g
t
h
co
m
p
le
m
en
tar
y
m
etal
o
x
id
e
s
e
m
ico
n
d
u
cto
r
(
C
MO
S)
s
en
s
o
r
co
n
s
is
ti
n
g
o
f
f
o
u
r
lig
h
t
s
o
u
r
ce
s
,
a
b
atter
y
,
r
ad
io
tr
an
s
m
itter
a
n
d
o
th
er
s
m
al
l
-
sc
ale
s
eg
m
e
n
ts
as
ap
p
ea
r
ed
in
F
ig
u
r
e
1
.
T
h
e
co
n
ta
in
er
is
g
u
lp
ed
b
y
p
atie
n
t,
ta
k
i
n
g
p
ict
u
r
es
f
o
r
e
v
er
y
s
ec
o
n
d
(
f
p
s
)
at
a
s
p
ee
d
o
f
t
w
o
i
m
ag
e
s
.
T
h
e
en
d
o
s
co
p
ic
co
n
tai
n
er
tr
av
el
t
h
r
o
u
g
h
th
e
GI
tr
ac
t
,
ca
tch
i
n
g
a
n
d
r
e
m
o
tel
y
tr
a
n
s
m
itt
in
g
i
n
ex
ce
s
s
o
f
5
5
0
0
0
f
r
a
m
es
to
th
e
r
e
co
r
d
er
attac
h
ed
to
th
e
p
atien
t
's
m
id
r
i
f
f
.
W
h
er
ea
s
W
C
E
ca
tch
e
s
co
lo
r
i
m
a
g
es
o
f
t
h
e
GI
tr
ac
t
f
o
r
r
o
u
g
h
l
y
8
h
o
u
r
s
an
d
tr
a
n
s
m
i
t
s
t
h
e
m
r
e
m
o
tel
y
to
an
o
u
t
s
id
e
in
f
o
r
m
atio
n
r
ec
o
r
d
in
g
g
ad
g
et
w
o
r
n
b
y
t
h
e
p
atien
t
ar
o
u
n
d
t
h
e
m
id
s
ec
tio
n
.
T
h
ese
i
m
ag
e
s
ar
e
th
e
n
d
o
w
n
lo
ad
ed
to
a
co
m
p
u
ter
w
o
r
k
s
tatio
n
,
an
d
ass
ess
ed
b
y
cli
n
icia
n
s
to
s
ett
le
o
n
r
esto
r
ativ
e
an
a
l
y
tic
c
h
o
ic
e.
I
n
s
p
ite
o
f
th
e
f
ac
t
t
h
a
t
W
C
E
i
s
a
s
p
ec
ialized
ac
h
iev
e
m
e
n
t,
t
h
e
lo
o
k
i
n
g
i
n
to
o
f
v
id
eo
an
d
e
lu
cid
atio
n
o
f
th
e
en
tire
in
e
x
ac
t
5
0
,
0
0
0
i
m
ag
e
s
f
o
r
ev
er
y
p
atien
t
tak
e
a
r
o
u
n
d
t
w
o
h
o
u
r
s
f
o
r
an
ac
co
m
p
lis
h
ed
clin
ic
ian
[
5
]
.
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:
2088
-
8708
A
n
efficien
t m
eth
o
d
to
cla
s
s
ify
GI
tr
a
ct
ima
g
es fr
o
m
WC
E
u
s
in
g
visu
a
l wo
r
d
s
(
R
.
P
o
n
n
u
s
a
my
)
5679
T
h
e
ab
n
o
r
m
al
i
m
ag
e
s
f
o
r
th
e
m
o
s
t
p
ar
t
p
o
s
s
e
s
s
j
u
s
t
u
n
d
er
5
%
o
f
t
h
e
e
n
tire
o
n
es.
B
esid
es,
s
p
atial
q
u
al
ities
o
f
an
o
m
alo
u
s
i
m
a
g
es,
t
h
e
s
h
ap
e,
s
u
r
f
ac
e,
esti
m
ate,
a
n
d
th
e
d
i
v
er
g
e
f
r
o
m
th
eir
e
n
co
m
p
as
s
i
n
g
c
h
an
g
e
ar
e
h
ar
d
f
o
r
clin
icia
n
s
to
r
ec
o
g
n
ize
v
ar
ia
tio
n
f
r
o
m
t
h
e
i
m
a
g
es
i
n
all
cir
cu
m
s
tan
ce
s
[
6
]
.
A
t
th
is
p
o
in
t,
p
lan
n
i
n
g
a
n
au
to
m
at
ic
co
m
p
u
ter
-
aid
ed
s
y
s
te
m
i
s
es
s
en
t
ial
to
h
e
lp
clin
ic
i
an
s
i
n
v
esti
g
ate
ab
n
o
r
m
al
i
m
a
g
es.
Mo
r
e
w
o
r
k
h
as
as
o
f
n
o
w
b
ee
n
e
m
b
r
ac
ed
o
n
au
to
m
at
ic
ab
n
o
r
m
al
i
m
ag
e
s
d
etec
tio
n
in
W
C
E
v
id
eo
s
.
Sev
er
a
l
w
o
r
k
s
ar
e
u
s
ed
to
ca
r
r
ied
o
u
t to
d
ef
in
e
a
p
ar
ticu
l
ar
d
is
ea
s
e.
Fig
u
r
e
1
.
Diag
r
a
m
f
o
r
co
m
p
o
n
en
ts
o
f
W
C
E
,
1
)
o
p
tical
d
o
m
e,
2
)
f
o
ca
l p
o
in
t h
o
ld
er
,
3
)
f
o
ca
l p
o
in
t,
4
)
lig
h
ti
n
g
u
p
s
o
u
r
c
e,
5
)
C
MO
S p
ictu
r
es,
6
)
b
atter
y
,
7
)
tr
an
s
m
itter
,
8
)
r
ec
ep
tio
n
ap
p
ar
atu
s
T
h
e
m
a
n
u
s
cr
ip
t
is
ar
r
an
g
ed
ac
co
r
d
in
g
to
th
e
f
o
llo
w
i
n
g
.
Sec
tio
n
2
,
w
e
d
is
c
u
s
s
ed
t
h
e
ass
o
ciate
d
w
o
r
k
s
.
T
h
e
m
a
n
u
s
cr
ip
t
h
as
b
e
en
o
r
g
an
ized
in
t
h
is
w
a
y
.
I
n
s
ec
tio
n
3
,
d
ea
ls
ab
o
u
t
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
w
h
er
e
w
e
i
n
tr
o
d
u
ce
d
th
e
SIFT
alg
o
r
ith
m
,
C
S
-
L
B
P
an
d
A
C
C
f
ea
t
u
r
es.
T
h
en
t
h
e
co
m
p
u
tatio
n
o
f
f
ea
tu
r
e
v
ec
to
r
b
y
co
m
b
i
n
i
n
g
th
e
s
e
t
h
r
ee
f
ea
t
u
r
e
s
w
er
e
also
d
is
c
u
s
s
ed
i
n
s
ec
t
io
n
3
.
T
h
e
v
is
u
al
b
ag
o
f
w
o
r
d
s
d
ep
ictio
n
u
s
i
n
g
K
-
m
ea
n
s
i
s
d
is
c
u
s
s
ed
in
s
ec
ti
o
n
4
.
W
e
b
r
ief
l
y
d
is
c
u
s
s
t
h
e
c
lass
i
f
icatio
n
al
g
o
r
ith
m
i
n
s
ec
ti
o
n
5
.
Sectio
n
6
h
a
s
ex
p
er
i
m
e
n
tal
r
es
u
lt
s
w
h
er
e
we
h
av
e
a
b
r
ie
f
d
es
cr
ip
tio
n
ab
o
u
t
th
e
d
ata
s
et
w
it
h
d
is
c
u
s
s
io
n
s
ab
o
u
t
th
e
r
es
u
lt
s
an
d
co
m
p
ar
is
o
n
w
it
h
t
h
e
ex
i
s
ti
n
g
s
y
s
te
m
s
.
Fi
n
all
y
,
s
ec
tio
n
7
is
w
h
er
e
t
h
e
co
n
cl
u
s
io
n
is
p
r
o
v
id
ed
.
2.
RE
L
AT
E
D
WO
RK
S
I
n
m
a
n
y
w
o
r
k
s
i
n
W
C
E
v
i
d
eo
s
,
au
to
m
atic
id
en
ti
f
icatio
n
o
f
m
u
lti
-
an
o
m
alies
d
i
s
ea
s
e
h
as
b
ee
n
s
u
g
g
e
s
ted
.
T
h
e
m
o
s
t
co
m
m
o
n
d
is
ea
s
e
s
in
th
e
GI
tr
ac
t
ar
e
b
leed
in
g
[
7
]
,
co
lo
n
[
8
]
,
p
o
l
y
p
[
9
]
,
tu
m
o
r
[
1
0
]
,
s
to
m
ac
h
[
1
1
]
an
d
u
lcer
[
1
2
]
d
i
s
ea
s
e.
T
h
e
ex
is
ti
n
g
s
y
s
te
m
f
o
r
th
e
d
etec
tio
n
o
f
W
C
E
i
m
ag
e
co
n
s
id
er
s
o
n
l
y
o
n
e
ab
n
o
r
m
alit
y
li
k
e
b
leed
in
g
o
r
u
lcer
o
r
tu
m
o
r
a
n
d
also
m
u
lt
i
-
ab
n
o
r
m
a
lit
y
d
etec
tio
n
i
s
f
ar
f
r
o
m
s
at
is
f
ac
to
r
y
.
Ho
w
e
v
er
,
m
u
ch
w
o
r
k
r
elate
d
to
GI
ab
n
o
r
m
a
lit
y
d
etec
tio
n
h
as
d
o
n
e
in
W
C
E
v
id
eo
s
[
1
3
,
1
4
]
.
A
s
in
g
le
W
C
E
f
r
a
m
e
o
r
p
ictu
r
e
i
n
cl
u
d
es
d
is
ti
n
ct
p
r
o
b
le
m
s
i
n
cl
u
d
in
g
d
is
ti
n
c
t
co
lo
u
r
s
,
p
o
o
r
co
n
tr
a
s
t,
f
u
zz
y
ar
ea
s
,
co
m
p
lica
ted
b
ac
k
g
r
o
u
n
d
,
f
o
r
m
o
f
lesi
o
n
s
,
d
ata
o
n
tex
tu
r
e,
etc.
[
1
5
]
.
B
leed
in
g
is
a
p
r
ev
alen
t
s
y
m
p
to
m
o
f
m
an
y
g
astro
in
test
i
n
al
(
GI
)
d
is
ea
s
e
s
,
an
d
t
h
e
id
e
n
ti
f
icatio
n
o
f
b
leed
in
g
i
s
t
h
er
ef
o
r
e
o
f
e
x
ce
lle
n
t
cl
in
ical
s
i
g
n
if
ican
c
e
in
th
e
d
ia
g
n
o
s
i
s
o
f
ap
p
r
o
p
r
iate
d
is
ea
s
es.
L
i
an
d
Me
n
g
[
1
6
]
d
escr
ib
e
ch
r
o
m
in
a
n
ce
ti
m
e
as
a
co
lo
r
ch
ar
ac
ter
is
tic
a
n
d
p
er
io
d
ic
L
o
ca
l
b
in
ar
y
p
atter
n
(
L
B
P
)
as
a
tex
t
u
r
e
ch
ar
ac
ter
is
tic
f
o
r
id
en
t
if
y
in
g
t
h
e
b
leed
in
g
ar
ea
s
w
it
h
i
n
a
W
C
E
f
r
a
m
e.
T
h
e
tech
n
iq
u
e
s
w
er
e
ev
al
u
at
ed
u
s
in
g
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
S
VM
)
,
lin
er
d
is
cr
i
m
i
n
an
t
an
al
y
s
i
s
(
L
D
A
)
an
d
k
-
n
ea
r
est
n
ei
g
h
b
o
r
(
KNN)
class
if
ier
w
er
e
u
ti
lized
.
A
s
m
all
g
r
o
u
p
o
f
ce
lls
th
at
d
ev
elo
p
s
o
n
th
e
co
lo
n
's
l
in
i
n
g
ar
e
ca
lled
as
p
o
ly
p
s
d
u
e
to
u
n
u
s
u
al
ce
ll
g
r
o
w
th
.
Fo
r
i
m
ag
e
p
r
o
ce
s
s
i
n
g
b
ased
o
n
lin
ea
r
d
ata,
p
o
ly
p
id
en
ti
f
icatio
n
u
s
i
n
g
th
e
L
o
g
Gab
o
r
f
ilter
s
an
d
th
e
S
US
A
N
ed
g
e
d
etec
tio
n
alg
o
r
ith
m
i
s
d
o
n
e
i
n
k
ar
ar
g
y
r
i
s
et
al.
