I
S
S
N
:
1693
-6
930
T
era
k
r
edit
a
s
i
D
I
K
T
I
,
S
K
N
o:
51/
D
I
K
T
I
/
K
ep/
2010
175
A
U
nif
ied
I
mage
E
ner
gy
A
p
proa
c
h
f
o
r
S
egment
at
ion
us
ing
B
S
plin
e
S
nak
e
(A
gun
g
A
lf
ians
y
ah
)
A
UNI
F
I
E
D
E
NE
RG
Y
AP
P
RO
AC
H
F
O
R
B
-S
P
L
I
NE
S
NAK
E
I
N
M
E
DI
CAL
I
M
AG
E
S
E
G
M
E
NT
AT
I
O
N
A
g
u
n
g
A
l
fi
a
n
s
y
ah
D
ept
.
of
E
lec
t
ric
al
E
n
ginee
r
ing
,
F
ac
ult
y
o
f
I
ndus
t
rial
E
n
ginee
ring,
I
nd
one
s
ia
I
s
lam
i
c
U
niv
e
rs
it
y
K
am
pu
s
T
er
p
adu
U
I
I
,
J
ala
n
K
aliura
ng
,
K
M
14.
5,
Y
ogy
ak
art
a,
I
ndo
nes
i
a.
e-m
ail:
ag
ung
.
alf
ians
y
ah
@
gm
ail.
c
o
m
A
b
s
tr
ak
M
odel
d
ef
or
mabel
par
amet
rik
ban
y
a
k
dipilih
s
e
ba
gai
pe
nde
k
at
an
u
nt
uk
me
lak
u
k
a
n
ek
s
t
ra
s
i
obj
e
k
da
ri
c
it
r
a
k
aren
a
ala
s
a
n
k
e
s
ed
erh
an
aan
da
n
ef
f
e
s
ien
s
i.
N
amun
met
ode
ini
juga
memilik
i
beberap
a
k
et
er
b
at
as
a
n.
T
uli
s
an
ini
membaha
s
menge
nai
s
at
u
t
ipe
model
def
or
mabel
y
an
g
di
ny
at
a
k
an
s
ec
ar
a
ek
s
pli
s
it
de
n
gan
k
u
r
v
a
B
S
pline
u
nt
uk
k
e
pe
rluan
s
egment
as
i
c
i
t
ra.
S
elain
membaha
s
b
ebe
ra
pa
h
al
y
an
g
membat
as
i
k
et
er
bat
a
s
an
model
d
ef
or
mabel
e
k
s
pli
s
it
.
P
aper
ini
ju
ga
menaw
a
r
k
an
b
ebe
rap
a
s
olu
s
i
ef
is
en
unt
u
k
mengat
a
s
in
y
a.
M
et
oda
y
a
ng
dik
emban
gk
a
n
t
eri
ns
pira
s
i
da
ri
model
k
la
s
ik
y
ang
dit
aw
a
rk
an
o
leh
K
a
s
s
d
e
ngan
be
ber
a
p
a
adapt
a
s
i
pad
a
apli
k
a
s
i
k
u
rv
a
p
aramet
ri
k
.
T
uli
s
a
n
ini
juga
menaw
ark
an
s
at
u
d
ef
inis
i
bar
u
d
ar
i
t
erm
energi
y
ang
dit
ur
un
k
a
n
dari
c
it
r
a
u
nt
uk
mengga
bung
k
a
n
ene
rgi
ber
ba
s
is
t
epi
dan
w
il
ay
ah
agar
menin
g
k
at
k
an
unj
uk
k
erja
mode
l
def
ormable
ini.
T
ujuan
dik
emban
gk
a
nny
a
met
oda
ini
adala
h
mem
bant
u
par
a
d
ok
t
er
melak
u
k
an
e
k
s
t
r
ak
s
i
org
an
anat
o
mik
dari
c
it
r
a
medis
s
e
c
ar
a
ot
omat
is
,
dim
ana
h
al
ini
s
a
ngat
s
ulit
dila
k
u
k
an
s
e
c
a
ra
manual.
S
es
udah
pr
os
es
s
egment
a
s
i
i
ni,
orga
n
anat
o
mik
pas
ien
b
is
a
diu
k
u
r
da
n
dianali
s
is
l
ebih
lanjut
u
nt
uk
menget
ahui
uk
ur
an
dan
anomali
be
nt
uk
y
a
ng
ada
di
dal
am
org
a
n
t
er
s
e
but
.
H
as
il
pen
elit
ian
menunju
k
k
a
n
ba
hw
a
met
o
de
y
an
g
di
us
ulk
an
t
ela
h
t
e
rb
uk
t
i
s
e
c
a
ra
k
ualit
at
if
be
rha
s
il
p
ada
s
eg
ment
as
i
b
ebe
rapa
c
it
ra
medis
y
an
g
ber
bed
a
.