,
[
17
]
.
I
n
[
1
8
]
,
a
m
et
h
o
d
ca
lled
g
lo
b
al
g
eo
m
etr
ic
co
n
s
tr
ain
t
s
o
f
p
o
l
y
p
a
n
d
lo
ca
l
p
atter
n
s
o
f
i
n
te
n
s
i
t
y
v
ar
i
atio
n
ac
r
o
s
s
p
o
l
y
p
b
o
u
n
d
ar
ies
is
s
u
g
g
e
s
ted
f
o
r
cla
s
s
i
f
y
i
n
g
d
ig
est
iv
e
o
r
g
an
s
,
th
e
d
ee
p
C
NN
is
u
s
ed
f
o
r
W
C
E
i
m
ag
e
s
.
Fo
r
u
lcer
id
en
t
if
ica
tio
n
i
n
W
C
E
i
m
a
g
es,
n
e
u
r
al
n
et
w
o
r
k
s
ca
n
b
e
u
s
ed
f
o
r
t
h
e
r
e
m
o
v
al
o
f
ch
ar
ac
ter
is
tic
s
b
y
Gab
o
r
f
ilte
r
s
an
d
co
lo
r
s
an
d
tex
tu
r
e
ch
ar
ac
ter
is
tics
.
f
o
r
class
i
f
y
in
g
i
m
ag
e
s
.
T
h
e
au
th
o
r
s
o
f
[
1
9
]
p
r
o
p
o
s
ed
to
d
etec
t
th
e
u
lcer
ab
n
o
r
m
alit
y
u
s
in
g
A
d
aB
o
o
s
t
lear
n
in
g
m
et
h
o
d
.
Desp
ite
th
e
ef
f
icac
y
o
f
A
d
aB
o
o
s
t,
a
s
tr
ai
g
h
t
f
o
r
w
ar
d
R
GB
v
al
u
e
a
s
a
h
i
n
t
f
o
r
t
h
e
as
s
ig
n
m
en
t
o
f
u
lcer
d
i
s
cr
i
m
in
at
i
o
n
is
u
s
ed
to
o
b
tain
th
e
s
p
ec
if
ic
lo
ca
l
an
d
g
lo
b
al
v
is
u
al
f
ea
t
u
r
es
[
1
9
].
I
n
[
2
0
]
,
ch
ar
ac
ter
is
tics
s
u
ch
as
p
r
o
b
ab
ilit
y
o
f
b
it
p
lan
e
an
d
w
a
v
elet
-
b
ased
ch
ar
ac
ter
is
t
ics
w
er
e
r
e
m
o
v
ed
f
r
o
m
th
e
r
e
co
g
n
ized
f
ie
ld
s
an
d
u
s
ed
to
ch
ar
ac
ter
ize
u
lcer
.
Ho
g
h
a
n
et
al
.
,
[
2
1
]
d
is
cu
s
s
ed
,
th
e
Ho
o
k
w
o
r
m
p
r
ese
n
t
i
n
th
e
s
to
m
ac
h
r
elate
d
tr
ac
t
o
f
th
e
h
u
m
a
n
f
r
o
m
W
C
E
i
m
a
g
es
u
s
i
n
g
co
lo
r
m
o
d
el
-
b
as
ed
r
ec
o
g
n
itio
n
w
er
e
th
e
s
h
ad
in
g
m
o
d
els
ac
t
u
alize
d
h
er
e
ar
e
th
e
R
GB
an
d
H
S
V
m
o
d
el
s
.
Yu
a
n
et
al.
,
[
2
2
]
s
u
g
g
ested
an
e
n
h
a
n
ce
d
b
ag
o
f
f
e
atu
r
es
(
B
o
F)
tech
n
i
q
u
e
to
h
el
p
class
if
y
p
o
l
y
p
s
i
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
.
6
,
Dec
em
b
er
2
0
2
0
:
5
6
7
8
-
5
6
8
6
5680
W
C
E
i
m
a
g
es
.
I
n
s
tead
o
f
u
s
in
g
a
s
in
g
le
s
ca
le
-
i
n
v
ar
ia
n
t
f
u
n
ct
io
n
tr
a
n
s
f
o
r
m
(
SIFT
)
in
t
h
e
tr
ad
itio
n
a
l
B
o
F
m
et
h
o
d
,
d
if
f
er
en
t
tex
t
u
r
al
f
ea
tu
r
es
s
u
c
h
as
L
B
P
,
u
n
i
L
B
P
,
C
L
B
P
an
d
HOG
f
r
o
m
th
e
k
e
y
p
o
in
t
s
n
eig
h
b
o
r
h
o
o
d
s
ar
e
in
te
g
r
ated
as
s
y
n
th
et
ic
d
escr
ip
to
r
s
to
p
er
f
o
r
m
cla
s
s
i
f
icatio
n
.
T
h
e
ass
o
ciate
d
w
o
r
k
s
s
h
o
w
th
at
s
e
v
er
al
w
o
r
k
s
ar
e
ca
r
r
ied
o
u
t
to
d
ef
in
e
t
h
e
p
ar
ticu
lar
d
is
ea
s
e.
C
o
lo
r
an
d
L
B
P
[
1
6
]
ch
ar
ac
ter
is
tics
ar
e
u
s
e
d
f
o
r
th
e
id
e
n
ti
f
icatio
n
o
f
b
lee
d
in
g
.
Fo
r
t
h
e
u
lcer
th
e
Gab
o
r
f
ilter
[
1
7
]
,
co
lo
r
an
d
tex
tu
r
e
ar
e
u
s
ed
an
d
f
o
r
th
e
d
etec
tio
n
o
f
p
o
l
y
p
.
I
n
f
ac
t,
th
e
class
i
f
icatio
n
p
r
ec
is
io
n
o
f
th
e
ab
o
v
e
-
m
e
n
tio
n
ed
s
c
h
e
m
es
is
n
o
t
y
et
w
ell
ac
h
iev
ed
an
d
t
h
er
e
w
er
e
n
o
SIFT
ch
ar
ac
ter
is
tics
th
a
t
ar
e
lin
ea
r
in
s
ca
le,
r
o
tatio
n
an
d
illu
m
in
a
tio
n
,
ex
c
ep
t p
o
ly
p
.
T
o
o
v
er
co
m
e
t
h
is
p
r
o
b
le
m
a
n
d
to
ad
d
r
ess
th
e
m
u
l
ticlas
s
d
is
ea
s
e
clas
s
i
f
icatio
n
o
f
W
C
E
i
m
ag
e
s
an
d
to
co
n
s
id
er
ab
l
y
i
n
cr
ea
s
e
t
h
e
a
cc
u
r
ac
y
o
f
d
is
ea
s
e
p
r
ed
ictio
n
,
w
e
p
r
o
p
o
s
ed
a
s
y
s
te
m
u
s
i
n
g
th
e
co
m
b
i
n
atio
n
o
f
SIFT
,
C
S
-
L
B
P
an
d
A
C
C
i
s
co
m
p
ar
ed
w
it
h
co
l
o
r
,
L
B
P
an
d
B
OF
f
o
r
b
leed
in
g
an
d
p
o
ly
p
r
esp
ec
ti
v
el
y
.
T
h
e
p
r
o
p
o
s
ed
co
m
b
in
a
tio
n
i
s
also
test
ed
a
g
ain
s
t
t
h
e
a
f
o
r
e
m
en
tio
n
ed
s
y
s
te
m
s
f
o
r
tu
m
o
r
,
c
o
lo
n
,
s
to
m
ac
h
a
n
d
u
lcer
d
is
ea
s
e
s
.
Fi
g
u
r
e
2
s
h
o
w
s
t
h
e
p
r
o
p
o
s
ed
w
o
r
k
f
o
r
th
e
class
if
icatio
n
o
f
m
u
lti
-
ab
n
o
r
m
alitie
s
in
GI
tr
ac
t
u
s
i
n
g
W
C
E
i
m
ag
e
s
.
I
n
t
h
is
p
r
o
p
o
s
ed
w
o
r
k
,
a
r
a
n
g
e
o
f
an
o
m
alies
d
is
ea
s
e
i
m
a
g
es
is
d
r
a
w
n
f
r
o
m
GI
tr
ac
t
u
s
i
n
g
W
C
E
an
d
a
n
o
v
el
m
et
h
o
d
is
p
r
o
p
o
s
ed
b
ased
o
n
B
OF
b
y
i
n
te
g
r
atin
g
SIFT
,
C
S
-
L
B
P
an
d
AC
C
.
T
h
ese
f
ea
t
u
r
es
ar
e
ab
le
to
d
etec
t
th
e
in
ter
est
p
o
in
t,
t
ex
t
u
r
e
an
d
co
lo
r
in
f
o
r
m
atio
n
i
n
a
n
i
m
a
g
e
m
o
r
e
e
f
f
e
ctiv
el
y
.
Feat
u
r
es
ar
e
ca
lcu
lated
to
cr
ea
te
a
h
ig
h
-
d
i
m
e
n
s
io
n
al
d
escr
ip
to
r
f
o
r
ea
c
h
p
ictu
r
e
an
d
th
is
d
escr
ip
to
r
is
g
r
o
u
p
ed
u
s
i
n
g
th
e
K
-
m
ea
n
s
tec
h
n
iq
u
e
r
ef
e
r
r
ed
to
as
v
is
u
al
b
ag
o
f
f
e
atu
r
es,
th
e
n
SVM
m
eth
o
d
i
s
u
s
ed
to
class
i
f
y
th
e
m
u
lt
ip
le
ab
n
o
r
m
alitie
s
o
f
d
is
ea
s
e
p
r
esen
t i
n
GI
tr
ac
t a
u
t
o
m
a
ticall
y
an
d
m
o
r
e
e
f
f
ec
t
iv
el
y
.
Fi
g
u
r
e
2
.
P
r
o
p
o
s
ed
w
o
r
k
f
o
r
th
e
class
if
icatio
n
o
f
m
u
lti
-
ab
n
o
r
m
alitie
s
in
GI
tr
ac
t u
s
i
n
g
W
C
E
i
m
ag
e
s
3.
F
E
AT
U
RE
E
XT
RAC
T
I
O
N
3
.
1
.
SI
F
T
Gau
s
s
i
a
n
'
s
L
ap
lacia
n
is
g
o
o
d
to
f
in
d
i
n
ter
esti
n
g
p
o
in
t
s
(
o
r
m
ain
p
o
in
t
s
)
in
a
p
ict
u
r
e
th
a
t
ar
e
m
a
x
i
m
a
an
d
m
in
i
m
a
i
n
t
h
e
Ga
u
s
s
ia
n
p
ictu
r
e
d
is
ti
n
ctio
n
.
Up
o
n
d
etec
tio
n
o
f
i
n
ter
est
p
o
in
ts
,
ch
a
r
ac
ter
is
tics
s
u
c
h
a
s
SIFT
ar
e
o
u
tlin
ed
.
SIFT
is
an
alg
o
r
ith
m
f
o
r
th
e
d
etec
t
io
n
a
n
d
d
escr
ip
tio
n
o
f
lo
ca
l
ch
ar
ac
ter
is
tics
i
n
p
ict
u
r
es
th
at
Dav
id
L
o
w
e
r
elea
s
ed
i
n
1
9
9
9
[2
3
]
.
A
cir
cu
lar
r
eg
io
n
o
f
p
ictu
r
e
w
it
h
o
r
ien
tatio
n
is
a
SIFT
k
e
y
p
o
in
t.
Fo
u
r
p
ar
am
e
ter
s
in
t
h
is
tec
h
n
iq
u
e
ar
e
k
e
y
p
o
i
n
t
ce
n
ter
,
th
e
s
ca
le
(
th
e
ar
ea
r
ad
iu
s
)
a
n
d
its
o
r
ien
tatio
n
(
an
an
g
le
e
x
p
r
ess
ed
in
r
ad
ian
s
)
.
SIFT
d
etec
to
r
is
s
tated
to
b
e
in
v
ar
ian
t
an
d
r
o
b
u
s
t
i
n
tr
an
s
lat
io
n
,
r
o
tatio
n
,
s
ca
lin
g
,
an
d
p
ar
tl
y
i
n
v
ar
ian
t i
n
o
r
d
er
t
o
af
f
in
a
te
ch
a
n
g
e
s
i
n
d
is
to
r
tio
n
an
d
li
g
h
t
in
g
[2
3
].
3.
2
.
L
o
ca
l bina
ry
pa
t
t
er
n (
L
B
P
)
A
s
tr
o
n
g
f
u
n
ct
io
n
f
o
r
te
x
tu
r
e
class
i
f
ica
tio
n
i
s
k
n
o
w
n
to
b
e
th
e
L
B
P
[2
3
]
.
I
n
2
0
0
9
,
L
B
P
an
d
h
is
to
g
r
a
m
o
f
o
r
ien
ti
n
g
g
r
ad
ien
ts
(
HO
G)
s
h
o
w
ed
t
h
at
d
etec
tio
n
ef
f
icie
n
c
y
w
as
lar
g
el
y
i
m
p
r
o
v
ed
b
y
W
an
t
et
a
l.
[
2
4
]
.
I
n
[
2
5
]
,
L
B
P
w
a
s
u
s
ed
a
s
a
n
e
f
f
icie
n
t,
n
o
n
p
ar
a
m
etr
ic
tec
h
n
iq
u
e
f
o
r
te
x
t
u
r
e
an
al
y
s
i
s
b
y
U
n
a
y
an
d
E
k
i
n
.
L
B
P
w
as
u
s
ed
to
e
x
tr
ac
t
v
a
lu
ab
le
d
ata
f
r
o
m
m
ed
ical
i
m
a
g
es,
e
s
p
ec
iall
y
m
a
g
n
etic
b
r
ain
r
eso
n
a
n
ce
i
m
a
g
es.
A
co
n
ten
t
-
b
ased
p
ictu
r
e
r
ec
o
v
er
y
alg
o
r
it
h
m
w
a
s
u
s
e
d
to
ex
tr
ac
t
th
e
ch
a
r
ac
ter
is
tics
.
T
h
eir
ex
p
er
i
m
en
t
h
as
s
h
o
w
n
th
a
t
th
e
te
x
t
u
r
e
d
ata
alo
n
g
w
i
th
s
p
atial
ch
ar
ac
te
r
is
tics
i
s
b
etter
th
a
n
o
n
l
y
tex
t
u
r
e
ch
ar
ac
ter
is
t
ics
b
ased
o
n
in
te
n
s
i
t
y
.
I
n
2
0
0
7
,
th
e
m
icr
o
-
m
atter
n
s
w
er
e
r
e
m
o
v
ed
w
it
h
L
B
P
b
y
O
liv
er
et
al.
[
2
6
]
f
r
o
m
m
a
m
m
o
g
r
a
m
s
.
T
h
ese
m
as
s
es
ar
e
class
i
f
ied
as
b
e
n
ig
n
o
r
m
a
lig
n
an
t
w
it
h
SVM.
T
h
e
f
i
n
d
in
g
s
o
f
t
h
eir
r
esear
ch
s
h
o
w
ed
L
B
P
's ef
f
icie
n
c
y
,
as t
h
e
a
m
o
u
n
t o
f
f
al
s
e
p
o
s
iti
v
e
ch
ar
ac
ter
is
tic
s
d
ec
r
ea
s
ed
in
all
m
ass
s
ize
s
[
2
7
]
.
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:
2088
-
8708
A
n
efficien
t m
eth
o
d
to
cla
s
s
ify
GI
tr
a
ct
ima
g
es fr
o
m
WC
E
u
s
in
g
visu
a
l wo
r
d
s
(
R
.
P
o
n
n
u
s
a
my
)
5681
3
.
3
.
Cent
er
-
s
y
mm
et
ric
lo
ca
l bina
ry
pa
t
t
er
n (
L
B
P
)
A
d
escr
ip
tio
n
o
f
t
h
e
r
eg
io
n
o
f
co
n
ce
r
n
h
as
b
ee
n
cr
ea
ted
f
o
r
ce
n
ter
-
s
y
m
m
etr
ic
lo
ca
l
b
in
ar
y
p
atter
n
s
(
C
S
-
L
B
P
)
[
28
,
29
]
.
C
S
-
L
B
P
s
ee
k
s
to
g
e
n
er
ate
s
h
o
r
ter
h
is
to
g
r
a
m
s
f
o
r
a
lar
g
er
a
m
o
u
n
t
o
f
L
B
P
lab
els
t
h
at
ar
e
m
o
r
e
s
u
itab
le
f
o
r
u
s
e
i
n
r
eg
io
n
al
d
escr
ip
to
r
s
.
I
n
f
la
t
p
ictu
r
e
ar
ea
s
,
C
S
-
L
B
P
w
as
al
s
o
in
te
n
d
ed
to
h
av
e
g
r
ea
ter
s
tab
ilit
y
.
I
n
C
S
-
L
B
P
,
p
ix
el
v
alu
es
ar
e
co
m
p
ar
ed
s
y
m
m
etr
i
ca
ll
y
w
i
th
th
e
ce
n
ter
p
i
x
el,
n
o
t
w
it
h
t
h
e
ce
n
ter
p
ix
el,
b
u
t
w
it
h
t
h
e
o
p
p
o
s
in
g
p
ix
el.
I
n
ad
d
itio
n
,
r
o
b
u
s
t
n
es
s
is
ac
h
ie
v
ed
in
f
la
t
p
ictu
r
e
ar
e
as
b
y
t
h
r
esh
o
ld
i
n
g
v
ar
iatio
n
s
in
g
r
a
y
le
v
els
w
i
th
a
tin
y
v
al
u
e
T
.
−
,
,
(
,
)
=
∑
(
−
+
/
2
)
2
/
2
−
1
=
0
,
(
)
=
{
1
>
0
ℎ
W
h
er
e
n
i
a
n
d
n
i
+
(N/2)
eq
u
all
y
s
p
ac
ed
p
ix
els
o
n
a
c
ir
cle
o
f
r
ad
iu
s
R
co
r
r
esp
o
n
d
s
to
th
e
g
r
a
y
v
al
u
es
o
f
th
e
ce
n
tr
al
-
s
y
m
m
e
tr
ic
p
air
s
o
f
N
-
P
ix
e
ls
.
T
h
e
T
li
m
it
v
al
u
e
i
n
o
u
r
e
x
p
er
i
m
e
n
t
s
i
s
ab
o
u
t
1
%
o
f
th
e
p
ix
el
v
al
u
e
r
an
g
e.
Si
n
ce
in
f
o
r
m
a
tio
n
f
o
r
th
e
r
eg
io
n
r
a
n
g
e
f
r
o
m
0
to
1
,
T
is
s
e
t
to
0
.
0
1
.
T
h
e
s
ize
o
f
th
e
ar
ea
is
8
.
T
h
e
r
ad
iu
s
is
s
et
to
2
.
I
t is
w
o
r
th
n
o
tin
g
t
h
at
C
S
-
L
B
P
's ad
v
a
n
tag
e
o
v
er
L
B
P
is
n
o
t j
u
s
t b
ec
a
u
s
e
o
f
th
e
d
ec
r
ea
s
e
o
f
t
h
e
d
i
m
en
s
io
n
s
,
b
u
t
b
ec
au
s
e
th
e
C
S
-
L
B
P
is
b
etter
ab
le
t
o
ca
p
tu
r
e
t
h
e
g
r
ad
ien
t
d
ata
th
an
t
h
e
f
u
n
d
a
m
e
n
ta
l
L
B
P
.
L
B
P
an
d
C
S
-
L
B
P
ex
p
er
im
e
n
t
s
h
a
v
e
d
e
m
o
n
s
tr
ate
d
C
S
-
L
B
P
'
s
ad
v
a
n
ta
g
es
o
v
e
r
L
B
P
,
p
ar
ticu
lar
l
y
s
u
b
s
ta
n
tial d
ec
r
ea
s
e
i
n
d
i
m
e
n
s
io
n
w
h
ile
m
ai
n
t
ai
n
i
n
g
d
i
s
tin
g
u
is
h
ab
ili
t
y
.
3
.
4
.
Aut
o
co
lo
r
c
o
rr
elo
g
ra
m
T
h
e
s
u
g
g
e
s
ted
s
c
h
e
m
e
u
s
es
th
e
au
to
co
r
r
elo
g
r
a
m
to
ca
lc
u
lat
e
th
e
co
lo
r
f
ea
tu
r
e.
L
et
[
D]
d
en
o
te
a
s
e
t
o
f
{d
1
.
.
.
d
D}
f
ix
ed
d
is
ta
n
ce
s
.
T
h
en
th
e
p
ictu
r
e
I
co
r
r
elo
g
r
a
m
is
d
escr
ib
ed
at
a
r
an
g
e
d
f
o
r
lev
el
p
air
(
,
)
.
,
(
)
(
)
≡
∈
,
∈
⌊
2
∈
|
1
−
2
=
|
⌋
W
h
ich
p
r
o
v
id
es
t
h
e
p
o
s
s
ib
ili
t
y
t
h
at
i
f
a
p
ix
el
p
1
is
lev
e
l,
a
p
ix
el
p
2
is
lev
el
at
t
h
e
r
a
n
g
e
d
in
s
o
m
e
d
ir
ec
tio
n
f
r
o
m
t
h
e
p
ix
el
p
1
.
T
h
e
s
p
atial
co
r
r
elatio
n
o
f
th
e
s
a
m
e
co
n
c
e
n
tr
atio
n
s
is
f
o
u
n
d
in
t
h
e
au
to
co
r
r
elo
g
r
am
.
∝
(
)
(
)
=
,
(
)
(
)
I
t
p
r
o
v
id
es
th
e
lik
eli
h
o
o
d
th
at
p
i
x
els
p
1
an
d
p
2
,
d
s
ep
ar
ate
f
r
o
m
ea
ch
o
th
er
,
ar
e
o
f
th
e
s
a
m
e
le
v
el
.
T
h
e
r
an
g
e
m
ea
s
u
r
e
m
en
t
b
et
w
ee
n
h
i
s
to
g
r
a
m
s
,
a
u
to
co
r
r
elo
g
r
a
m
s
,
an
d
co
r
r
elo
g
r
a
m
s
is
t
h
e
L
1
s
ta
n
d
ar
d
th
at
is
a
co
m
p
u
tati
o
n
a
ll
y
f
as
t te
ch
n
iq
u
e
u
s
ed
in
[
3
0
-
3
3
].
3.
5
.
CS
-
L
B
P
,
SI
F
T
a
nd
ACC
f
ea
t
ures int
eg
ra
t
io
n
I
f
t
h
e
b
ac
k
g
r
o
u
n
d
i
s
co
m
p
l
ic
ated
o
r
co
r
r
u
p
ted
w
i
th
n
o
is
e,
SIFT
ca
n
p
er
f
o
r
m
b
ad
l
y
,
C
S
-
L
B
P
w
i
th
s
tan
d
ar
d
ized
p
atter
n
s
i
s
co
m
p
le
m
e
n
tar
y
to
SIFT
b
y
f
i
lter
in
g
o
u
t
t
h
e
s
e
n
o
i
s
es
[
34
]
.
W
e
b
eliev
e
th
at
th
e
ch
ar
ac
ter
is
t
ics
o
f
an
ite
m
i
n
a
i
m
a
g
e
ca
n
b
e
f
aster
r
ec
o
r
d
ed
b
y
m
i
x
i
n
g
t
h
ese
t
h
r
ee
f
ea
t
u
r
es.
T
h
is
r
esear
ch
th
er
ef
o
r
e
p
r
o
p
o
s
es
th
e
in
cl
u
s
i
o
n
o
f
SIFT
,
C
S
-
L
B
P
an
d
A
C
C
at
p
atch
lev
el
a
n
d
p
ictu
r
e
lev
el.
W
e
d
escr
ib
e
p
i
(
x
,
y
,
σ
,
θ)
as
a
k
ey
p
o
in
t
s
p
o
tted
b
y
SIFT
ap
p
r
o
ac
h
,
w
h
er
e
(
x
,
y
)
is
th
e
p
o
s
it
io
n
o
f
p
ix
el
p
i
in
th
e
o
r
ig
i
n
al
i
m
a
g
e,
σ
an
d
θ
is
th
e
s
ca
le
a
n
d
m
ai
n
d
ir
ec
tio
n
o
f
p
i
r
esp
e
ctiv
el
y
.
σ
m
ea
n
s
to
th
e
co
n
f
i
d
en
t
lev
el
o
f
p
i
in
Gau
s
s
ia
n
P
y
r
a
m
id
.
T
ak
e
a
r
eg
io
n
w
it
h
s
ize
o
f
8
×
8
as
a
p
atch
w
h
er
e
p
i
is
th
e
ce
n
ter
o
f
th
e
p
atc
h
,
th
en
t
h
e
SIFT
,
C
S
-
L
B
P
an
d
AC
C
d
escr
ip
to
r
s
ar
e
b
u
ilt as
f
o
ll
o
w
s
:
Step
1
.
Use
1
2
8
-
d
im
e
n
s
io
n
al
SIFT
d
escr
ip
t
o
r
t
o
d
escr
ib
e
e
ac
h
k
e
y
p
o
i
n
t
p
i
in
a
p
atch
,
d
e
n
o
ted
as
SIFT
i
th
e
i
m
ag
e.
Step
2
.
C
h
o
o
s
e
a
4
×
4
r
eg
io
n
ar
o
u
n
d
p
i
an
d
co
m
p
u
te
th
e
u
n
i
f
o
r
m
p
atter
n
o
f
ea
c
h
p
ix
el.
T
h
ese
d
es
cr
ip
to
r
s
ar
e
co
m
p
o
s
ed
as a
6
4
-
d
i
m
e
n
s
i
o
n
al
v
ec
to
r
,
i.e
.
−
=
[
−
4
,
1
1
,
−
4
,
1
2
…
…
−
4
,
1
16
]
Step
3
.
Fo
r
ev
er
y
p
atch
AC
C
f
ea
tu
r
e
ar
e
ca
lcu
lated
Step
4
.
Fin
all
y
,
t
h
e
f
ea
tu
r
e
v
ec
to
r
co
m
p
u
ted
b
y
co
m
b
i
n
i
n
g
th
e
th
r
ee
f
ea
t
u
r
es is
d
escr
ib
ed
as
(
,
−
,
,
)
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
.
6
,
Dec
em
b
er
2
0
2
0
:
5
6
7
8
-
5
6
8
6
5682
4.
VIS
UA
L
B
AG
O
F
WO
RD
S
4
.
1
.
K
-
m
ea
ns
a
lg
o
rit
h
m
T
h
e
aim
o
f
k
-
m
ea
n
s
alg
o
r
it
h
m
is
to
clu
s
ter
th
e
i
n
f
o
r
m
ati
o
n
an
d
is
o
n
e
o
f
t
h
e
ea
s
ies
t
m
et
h
o
d
s
o
f
clu
s
ter
i
n
g
th
e
p
ar
titi
o
n
s
.
C
l
u
s
t
er
in
g
t
h
e
p
ictu
r
e
co
n
s
is
t
s
o
f
g
r
o
u
p
in
g
t
h
e
p
ix
el
s
ac
co
r
d
in
g
t
o
ce
r
tain
f
ea
t
u
r
es.
W
e
m
u
s
t
o
r
ig
i
n
all
y
id
e
n
ti
f
y
th
e
n
u
m
b
er
o
f
cl
u
s
ter
s
i
n
t
h
e
k
-
m
ea
n
s
alg
o
r
it
h
m
[
3
5
]
.
T
h
en
th
e
ce
n
ter
o
f
th
e
k
-
cl
u
s
ter
is
r
a
n
d
o
m
l
y
s
el
ec
ted
.
T
h
e
d
is
tan
ce
to
ea
c
h
clu
s
ter
ce
n
ter
b
et
w
ee
n
ea
ch
p
ix
el
i
s
ca
lc
u
lated
.
E
u
clid
ea
n
d
is
ta
n
ce
is
u
s
ed
i
n
p
ar
ticu
lar
.
Usi
n
g
th
e
r
a
n
g
e
f
o
r
m
u
la,
s
i
n
g
le
p
ix
els
ar
e
li
k
en
ed
to
all
clu
s
ter
ce
n
ter
s
.
T
h
e
p
ix
el
is
tr
a
n
s
f
er
r
ed
to
a
s
p
ec
if
i
c
clu
s
ter
w
it
h
t
h
e
s
h
o
r
test
r
an
g
e
b
et
w
ee
n
all.
T
h
e
ce
n
ter
is
th
e
n
r
ea
s
s
ess
ed
.
Ag
ai
n
,
ea
ch
p
i
x
el
is
co
m
p
ar
ed
to
all
ce
n
tr
o
id
s
a
n
d
th
e
p
r
o
ce
s
s
es
m
e
n
tio
n
ed
ab
o
v
e
ar
e
co
n
tin
u
ed
u
n
t
il th
e
p
i
x
el
ar
e
g
r
o
u
p
ed
in
t
o
a
s
u
itab
le
clu
s
ter
w
it
h
th
e
f
o
llo
w
i
n
g
alg
o
r
it
h
m
d
e
s
cr
ib
ed
.
4
.
2
.
B
a
g
o
f
f
ea
t
ures
T
h
e
B
ag
-
of
-
f
ea
t
u
r
es
(
B
OF)
te
ch
n
iq
u
es
is
m
ai
n
l
y
i
n
f
l
u
en
ce
d
b
y
t
h
e
n
o
tio
n
o
f
b
a
g
-
of
-
w
o
r
d
s
[
3
6
]
th
at
w
a
s
w
id
el
y
u
s
ed
i
n
te
x
t
m
i
n
i
n
g
.
I
n
t
h
e
B
OW
m
o
d
el,
e
v
er
y
ter
m
is
co
n
s
id
er
ed
to
b
e
au
to
n
o
m
o
u
s
alt
h
o
u
g
h
v
er
y
co
n
tr
a
in
t
u
it
iv
e
a
n
d
well
u
til
ized
w
ith
o
u
ts
tan
d
i
n
g
p
er
f
o
r
m
a
n
ce
i
n
s
p
a
m
f
ilt
r
atio
n
an
d
to
p
ical
m
o
d
eli
n
g
[
3
7
]
.
E
ac
h
im
a
g
e
is
ch
ar
ac
ter
ized
b
y
a
s
et
o
f
o
r
d
er
less
lo
ca
l
ch
ar
ac
ter
is
tics
i
n
th
e
B
OF
m
o
d
el,
later
s
tu
d
y
h
as
s
h
o
w
n
it
e
f
f
ica
c
y
in
i
m
ag
e
p
r
o
ce
s
s
i
n
g
.
I
t
h
a
s
t
w
o
m
a
in
co
n
ce
p
t
s
:
lo
ca
l
f
ea
t
u
r
es
an
d
co
d
eb
o
o
k
.
T
h
e
ess
en
tial
a
s
p
ec
t
o
f
t
h
e
B
o
F
co
n
ce
p
t
is
to
ex
tr
ac
t
g
l
o
b
al
im
a
g
e
d
escr
ip
to
r
w
h
ic
h
ar
e
co
m
p
u
ted
f
r
o
m
th
e
co
llectio
n
o
f
lo
ca
l
p
r
o
p
er
t
ies
lik
e
SIFT
,
C
S
-
L
B
P
an
d
AC
C
.
T
h
e
SIFT
p
atch
es
ar
e
tin
y
r
ec
ta
n
g
u
lar
ar
ea
s
w
it
h
a
f
o
cu
s
o
n
p
o
in
t
o
f
co
n
c
er
n
an
d
th
e
C
S
-
L
B
P
p
atch
es
ar
e
tin
y
r
o
u
n
d
zo
n
es
w
i
th
t
h
e
r
eq
u
ir
ed
r
ad
iu
s
an
d
s
ev
er
al
s
a
m
p
li
n
g
p
o
in
ts
.
A
u
t
o
co
r
r
elo
g
r
am
co
llect
s
o
n
l
y
id
en
tical
co
lo
r
v
alu
es i
n
t
h
e
s
p
atial
co
r
r
elatio
n
.
C
o
d
eb
o
o
k
is
a
w
a
y
to
r
ep
r
esen
t
an
i
m
a
g
e
b
y
a
s
et
o
f
lo
ca
l
f
ea
t
u
r
es
[
38
]
.
T
h
e
id
ea
is
to
g
r
o
u
p
th
e
f
ea
t
u
r
e
d
escr
ip
to
r
s
f
o
r
all
p
atch
es
o
n
t
h
e
b
asis
o
f
a
clu
s
ter
n
u
m
b
er
a
n
d
ea
ch
clu
s
ter
is
a
v
is
u
al
w
o
r
d
to
f
o
r
m
a
co
d
eb
o
o
k
.
E
ac
h
i
m
ag
e
ca
n
b
e
d
ep
icted
,
af
ter
th
e
co
d
eb
o
o
k
h
as
b
ee
n
o
b
tain
ed
,
b
y
t
h
e
v
i
s
u
a
l
co
d
eb
o
o
k
g
r
ap
h
ic
f
r
eq
u
en
c
y
h
is
to
g
r
a
m
B
o
F.
5.
CL
AS
SI
F
I
CAT
I
O
N
US
I
N
G
SVM
I
n
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
,
th
e
SVM
[
3
9
]
is
em
p
lo
y
ed
to
class
if
y
t
h
e
W
C
E
i
m
a
g
es.
SVM
cl
ass
i
f
ier
is
th
e
b
est
o
p
tio
n
to
class
if
y
p
r
o
b
le
m
,
s
i
n
ce
o
u
r
p
r
o
b
le
m
i
s
to
class
i
f
y
s
e
v
en
clas
s
es
o
f
ab
n
o
r
m
alitie
s
p
r
esen
t
in
GI
tr
ac
t.
C
o
n
s
id
er
i
n
g
a
tr
ain
in
g
d
ataset
w
h
ic
h
co
n
s
is
t
s
o
f
N
i
m
a
g
es
w
i
th
f
ea
t
u
r
e
v
ec
to
r
s
x
i
,
i=1
,
2
,
.
.
.
N,
w
h
e
r
e
e
a
c
h
M
d
i
m
e
n
s
i
o
n
a
l
e
x
p
r
e
s
s
i
o
n
p
r
o
f
i
l
e
x
i
=
x
i
(
n
)
,
n
=
1
,
2
,
,
.
.
.
,
M
i
s
a
s
s
o
c
i
a
t
e
d
w
i
t
h
a
f
e
a
t
u
r
e
v
a
l
u
e
y
i
(
+
1
,
-
1
)
.
T
h
e
o
b
j
ec
tiv
e
is
to
f
i
n
d
an
d
M
d
i
m
e
n
s
io
n
al
d
ec
i
s
io
n
v
ec
t
o
r
w
=
[
w
1
,
w
2
.
.
.
.
w
M
]
T
d
u
e
t
o
th
e
d
i
s
cr
i
m
i
n
ati
n
g
f
u
n
ctio
n
f
(
x
)
=f
(
w
,
x
)
.
Su
c
h
th
a
t
:
w
T
x
i
+b
+1
,
f
o
r
all
p
o
s
itiv
e
x
i
w
T
x
i
+b
≤
-
1
,
f
o
r
all
n
e
g
ati
v
e
x
i
C
o
n
s
id
er
in
g
a
n
e
m
p
ir
ical
v
ec
to
r
a
an
d
(
N*
N)
d
i
m
en
s
io
n
a
l
k
er
n
el
m
atr
i
x
K
w
it
h
it
s
(
i,
j
)
th
ele
m
e
n
t
K(
x
i
,
y
i
)
,
th
e
d
ec
is
io
n
b
o
u
n
d
ar
y
is
c
h
ar
ac
ter
ized
b
y
f
(
x
)
=
0
,
W
h
er
e
:
f
(
x
)
=
∑
(
,
=
1
)
+b
w
h
er
e
b
an
d
a
i
ar
e
b
i
as a
n
d
w
e
ig
h
ts
r
esp
ec
ti
v
el
y
.
6.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
S
6
.
1
.
Da
t
a
s
et
T
h
e
d
ataset
is
co
llected
f
r
o
m
Kv
asir
co
n
tai
n
i
n
g
i
m
a
g
es
o
f
GI
tr
ac
t.
T
h
e
an
ato
m
ical
f
e
atu
r
es
ar
e
Z
-
li
n
e,
p
y
lo
r
u
s
a
n
d
ce
c
u
m
,
w
h
ile
eso
p
h
a
g
iti
s
,
b
leed
in
g
,
p
o
ly
p
s
an
d
u
lcer
ated
co
liti
s
ar
e
th
e
p
at
h
o
lo
g
ical
f
i
n
d
in
g
s
.
T
h
e
d
ataset
co
n
tain
s
i
m
ag
e
s
o
f
m
u
lti
-
ab
n
o
r
m
a
liti
e
s
d
is
ea
s
es
w
h
ich
h
as
3
5
0
0
o
f
i
m
a
g
es
in
7
class
e
s
i.e
.
5
0
0
im
a
g
es
f
r
o
m
ea
c
h
class
f
o
r
4
0
p
atien
ts
.
T
h
e
s
et
o
f
im
ag
e
s
in
ea
c
h
clas
s
is
d
iv
id
e
i
n
to
t
w
o
ca
te
g
o
r
ies
:
tr
ain
i
n
g
a
n
d
test
in
g
s
et.
A
f
i
v
e
-
f
o
ld
s
tr
ate
g
y
to
cr
o
s
s
v
alid
a
tio
n
h
as
b
ee
n
i
m
p
le
m
en
ted
in
th
e
p
r
o
p
o
s
ed
w
o
r
k
.
I
n
t
h
is
w
o
r
k
,
8
0
%
o
f
t
h
e
i
m
a
g
es
w
er
e
r
an
d
o
m
l
y
s
elec
ted
f
o
r
tr
ain
in
g
f
o
r
ea
c
h
cla
s
s
a
n
d
th
e
o
th
er
2
0
%
f
o
r
test
i
n
g
.
T
h
e
m
u
lti
-
ab
n
o
r
m
al
iti
es
d
is
ea
s
e
co
n
ta
in
s
t
h
e
Z
-
L
i
n
e
,
B
leed
in
g
,
P
y
lo
r
u
s
,
C
ec
u
m
,
E
s
o
p
h
ag
iti
s
,
P
o
l
y
p
s
an
d
Ulce
r
ativ
e
C
o
liti
s
.
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:
2088
-
8708
A
n
efficien
t m
eth
o
d
to
cla
s
s
ify
GI
tr
a
ct
ima
g
es fr
o
m
WC
E
u
s
in
g
visu
a
l wo
r
d
s
(
R
.
P
o
n
n
u
s
a
my
)
5683
6
.
2
.
Resul
t
s
a
nd
dis
cu
s
s
io
n
I
n
o
u
r
p
r
o
p
o
s
e
d
m
e
th
o
d
,
th
e
W
C
E
im
ag
es
a
r
e
d
i
v
i
d
e
d
in
to
n
u
m
b
e
r
o
f
p
a
t
ch
es
.
F
r
o
m
th
e
p
a
tc
h
e
d
im
ag
es
f
e
atu
r
e
s
a
r
e
ex
t
r
a
c
te
d
u
s
in
g
th
r
e
e
m
et
h
o
d
s
n
am
e
ly
C
S
-
L
B
P,
S
I
F
T
a
n
d
c
o
m
b
in
e
d
CS
-
L
B
P+
S
I
F
T
+A
C
C
.
F
i
r
s
t
,
s
tu
d
y
o
f
c
o
d
e
b
o
o
k
s
i
z
e
f
o
r
C
S
-
L
B
P,
S
I
F
T
a
n
d
C
S
-
L
B
P
+
S
I
F
T
+A
C
C
w
i
th
S
VM
is
c
al
cu
l
at
e
d
.
W
e
s
e
l
e
ct
e
d
t
h
e
s
i
z
e
o
f
c
o
d
e
b
o
o
k
f
r
o
m
{
2
5
0
,
5
0
0
,
7
5
0
,
1
0
0
0
}
.
T
h
e
p
e
r
f
o
r
m
an
c
e
is
s
h
o
w
n
in
F
ig
u
r
e
3
.
F
r
o
m
th
e
ex
p
e
r
im
en
ts
,
w
e
o
b
t
a
in
e
d
th
e
b
es
t
s
i
z
e
o
f
c
o
d
e
b
o
o
k
is
7
5
0
w
ith
p
at
ch
s
i
z
e
8
*
8
f
o
r
th
e
d
a
t
as
e
t
c
o
n
s
i
d
e
r
e
d
i
n
th
e
e
x
p
e
r
im
en
t
s
.
F
o
r
m
ea
s
u
r
in
g
th
e
ac
cu
r
a
cy
,
th
e
s
en
s
i
t
iv
ity
an
d
s
p
e
ci
f
i
c
ity
m
e
t
h
o
d
is
u
s
e
d
.
Fig
u
r
e
3
.
P
er
f
o
r
m
a
n
ce
o
f
v
ar
y
in
g
co
d
eb
o
o
k
s
ize
Sp
ec
if
icit
y
:
T
h
e
a
m
o
u
n
t
o
f
r
ig
h
t
ad
v
er
s
e
s
tate
m
e
n
ts
s
ep
ar
ated
b
y
th
e
co
m
p
lete
n
eg
a
ti
v
e
n
u
m
b
er
s
i
s
ca
lcu
lated
.
Sp
ec
if
icit
y
=
+
Sen
s
iti
v
it
y
: I
t is ca
lc
u
lated
th
e
n
u
m
b
er
o
f
n
e
g
ati
v
e
p
r
ed
ictio
n
s
d
iv
id
ed
b
y
t
h
e
to
tal
n
u
m
b
er
o
f
n
eg
a
tiv
e
s
.
Sen
s
iti
v
it
y
=
+
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
ab
n
o
r
m
a
liti
es
u
s
i
n
g
C
S
-
L
B
P
,
SIFT
an
d
co
m
b
in
ed
C
S
-
L
B
P
+SI
FT+
AC
C
w
it
h
SVM
s
h
o
w
n
in
T
ab
le
1
.
W
h
en
co
m
p
ar
i
n
g
ea
c
h
cla
s
s
,
t
h
e
p
er
f
o
r
m
an
ce
o
f
co
m
b
i
n
ed
C
S
-
L
B
P
+SI
FT
s
h
o
w
s
a
n
ac
cu
r
ac
y
o
f
7
9
.
9
1
%
f
o
r
E
s
o
p
h
ag
i
t
is
,
7
6
.
2
1
%
f
o
r
Z
-
L
in
e,
8
0
.
2
3
%
f
o
r
ce
cu
m
,
8
4
.
3
9
%
f
o
r
P
o
ly
p
s
,
8
5
.
9
2
%
f
o
r
Ulce
r
ativ
e
C
o
liti
s
,
9
0
.
9
1
%
f
o
r
P
y
lo
r
u
s
a
n
d
9
4
.
8
0
%
b
leed
in
g
is
o
b
tain
ed
.
T
ab
le
2
s
h
o
w
s
t
h
e
co
n
f
u
s
io
n
m
atr
i
x
f
o
r
W
C
E
i
m
a
g
e
class
if
icatio
n
.
6
.
3
.
Co
m
pa
ri
s
o
n w
it
h e
x
is
t
i
ng
w
o
rk
Fu
r
t
h
er
t
h
e
p
r
o
p
o
s
ed
w
o
r
k
i
s
co
m
p
ar
ed
w
i
th
ex
i
s
t
in
g
W
C
E
ab
n
o
r
m
ali
t
y
cla
s
s
i
f
icatio
n
tech
n
i
q
u
es
[
2
2
,
4
0
-
42
]
.
I
n
[
2
2
]
,
Yu
an
et
al.
,
class
if
ied
t
h
e
im
ag
e
s
in
to
t
h
e
n
o
r
m
al
o
n
e
s
an
d
p
o
ly
p
s
u
s
i
n
g
V
Q
an
d
VQ
is
u
s
ed
to
en
co
d
e
th
e
f
ea
tu
r
e
s
an
d
also
it
s
h
o
w
s
lo
ca
l
f
e
at
u
r
es
b
y
th
e
ir
n
ea
r
est
c
o
d
e
w
o
r
d
s
.
I
n
[
41
]
,
s
tatis
t
ical
b
ased
co
lo
r
,
s
p
atial
an
d
tex
t
u
r
e
ar
e
f
ea
tu
r
es
u
s
i
n
g
b
ag
o
f
v
is
u
al
m
et
h
o
d
s
ar
e
p
r
o
p
o
s
ed
b
y
H
w
a
n
g
.
Na
w
ar
at
h
a
n
et
al.
,
[
42
]
p
r
o
p
o
s
ed
a
m
et
h
o
d
to
d
en
o
te
i
m
ag
e
f
ea
tu
r
e
b
y
tex
to
n
h
is
to
g
r
a
m
w
h
er
e
L
B
P
f
ea
t
u
r
e
s
ar
e
clu
s
ter
to
o
b
tain
ed
tex
to
n
s
.
T
h
e
ac
cu
r
ac
y
o
f
SVM
w
i
t
h
C
S
-
L
B
P
+SI
FT
+A
C
C
f
o
r
Mu
lti
-
ab
n
o
r
m
alitie
s
class
i
f
icatio
n
is
s
h
o
w
n
in
F
ig
u
r
e
4
.
I
n
p
r
o
p
o
s
ed
w
o
r
k
,
co
m
b
i
n
ati
o
n
SIFT
+CS
-
L
B
P
+A
C
C
i
s
u
s
ed
f
o
r
m
u
lti
-
ab
n
o
r
m
al
it
y
cla
s
s
i
f
icatio
n
.
T
ab
le
3
s
h
o
w
s
t
h
e
ac
cu
r
ac
y
o
f
p
r
o
p
o
s
ed
ab
n
o
r
m
alitie
s
w
i
th
ex
is
ti
n
g
w
o
r
k
.
Ma
j
o
r
ity
o
f
ex
is
ti
n
g
w
o
r
k
is
f
o
cu
s
s
ed
o
n
o
n
l
y
o
n
e
o
r
t
w
o
ty
p
e
ab
n
o
r
m
al
it
y
d
etec
tio
n
an
d
th
e
ac
cu
r
ac
y
i
s
also
n
o
t
u
p
to
th
e
lev
el
o
f
s
atis
f
ac
to
r
y
.
B
u
t
w
e
f
o
cu
s
ed
o
n
all
clas
s
es
o
f
GI
tr
ac
t
d
is
e
ases
an
d
w
e
o
b
tai
n
ed
r
e
m
ar
k
a
b
le
i
m
p
r
o
v
e
m
e
n
t
i
n
all
class
es
o
f
GI
tr
ac
t
d
is
ea
s
e
s
an
d
th
e
o
v
er
all
ac
cu
r
ac
y
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
is
s
i
g
n
i
f
i
ca
n
tl
y
h
i
g
h
a
n
d
is
8
4
.
6
2
%
w
h
ic
h
is
o
w
in
g
to
th
e
co
m
b
in
atio
n
o
f
p
r
o
p
o
s
ed
ef
f
ec
tiv
e
te
x
t
u
r
e
an
d
co
lo
r
f
e
atu
r
es.
T
h
e
r
esu
lt
s
o
b
tain
ed
f
o
r
p
o
l
y
p
,
u
lcer
,
b
leed
in
g
,
E
s
o
p
h
ag
i
tis
,
Z
-
li
n
e,
C
ec
u
m
,
p
y
lo
r
u
s
is
8
4
.
3
9
%,
8
5
.
9
2
%,
9
6
.
8
5
%
,
7
9
.
9
1
%,
7
6
.
2
1
%,
8
0
.
2
3
% a
n
d
9
0
.
9
1
% r
esp
ec
tiv
ely
.
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
.
6
,
Dec
em
b
er
2
0
2
0
:
5
6
7
8
-
5
6
8
6
5684
Fig
u
r
e
4
.
S
h
o
w
s
t
h
e
ac
cu
r
ac
y
o
f
SVM
w
it
h
C
S
-
L
B
P
+SI
FT
+A
C
C
f
o
r
Mu
l
ti
-
ab
n
o
r
m
alitie
s
class
i
f
icatio
n
T
ab
le
1
.
P
er
f
o
r
m
a
n
ce
o
f
A
b
n
o
r
m
alitie
s
u
s
in
g
SV
M
w
it
h
C
S
-
L
B
P
,
SIFT
an
d
co
m
b
in
ed
C
S
-
L
B
P
w
it
h
SIFT
A
b
n
o
r
mal
i
t
i
e
s
P
e
r
f
o
r
man
c
e
(
in
%)
CS
-
L
B
P
+
S
I
F
T
CS
-
L
B
P
+
S
I
F
T
+
A
C
C
Eso
p
h
a
g
i
t
i
s
A
c
c
u
r
a
c
y
7
2
.
5
0
7
9
.
9
1
S
e
n
si
t
i
v
i
t
y
7
2
.
3
0
7
8
.
9
5
S
p
e
c
i
f
i
c
i
t
y
9
4
.
2
8
9
6
.
1
Z
-
L
i
n
e
A
c
c
u
r
a
c
y
7
5
.
0
1
7
6
.
2
1
S
e
n
si
t
i
v
i
t
y
7
4
.
0
5
7
5
.
2
1
S
p
e
c
i
f
i
c
i
t
y
9
5
.
2
5
9
5
.
1
9
C
e
c
u
m
A
c
c
u
r
a
c
y
7
9
.
2
1
8
0
.
2
3
S
e
n
si
t
i
v
i
t
y
7
8
.
9
5
8
0
.
0
1
S
p
e
c
i
f
i
c
i
t
y
9
5
.
6
9
0
.
9
P
o
l
y
p
s
A
c
c
u
r
a
c
y
8
3
.
9
1
8
4
.
3
9
S
e
n
si
t
i
v
i
t
y
8
3
.
2
0
8
4
.
2
1
S
p
e
c
i
f
i
c
i
t
y
9
5
.
1
2
9
0
.
9
U
l
c
e
r
a
t
i
v
e
C
o
l
i
t
i
s
A
c
c
u
r
a
c
y
8
0
.
2
0
8
5
.
9
2
S
e
n
si
t
i
v
i
t
y
8
0
.
7
1
8
4
.
2
1
S
p
e
c
i
f
i
c
i
t
y
9
3
.
2
5
9
3
.
6
4
P
y
l
o
r
u
s
A
c
c
u
r
a
c
y
8
9
.
0
1
9
0
.
9
1
S
e
n
si
t
i
v
i
t
y
8
8
.
2
0
8
9
.
5
7
S
p
e
c
i
f
i
c
i
t
y
9
3
.
2
5
9
3
.
2
5
B
l
e
e
d
i
n
g
A
c
c
u
r
a
c
y
8
8
.
9
2
9
6
.
8
5
S
e
n
si
t
i
v
i
t
y
8
8
.
2
3
9
0
.
1
1
S
p
e
c
i
f
i
c
i
t
y
9
2
.
2
8
9
7
.
1
2
T
ab
le
2
.
C
o
n
f
u
s
io
n
m
atr
i
x
f
o
r
W
C
E
i
m
a
g
e
class
if
icatio
n
Eso
p
h
a
g
i
t
i
s
Z
-
L
i
n
e
C
e
c
u
m
P
o
l
y
p
s
U
l
c
e
r
a
t
i
v
e
C
o
l
i
t
i
s
P
y
l
o
r
u
s
B
l
e
e
d
i
n
g
S
e
n
si
t
i
v
i
t
y
Eso
p
h
a
g
i
t
i
s
7
8
.
9
0
2
1
.
0
5
0
0
0
0
0
7
8
.
9
5
Z
-
L
i
n
e
1
0
.
0
5
7
5
.
2
1
0
8
.
2
3
0
6
.
4
5
0
7
5
.
2
1
C
e
c
u
m
0
6
.
6
7
8
0
.
0
0
1
3
.
3
3
0
0
0
8
0
.
0
0
P
o
l
y
p
s
0
0
1
4
.
0
2
8
0
.
0
0
0
5
.
9
5
0
8
0
.
0
0
U
l
c
e
r
a
t
i
v
e
C
o
l
i
t
i
s
5
.
2
6
0
0
5
.
2
6
8
4
.
2
1
5
.
2
6
0
8
4
.
2
1
P
y
l
o
r
u
s
5
.
2
0
2
.
3
0
2
.
9
3
0
0
8
9
.
5
7
0
8
9
.
5
7
B
l
e
e
d
i
n
g
0
0
0
3
.
1
5
0
0
9
6
.
8
5
9
6
.
8
5
S
p
e
c
i
f
i
c
i
t
y
9
6
.
1
0
9
5
.
1
9
9
0
.
9
9
0
.
9
0
9
3
.
6
4
9
3
.
2
5
9
8
.
1
2
T
ab
le
3
.
P
er
f
o
r
m
a
n
ce
co
m
p
ar
i
s
o
n
o
f
ab
n
o
r
m
a
liti
es
cla
s
s
i
f
ica
tio
n
w
it
h
p
r
o
p
o
s
ed
an
d
ex
is
tin
g
w
o
r
k
P
o
l
y
p
U
l
c
e
r
B
l
e
e
d
i
n
g
Eso
p
h
a
g
i
t
i
s
Z
-
L
i
n
e
C
e
c
u
m
P
y
l
o
r
u
s
C
o
i
mb
r
a
e
t
a
l
.
,
[
3
7
]
.
7
4
.
6
7
7
2
.
7
8
8
7
.
9
N
i
l
N
i
l
N
i
l
N
i
l
H
w
a
n
g
e
t
a
l
.
,
[
3
8
]
.
8
1
.
3
3
7
8
.
3
3
9
0
.
9
5
N
i
l
N
i
l
N
i
l
N
i
l
N
a
w
a
r
a
t
h
n
a
e
t
a
l
.
,
[
3
9
]
.
8
4
.
0
0
8
3
.
3
3
9
3
.
3
3
N
i
l
N
i
l
N
i
l
N
i
l
Y
u
a
n
e
t
a
l
.
,
[
2
2
]
.
8
3
.
5
0
8
0
.
3
3
9
6
.
6
0
N
i
l
N
i
l
N
i
l
N
i
l
P
r
o
p
o
se
d
sy
st
e
m
8
4
.
3
9
8
5
.
9
2
9
6
.8
5
7
9
.
9
1
7
6
.
2
1
8
0
.
2
3
9
0
.
9
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:
2088
-
8708
A
n
efficien
t m
eth
o
d
to
cla
s
s
ify
GI
tr
a
ct
ima
g
es fr
o
m
WC
E
u
s
in
g
visu
a
l wo
r
d
s
(
R
.
P
o
n
n
u
s
a
my
)
5685
7.
CO
NCLU
SI
O
N
C
las
s
i
f
y
i
n
g
t
h
e
ab
n
o
r
m
alitie
s
f
r
o
m
W
C
E
i
m
a
g
es
s
u
ch
as
Z
-
L
i
n
e,
B
leed
in
g
,
P
y
lo
r
u
s
,
C
ec
u
m
,
E
s
o
p
h
ag
iti
s
,
P
o
ly
p
s
,
Ulce
r
ati
v
e
C
o
liti
s
is
a
c
h
alle
n
g
i
n
g
tas
k
.
I
t
m
a
y
ta
k
e
2
h
o
u
r
s
p
er
p
atien
t
f
o
r
r
ev
ie
w
i
n
g
GI
tr
ac
t
d
is
ea
s
es.
I
t
i
s
h
ig
h
l
y
ti
m
e
co
n
s
u
m
i
n
g
an
d
i
n
cr
ea
s
e
s
h
ea
lt
h
ca
r
e
co
s
ts
co
n
s
id
er
ab
l
y
.
T
o
o
v
er
co
m
e
t
h
is
p
r
o
b
lem
,
w
e
p
r
o
p
o
s
ed
Mu
lti
-
A
b
n
o
r
m
alitie
s
clas
s
if
icatio
n
m
o
d
el
u
s
i
n
g
Vi
s
u
al
B
a
g
o
f
W
o
r
d
s
w
h
ic
h
i
s
co
n
s
tr
u
cted
u
s
i
n
g
CS
-
L
B
P
w
ith
SIFT
an
d
co
m
b
i
n
ed
CS
-
L
B
P
,
SIFT
w
i
th
AC
C
u
s
in
g
K
-
m
ea
n
s
cl
u
s
ter
i
n
g
ap
p
r
o
ac
h
.
T
h
e
SVM
is
u
s
ed
f
o
r
clas
s
if
icatio
n
.
B
y
v
ar
y
i
n
g
th
e
co
d
eb
o
o
k
s
ize
f
r
o
m
{2
5
0
,
5
0
0
,
7
5
0
,
1
0
0
0
},
w
e
o
b
tain
ed
7
5
0
as
b
est
s
i
ze
f
o
r
o
u
t
d
ataset
s
.
Fro
m
t
h
e
e
x
p
er
i
m
e
n
t
r
e
s
u
l
ts
,
it
i
s
d
e
m
o
n
s
tr
ated
t
h
at
th
e
ac
c
u
r
ac
y
o
f
7
9
.
9
1
%
f
o
r
E
s
o
p
h
ag
it
is
,
7
6
.
2
1
%
f
o
r
Z
-
L
in
e
,
8
0
.
2
3
%
f
o
r
ce
cu
m
,
8
4
.
3
9
%
f
o
r
P
o
ly
p
s
,
8
5
.
9
2
%
f
o
r
Ulce
r
ati
v
e
C
o
li
tis
,
9
0
.
9
1
%
f
o
r
P
y
lo
r
u
s
an
d
9
4
.
8
0
%
f
o
r
b
leed
in
g
is
o
b
tain
ed
f
o
r
co
m
b
i
n
e
d
CS
-
L
B
P
+SI
FT
+A
C
C
w
it
h
SVM
w
h
ic
h
is
s
i
g
n
if
ica
n
tl
y
h
i
g
h
er
th
a
n
th
e
e
x
is
t
in
g
ap
p
r
o
ac
h
es
an
d
th
e
co
m
b
i
n
atio
n
o
f
C
S
-
L
B
P
an
d
SIFT
.
I
n
f
u
t
u
r
e
,
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
m
a
y
b
e
o
p
ti
m
ize
d
u
s
i
n
g
v
ar
io
u
s
p
ar
a
m
eter
s
to
i
m
p
r
o
v
e
p
er
f
o
r
m
an
ce
.
RE
F
E
R
E
NC
E
S
[
1
]
G
.
I
d
d
a
n
,
G
.
M
e
r
o
n
,
A
.
G
l
u
k
h
o
v
s
k
y
,
a
n
d
P
.
S
w
a
i
n
,
“
W
i
r
e
l
e
s
s
c
a
p
s
u
l
e
e
n
d
o
s
c
o
p
y
,
”
N
a
t
u
r
e
,
v
o
l
.
4
0
5
,
n
o
.
6
7
8
5
,
p
p
.
4
1
7
-
4
1
8
,
2000.
[2
]
D.
K.
Ia
k
o
v
id
is
a
n
d
A
.
Ko
u
lao
u
z
id
is,
“
S
o
f
tw
a
re
f
o
r
e
n
h
a
n
c
e
d
v
i
d
e
o
c
a
p
su
le
e
n
d
o
sc
o
p
y
:
Ch
a
ll
e
n
g
e
s
f
o
r
e
ss
e
n
ti
a
l
p
ro
g
re
ss
,
”
Na
t.
Rev
.
Ga
str
o
e
n
ter
o
l
.
He
p
a
t
o
l.
,
v
o
l.
1
2
,
n
o
.
3
,
p
p
.
1
7
2
-
1
8
6
,
2
0
1
5
.
[
3
]
B
.
U
p
c
h
u
r
c
h
a
n
d
J
.
V
a
r
g
o
,
“
S
m
a
l
l
b
o
w
e
l
e
n
d
o
s
c
o
p
y
,
”
R
e
v
.
G
a
s
t
r
o
e
n
t
e
r
o
l
D
i
s
o
r
d
e
r
s
,
v
o
l
.
8
,
no.
3
,
p
p
.
1
6
9
-
1
7
7
,
2
0
0
7
.
[
4
]
M
a
u
r
o
M
a
n
n
o
,
C
a
r
m
e
l
o
B
a
r
b
e
r
a
,
H
e
l
g
a
B
e
r
t
a
n
i
,
R
a
f
f
a
e
l
e
M
a
n
t
a
,
V
i
n
c
e
n
z
o
G
i
o
r
g
i
o
M
i
y
r
a
n
t
e
D
a
b
i
z
z
i
,
A
n
g
e
l
o
C
a
r
u
s
o
,
F
l
a
v
i
a
P
i
g
o
,
G
i
a
m
p
i
e
r
o
O
l
i
v
e
t
t
i
,
a
n
d
R
i
t
a
C
o
n
i
g
l
i
a
r
o
,
"
S
i
n
g
l
e
-
b
a
l
l
o
o
n
e
n
d
o
s
c
o
p
y
:
T
e
c
h
n
i
c
a
l
a
s
c
e
p
t
s
a
n
d
c
l
i
n
i
c
a
l
a
p
p
l
i
c
a
t
i
o
n
s
"
i
n
W
o
r
l
d
J
o
u
r
n
a
l
o
f
G
a
s
t
r
o
i
n
t
e
s
t
i
n
a
l
E
n
d
o
s
c
o
p
y
,
pp.
28
-
32
,
2
0
1
2
.
[5
]
D.
K.
Ia
k
o
v
id
is
a
n
d
A
.
Ko
u
lao
u
z
id
is,
“
S
o
f
tw
a
re
f
o
r
e
n
h
a
n
c
e
d
v
i
d
e
o
c
a
p
su
le
e
n
d
o
sc
o
p
y
:
Ch
a
ll
e
n
g
e
s
f
o
r
e
ss
e
n
ti
a
l
p
ro
g
re
ss
,
”
Na
tu
re
Rev
.
Ga
str
o
e
n
t
e
ro
l,
v
o
l.
1
2
,
n
o
.
3
,
p
p
.
1
7
2
-
1
8
6
,
2
0
1
5
.
[6
]
Y.
Yu
a
n
,
B.
L
i,
a
n
d
M
.
Q.
-
H.
M
e
n
g
,
“
I
m
p
ro
v
e
d
b
a
g
o
f
f
e
a
tu
re
f
o
r
a
u
to
m
a
ti
c
p
o
l
y
p
d
e
tec
ti
o
n
in
wire
les
s
c
a
p
su
le
e
n
d
o
sc
o
p
y
i
m
a
g
e
s,”
IEE
E
T
ra
n
s.
Au
to
m.
S
c
i.
En
g
.
,
v
o
l.
1
3
,
n
o
.
2
,
p
p
.
5
2
9
-
5
3
5
,
2
0
1
6
.
[7
]
S
a
id
Ch
a
rf
i,
M
o
h
a
m
e
d
El
A
n
sa
ri
,
“
C
o
m
p
u
ter
-
a
id
e
d
d
iag
n
o
sis
sy
st
e
m
f
o
r
c
o
lo
n
a
b
n
o
rm
a
li
ti
e
s
d
e
tec
ti
o
n
in
w
irele
ss
c
a
p
su
le en
d
o
sc
o
p
y
ima
g
e
s,”
M
u
lt
ime
d
ia
T
o
o
ls a
n
d
Ap
p
li
c
a
ti
o
n
s,
p
p
.
1
-
1
8
,
2
0
1
7
[8
]
T
.
G
h
o
sh
,
S
.
A
.
F
a
tt
a
h
,
a
n
d
K.
A
.
Wah
id
,
“
A
u
to
m
a
ti
c
Co
m
p
u
ter
A
id
e
d
Blee
d
in
g
De
tec
ti
o
n
S
c
h
e
m
e
f
o
r
W
irele
ss
Ca
p
su
le
En
d
o
sc
o
p
y
(
W
CE)
V
id
e
o
b
a
se
d
o
n
Hig
h
e
r
a
n
d
L
o
w
e
r
o
r
d
e
r
S
tatisti
c
a
l
F
e
a
tu
re
s
in
a
C
o
m
p
o
site
Co
lo
r,
”
J
o
u
rn
a
l
o
f
M
e
d
ica
l
a
n
d
Bi
o
l
o
g
ica
l
En
g
in
e
e
rin
g,
v
o
l
.
3
8
,
n
o
.
3
,
p
p
.
4
8
2
-
4
9
6
,
2
0
1
8
[9
]
Y.
Yu
a
n
,
M
a
x
Q.
-
H.
M
e
n
g
,
“
D
e
e
p
L
e
a
rn
in
g
f
o
r
P
o
ly
p
Re
c
o
g
n
i
ti
o
n
in
W
irele
ss
Ca
p
su
le
E
n
d
o
sc
o
p
y
I
m
a
g
e
s,”
Ame
ric
a
n
Asso
c
ia
t
io
n
o
f
Ph
y
sic
is
ts i
n
M
e
d
icin
e
,
v
o
l.
4
4
,
n
o
.
4
,
p
p
.
1
3
7
9
-
1
3
8
9
,
2
0
1
7
[1
0
]
M
.
A
li
z
a
d
e
h
,
e
t
a
l.
,
“
De
tec
ti
o
n
o
f
S
m
a
ll
Bo
w
e
l
T
u
m
o
r
in
W
irele
s
s
Ca
p
su
le
En
d
o
sc
o
p
y
i
m
a
g
e
s
u
sin
g
a
n
A
d
a
p
ti
v
e
Ne
u
ro
-
F
u
z
z
y
In
f
e
re
n
c
e
S
y
ste
m
,
”
T
h
e
J
o
u
r
n
a
l
o
f
Bi
o
me
d
ica
l
Res
e
a
rc
h
,
v
o
l.
3
1
,
n
o
.
5
,
p
p
.
4
1
9
-
4
2
7
,
2
0
1
7
[1
1
]
P
.
S
iv
a
k
u
m
a
r,
B.
M
u
th
u
Ku
m
a
r,
“
A
n
o
v
e
l
m
e
th
o
d
to
d
e
tec
t
b
lee
d
in
g
f
ra
m
e
a
n
d
re
g
io
n
in
w
irele
ss
c
a
p
su
le
e
n
d
o
sc
o
p
y
v
id
e
o
,
”
Clu
ste
r Co
mp
u
ti
n
g
,
p
p
.
1
-
7
,
2
0
1
7
[1
2
]
M
e
ry
e
m
S
o
u
a
id
i,
A
b
d
e
lk
a
h
e
r
A
it
A
b
d
e
lo
u
a
h
e
d
,
M
o
h
a
m
e
d
El
A
n
s
a
ri,
“
M
u
lt
i
-
sc
a
le
c
o
m
p
lete
d
lo
c
a
l
b
in
a
ry
p
a
tt
e
rn
s
f
o
r
u
lce
r
d
e
tec
ti
o
n
in
w
irele
ss
c
a
p
su
le en
d
o
sc
o
p
y
ima
g
e
s,”
M
u
lt
ime
d
ia
T
o
o
ls a
n
d
Ap
p
li
c
a
ti
o
n
s
,
p
p
.
1
-
1
8
,
2
0
1
8
[1
3
]
M
ich
a
e
l
D.
V
a
silak
a
k
is,
e
t
a
l.
,
“
Wea
k
l
y
su
p
e
rv
ise
d
m
u
lt
i
-
lab
e
l
c
las
si
f
ica
ti
o
n
f
o
r
se
m
a
n
ti
c
i
n
terp
re
tatio
n
o
f
e
n
d
o
sc
o
p
y
v
id
e
o
f
ra
m
e
s,”
Evo
lvin
g
S
y
ste
ms
,
p
p
.
1
-
1
3
,
2
0
1
8
[
1
4
]
Y
.
Y
a
n
a
g
a
w
a
,
e
t
a
l
.
,
“
A
b
n
o
r
m
a
l
i
t
y
t
r
a
c
k
i
n
g
d
u
r
i
n
g
v
i
d
e
o
c
a
p
s
u
l
e
e
n
d
o
s
c
o
p
y
u
s
i
n
g
a
n
a
f
f
i
n
e
t
r
i
a
n
g
u
l
a
r
c
o
n
s
t
r
a
i
n
t
b
a
s
e
d
o
n
s
u
r
r
o
u
n
d
i
n
g
f
e
a
t
u
r
e
s
,
”
I
P
S
J
T
r
a
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
1
-
1
0
,
2
0
1
7
[1
5
]
Dim
it
ris
K.
Ia
k
o
v
id
is,
e
t
a
l.
,
“
De
e
p
En
d
o
sc
o
p
ic
V
isu
a
l
M
e
a
su
re
m
e
n
ts,”
IEE
E
J
o
u
r
n
a
l
o
f
Bi
o
me
d
ic
a
l
a
n
d
He
a
lt
h
In
fo
rm
a
t
ics
,
p
p
.
1
-
9
,
2
0
1
8
[1
6
]
L
i,
B.
a
n
d
M
e
n
g
,
M
.
Q.H
.
,
"
Co
m
p
u
ter
a
id
e
d
d
e
tec
ti
o
n
o
f
b
le
e
d
in
g
re
g
io
n
s
f
o
r
c
a
p
su
le
e
n
d
o
sc
o
p
y
i
m
a
g
e
s,
"
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Bi
o
me
d
ica
l
En
g
in
e
e
rin
g
,
v
o
l
.
5
6
,
n
o
.
4
,
p
p
.
1
0
3
2
-
1
0
3
9
,
2
0
0
9
.
[1
7
]
A
.
Ka
ra
r
g
y
ri
s
a
n
d
N.
G
.
Bo
u
rb
a
k
is,
“
De
t
e
c
ti
o
n
o
f
s
m
a
ll
b
o
w
e
l
p
o
ly
p
s
a
n
d
u
lce
rs
in
w
irele
ss
c
a
p
su
le
e
n
d
o
sc
o
p
y
v
id
e
o
s,”
IEE
E
T
r
a
n
s.
Bi
o
me
d
.
E
n
g
.
,
v
o
l
.
5
8
,
n
o
.
1
0
,
p
p
.
2
7
7
7
-
2
7
8
6
,
2
0
1
1
.
[1
8
]
T
a
jb
a
k
h
sh
N,
G
u
ru
d
u
S
.
R
,
L
ian
g
J.,
“
A
u
to
m
a
ti
c
p
o
ly
p
d
e
tec
ti
o
n
u
sin
g
g
lo
b
a
l
g
e
o
m
e
tri
c
c
o
n
stra
in
ts
a
n
d
lo
c
a
l
in
ten
sity
v
a
ri
a
ti
o
n
p
a
tt
e
rn
s,”
17
t
h
In
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
me
d
ica
l
ima
g
e
c
o
mp
u
ti
n
g
a
n
d
c
o
mp
u
ter
-
a
ss
isted
in
ter
v
e
n
ti
o
n
,
pp.
1
7
9
-
1
8
7
,
2
0
1
4
.
[1
9
]
Htw
e
T
.
M
,
e
t
a
l.
,
“
A
d
a
b
o
o
st
lea
rn
in
g
f
o
r
sm
a
ll
u
lce
r
d
e
tec
ti
o
n
f
r
o
m
w
irele
s
s
c
a
p
su
le
e
n
d
o
sc
o
p
y
(
W
CE
)
i
m
a
g
e
s,”
Asia
P
a
c
if
ic sig
n
a
l
a
n
d
in
f
o
rm
a
ti
o
n
p
ro
c
e
ss
in
g
a
ss
o
c
ia
ti
o
n
(
AP
S
IP
A),
p
p
.
6
5
3
-
6
5
6
,
2
0
1
0
.
[2
0
]
H.
B.
Ba
h
a
r
,
e
t
a
l.
,
“
A
d
a
p
ted
b
it
-
pl
a
n
e
p
ro
b
a
b
il
i
ty
a
n
d
w
a
v
e
let
-
b
a
se
d
u
lce
r
d
e
tec
ti
o
n
i
n
w
irele
ss
c
a
p
su
le
e
n
d
o
sc
o
p
y
i
m
a
g
e
s,”
J
o
u
rn
a
l
o
f
B
io
me
d
ica
l
E
n
g
in
e
e
rin
g
:
Ap
p
li
c
a
t
i
o
n
s,
B
a
sis
a
n
d
C
o
mm
u
n
ic
a
ti
o
n
s
,
v
o
l.
2
8
,
n
o
.
4
,
p
p
.
1
6
5
0
0
2
9
-
1
-
1
0
,
2
0
1
6
.
[2
1
]
Ho
g
h
a
n
.
C
h
e
n
,
J.
Ch
e
n
,
Q.
P
e
n
g
,
G
.
S
u
n
,
a
n
d
T
.
G
a
n
,
“
A
u
to
m
a
ti
c
h
o
o
k
w
o
r
m
i
m
a
g
e
d
e
tec
ti
o
n
f
o
r
w
irele
ss
c
a
p
su
le
e
n
d
o
sc
o
p
y
u
sin
g
h
y
b
rid
c
o
l
o
r
g
ra
d
ien
t
a
n
d
c
o
n
t
o
u
rlet
tran
sf
o
rm
,
”
6
th
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
b
i
o
me
d
ica
l
e
n
g
in
e
e
rin
g
a
n
d
in
fo
rm
a
t
ics
Bi
o
me
d
ica
l
En
g
i
n
e
e
rin
g
a
n
d
I
n
fo
rm
a
ti
c
s (
BM
EI)
,
p
p
.
1
1
6
-
1
2
0
,
2
0
1
3
.
[2
2
]
Y.
Yu
a
n
,
Ba
o
p
u
L
i,
a
n
d
M
a
x
Q.
-
H.
M
e
n
g
,
“
W
CE
a
b
n
o
rm
a
li
t
y
d
e
tec
ti
o
n
b
a
se
d
o
n
sa
li
e
n
c
y
a
n
d
a
d
a
p
ti
v
e
lo
c
a
li
ty
-
c
o
n
stra
in
e
d
li
n
e
a
r
c
o
d
in
g
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Au
to
ma
ti
o
n
S
c
ie
n
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
,
p
p
.
1
4
9
-
1
5
9
,
2
0
1
5
.
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
.
6
,
Dec
em
b
er
2
0
2
0
:
5
6
7
8
-
5
6
8
6
5686
[2
3
]
D.T
.
Oja
la,
M
.
P
ietik
in
e
n
,
a
n
d
T
.
M
a
e
n
p
a
a
,
“
M
u
lt
ires
o
l
u
ti
o
n
g
ra
y
sc
a
le
a
n
d
ro
tatio
n
i
n
v
a
rian
t
tex
tu
r
e
c
las
si
f
ica
ti
o
n
w
it
h
lo
c
a
l
b
i
n
a
ry
p
a
tt
e
rn
s,”
IEE
E
T
ra
n
s
o
n
PA
M
I,
v
o
l.
2
4
,
n
o
.
7
,
p
p
.
9
7
1
-
9
8
7
,
2
0
0
2
.
[2
4
]
W
a
n
g
,
X
.
,
Ha
n
,
T
.
X
.
,
a
n
d
Ya
n
,
S
,
“
A
n
HO
G
-
L
BP
h
u
m
a
n
d
e
tec
to
r
w
it
h
p
a
rti
a
l
o
c
c
lu
sio
n
h
a
n
d
l
i
n
g
,
”
IEE
E
1
2
t
h
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
si
o
n
,
p
p
.
3
2
-
3
9
,
2
0
0
9
.
[2
5
]
Un
a
y
,
D.,
a
n
d
Ek
in
,
A
,
“
In
ten
sity
v
e
ru
s
te
x
tu
re
f
o
r
m
e
d
ica
l
i
m
a
g
e
se
a
rc
h
a
n
d
re
tri
e
v
a
l,
”
5
th
IEE
E
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m o
n
Bi
o
me
d
ica
l
Ima
g
i
n
g
:
Fr
o
m Na
n
o
to
M
a
c
ro
,
p
p
.
2
4
1
-
2
4
4
,
2
0
0
8
.
[2
6
]
Oliv
e
r,
A
.
,
L
lad
o
,
X.,
F
re
ix
e
n
e
t,
J.,
a
n
d
M
a
rta,
J.,
“
F
a
lse
p
o
siti
v
e
re
d
u
c
ti
o
n
in
m
a
m
m
o
g
ra
p
h
ic
m
a
ss
d
e
tec
ti
o
n
u
sin
g
lo
c
a
l
b
in
a
ry
p
a
tt
e
rn
s,”
M
e
d
ica
l
Ima
g
e
Co
mp
u
ti
n
g
a
n
d
Co
mp
u
ter
-
Assist
e
d
In
ter
v
e
n
ti
o
n
–
M
ICCAI
2
0
0
7
,
p
p
.
2
8
6
-
2
9
3
,
2
0
0
7
.
[2
7
]
M
a
rg
h
a
n
i,
K.
A
.
,
Dla
y
,
S
.
S
.
,
S
h
a
rif
,
B.
S
.
,
a
n
d
S
im
s,
A
.
J.,
“
M
o
r
p
h
o
lo
g
ica
l
a
n
d
tex
tu
re
f
e
a
tu
re
s
f
o
r
c
a
n
c
e
r
ti
ss
u
e
s
m
icro
sc
o
p
ic i
m
a
g
e
s,”
5
th
IEE
E
I
n
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m o
n
M
e
d
i
c
a
l
Ima
g
in
g
,
p
p
.
1
7
5
7
-
1
7
6
4
,
2
0
0
3
.
[2
8
]
He
ik
k
il
ä
,
M
.
,
P
ietik
ä
in
e
n
,
M
.
,
a
n
d
S
c
h
m
id
,
C.
,
“
De
sc
rip
ti
o
n
o
f
in
t
e
re
st
re
g
io
n
s
w
it
h
lo
c
a
l
b
i
n
a
ry
p
a
tt
e
rn
s,”
Pa
t
ter
n
Rec
o
g
n
it
io
n
,
v
o
l.
4
2
,
n
o
.
3
,
p
p
.
4
2
5
-
4
3
6
,
2
0
0
9
.
[2
9
]
Ha
n
a
n
e
.
Ra
m
i,
M
o
h
a
m
m
e
d
.
Ha
m
ri
a
n
d
L
h
o
u
c
in
e
.
M
a
sm
o
u
d
i,
“
Ob
jec
ts
T
r
a
c
k
in
g
in
I
m
a
g
e
s
S
e
q
u
e
n
c
e
Us
in
g
Ce
n
ter
-
S
y
m
m
e
tri
c
L
o
c
a
l
Bin
a
r
y
P
a
tt
e
r
n
(CS
-
L
BP
)
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
Ap
p
li
c
a
ti
o
n
s
T
e
c
h
n
o
lo
g
y
a
n
d
Res
e
a
rc
h
,
v
o
l
.
2
,
n
o
.
5
,
p
p
.
5
0
4
-
5
0
8
,
2
0
1
3
.
[3
0
]
K.
N
irm
a
la
a
n
d
A
.
S
u
b
ra
m
a
n
i,
"
Co
n
ten
t
Ba
se
d
Im
a
g
e
Re
tri
e
v
a
l
S
y
s
tem
U
sin
g
A
u
to
Co
lo
r
C
o
rre
lo
g
ra
m
,
"
J
o
u
rn
a
l
o
f
Co
mp
u
ter
A
p
p
li
c
a
ti
o
n
s (
J
CA),
v
o
l.
VI,
n
o
.
4
,
p
p
.
1
1
1
-
1
1
5
,
2
0
1
3
.
[3
1
]
Nid
h
i
S
in
g
h
a
i,
K.
S
h
a
n
d
il
y
a
,
“
A
S
u
rv
e
y
On
:
Co
n
ten
t
Ba
se
d
Im
a
g
e
Re
tri
e
v
a
l
S
y
ste
m
s,”
In
te
rn
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Co
mp
u
ter
A
p
p
li
c
a
ti
o
n
s,
v
o
l.
4
,
n
o
.
2
,
p
p
.
2
2
-
2
6
,
2
0
1
0
.
[3
2
]
C.
H.
L
in
,
R.
T
.
Ch
e
n
a
n
d
Y.
K.
Ch
a
n
,
“
A
s
m
a
rt
c
o
n
ten
t
-
b
a
se
d
ima
g
e
re
tri
e
v
a
l
s
y
ste
m
b
a
se
d
o
n
c
o
lo
r
a
n
d
tex
tu
re
f
e
a
tu
re
,
”
Ima
g
e
a
n
d
Vi
sio
n
C
o
mp
u
ti
n
g
,
v
o
l
.
2
7
,
n
o
.
6
,
p
p
.
6
5
8
-
6
6
5
,
2
0
0
9
.
[3
3
]
K.
S
e
e
th
a
ra
m
a
n
,
S
.
S
a
th
iam
o
o
rth
y
,
“
A
n
i
m
p
ro
v
e
d
e
d
g
e
d
irec
ti
o
n
h
isto
g
ra
m
a
n
d
e
d
g
e
o
rien
tati
o
n
a
u
to
c
o
rre
lo
g
ra
m
f
o
r
a
n
e
ff
ici
e
n
t
c
o
lo
r
im
a
g
e
r
e
tri
e
v
a
l,
”
2
0
1
3
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
Co
mp
u
ti
n
g
&
Co
mm
u
n
ica
ti
o
n
S
y
ste
ms
(
ICACCS
)
,
p
p
.
1
9
-
2
1
,
2
0
1
3.
[3
4
]
A
d
a
m
o
,
F
.
,
Ca
rc
a
g
n
i,
P
.
,
M
a
z
z
e
o
,
P
.
L
.
,
Dista
n
te,
C.
,
a
n
d
S
p
a
g
n
o
lo
,
P
.
,
“
T
LD
a
n
d
S
tru
c
k
:
A
F
e
a
tu
re
De
sc
rip
to
rs
C
o
m
p
a
r
a
t
i
v
e
S
t
u
d
y
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
W
o
r
k
s
h
o
p
o
n
A
c
t
i
v
i
t
y
M
o
n
i
t
o
r
i
n
g
b
y
M
u
l
t
i
p
l
e
D
i
s
t
r
i
b
u
t
e
d
S
e
n
s
i
n
g
,
p
p
.
5
2
-
6
3
,
2
0
1
4
.
[3
5
]
R.
V
isa
lak
sh
i,
R.
P
o
n
n
u
sa
m
y
,
a
n
d
K.
M
a
n
ik
a
n
d
a
n
,
“
L
it
e
ra
tu
re
S
u
rv
e
y
o
f
Da
ta
M
in
in
g
Clu
ste
rin
g
A
lg
o
rit
h
m
s,”
S
o
u
th
Asia
n
J
o
u
r
n
a
l
o
f
Res
e
a
rc
h
in
E
n
g
i
n
e
e
rin
g
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
1
,
p
p
.
3
1
0
-
3
1
3
,
2
0
1
6
.
[3
6
]
B.
T
rig
g
s
a
n
d
F
.
Ju
rie,
“
Cre
a
ti
n
g
e
ff
ici
e
n
t
c
o
d
e
b
o
o
k
s
f
o
r
v
isu
a
l
re
c
o
g
n
it
io
n
,
”
T
e
n
th
IEE
E
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
sio
n
(
ICCV'0
5
)
,
v
o
l.
1
,
p
p
.
6
0
4
-
6
1
0
,
2
0
0
5
.
[3
7
]
Q.
T
ian
,
a
n
d
S
.
Zh
a
n
g
,
“
De
sc
rip
ti
v
e
v
isu
a
l
w
o
rd
s
a
n
d
v
isu
a
l
p
h
r
a
se
s
f
o
r
im
a
g
e
a
p
p
li
c
a
ti
o
n
s,”
AC
M
M
u
lt
ime
d
ia
,
p
p
.
1
9
-
2
4
,
2
0
0
9
.
[3
8
]
A
.
S
treic
h
e
r,
H.
Bu
rk
h
a
rd
t,
a
n
d
J.
F
e
h
r,
“
A
b
a
g
o
f
f
e
a
tu
re
s
a
p
p
r
o
a
c
h
f
o
r
3
D
sh
a
p
e
re
tri
e
v
a
l,
”
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m o
n
Vi
su
a
l
C
o
mp
u
ti
n
g
,
IS
V
C
2
0
0
9
,
L
a
s V
e
g
a
s,
NV
,
US
A
,
p
p
.
3
4
-
4
3
,
2
0
0
9
.
[3
9
]
R.
P
o
n
n
u
sa
m
y
,
S
.
S
a
th
iam
o
o
rth
y
,
“
A
n
Eff
i
c
ien
t
G
a
stro
in
tes
ti
n
a
l
He
m
o
rrh
a
g
e
D
e
tec
ti
o
n
a
n
d
Dia
g
n
o
sis
M
o
d
e
l
f
o
r
W
irele
ss
C
a
p
su
le
En
d
o
sc
o
p
y
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
Rec
e
n
t
T
e
c
h
n
o
l
o
g
y
a
n
d
En
g
in
e
e
rin
g
(
IJ
RT
E),
v
o
l.
8
n
o
.
3
,
p
p
.
7
5
4
9
-
7
5
5
4
,
2
0
1
9
.
[4
0
]
P
.
S
a
leh
p
o
u
r,
H.
B.
Ba
h
a
r,
G
.
Ka
ri
m
ian
a
n
d
M
.
T
,
Co
im
b
ra
a
n
d
J.
P
.
S
.
Cu
n
h
a
,
“
M
P
EG
-
7
v
i
su
a
l
d
e
sc
rip
to
r
-
c
o
n
tri
b
u
ti
o
n
s
f
o
r
a
u
to
m
a
ted
f
e
a
tu
re
e
x
trac
ti
o
n
in
c
a
p
su
le
e
n
d
o
sc
o
p
y
,
”
IEE
E
T
ra
n
s.
Ciru
c
it
s
S
y
st.
Vi
d
e
o
T
e
c
h
n
o
l.
,
v
o
l.
1
6
,
n
o
.
5
,
p
p
.
6
2
8
-
6
3
7
,
2
0
0
6
.
[4
1
]
S
.
Hw
a
n
g
,
“
B
a
g
o
f
v
isu
a
l
w
o
rd
s
a
p
p
ro
a
c
h
t
o
a
b
n
o
rm
a
l
i
m
a
g
e
d
e
tec
ti
o
n
in
w
irele
ss
c
a
p
su
le
e
n
d
o
sc
o
p
y
v
id
e
o
s,”
7
th
I
n
ter
n
a
ti
o
n
a
l
S
y
mp
o
siu
m
A
d
v
a
n
c
e
s
in
Vi
su
a
l
Co
m
p
u
ti
n
g
IS
VC
2
0
1
1
,
L
a
s
Veg
a
s,
NV,
US
A,
P
a
rt
II,
p
p
.
3
2
0
-
3
2
7
,
2
0
1
1
.
[4
2
]
R.
Na
w
a
r
a
th
n
a
e
t
a
l.
,
“
A
b
n
o
rm
a
l
im
a
g
e
d
e
tec
ti
o
n
in
e
n
d
o
sc
o
p
y
v
id
e
o
s
u
sin
g
f
il
ter
b
a
n
k
a
n
d
l
o
c
a
l
b
i
n
a
ry
p
a
tt
e
rn
s,”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
1
4
4
,
p
p
.
70
-
91,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.