K
at
a
ku
n
ci
:
B
s
pline
S
n
ak
e,
def
ormabel
,
energi
t
epi,
ener
gi
w
ila
y
a
h,
s
egment
as
i
A
b
s
tr
ac
t
T
he
pa
ramet
ric
s
na
k
e
i
s
one
of
t
he
p
r
ef
err
ed
ap
pr
oac
h
e
s
in
f
e
at
ure
e
x
t
r
ac
t
i
on
f
rom
images
b
ec
a
us
e
of
t
hei
r
s
implic
it
y
an
d
ef
f
ic
ien
c
y
.
H
o
w
e
v
er
t
he
met
hod
has
al
s
o
limit
at
ions
.
I
n
t
his
p
ape
r
a
n
ex
plic
it
s
na
k
e
t
hat
r
ep
re
s
ent
ed
us
i
ng
B
S
pline
ap
pli
ed
f
or
image
s
egment
at
ion
is
c
on
s
id
ere
d.
I
n
t
his
pa
per,
w
e
i
dent
if
y
s
o
me
of
t
hes
e
probl
ems
a
nd
pr
opo
s
e
ef
f
ic
ient
s
olut
ion
s
t
o
get
ar
oun
d
t
hem.
T
he
pr
opo
s
ed
met
h
od
is
in
s
pir
e
d
by
c
l
as
s
i
c
a
l
s
na
k
e
f
rom
K
as
s
w
it
h
s
ome
adapt
io
n
f
or
paramet
ri
c
c
u
rv
e.
T
he
pap
er
al
s
o
pro
po
s
e
s
ne
w
d
ef
init
ions
of
ene
rgy
t
erms
in
t
he
model
t
o
brin
g
t
he
s
na
k
e
perf
o
rmanc
e
more
rob
us
t
and
ef
f
ic
i
ent
f
or
image
s
eg
ment
at
ion.
T
his
ener
gy
t
erm
unif
y
t
he
edg
e
bas
e
d
and
regio
n
ba
s
ed
energ
y
de
riv
ed
f
rom
t
he
i
mage
dat
a.
T
he
main
objec
t
i
v
e
of
de
v
elo
pe
d
w
o
rk
is
t
o
d
ev
el
op
a
n
aut
omat
ic
met
hod
t
o
s
e
gment
t
he
anat
om
ic
al
orga
ns
f
rom
medic
al
imag
es
w
hi
c
h
is
v
ery
h
ard
an
d
t
ediou
s
t
o
be
perf
o
rmed
manuall
y
.
A
f
t
er
t
his
s
egment
at
ion
,
t
he
anat
omic
al
o
bjec
t
c
a
n
be
f
urt
h
er
meas
ur
ed
an
d
anal
y
z
ed
t
o
diagn
os
e
t
he
anomaly
i
n
t
hat
or
gan.
T
he
r
es
ult
s
ha
v
e
s
h
ow
n
t
h
at
t
he
p
rop
o
s
ed
met
hod
has
bee
n
p
r
ov
e
n
qualit
at
iv
ely
s
uc
c
e
s
s
f
ul
in
s
egment
ing
dif
f
erent
t
y
p
es
o
f
medic
al
ima
ges
.
K
ey
w
o
r
d
s
:
B
s
pline
S
n
ak
e,
def
ormable
,
edge
ene
rg
y
,
region
en
er
gy
,
s
e
gment
at
ion
1.
I
N
T
R
O
D
U
C
T
I
O
N
S
egm
ent
at
io
n
is
a
pa
rt
it
ioning
pr
oc
e
s
s
of
an
im
age
d
om
ai
n
int
o
no
n
-ov
erlap
ping
c
on
ne
c
t
ed
r
egion
s
t
hat
c
o
rr
es
pond
t
o
s
i
gnif
ic
ant
an
at
om
i
c
al
s
t
ru
c
t
u
r
es
.
A
ut
om
at
ed
s
egm
ent
at
ion
of
m
edic
al
i
m
age
s
is
a
d
if
f
ic
ult
t
as
k
.
I
m
age
s
ar
e
of
t
en
nois
y
an
d
us
ually
c
ont
ain
m
ore
t
han
a
s
ingl
e
an
at
o
m
ic
al
s
t
r
uc
t
u
r
e
w
it
h
na
rro
w
di
s
t
an
c
e
s
b
et
w
e
en
o
rg
an
b
ound
ari
e
s
.
I
n
addit
ion
t
h
e
o
rgan
bo
und
ar
ies
m
ay
be
di
f
f
us
e.
A
lt
hou
gh
m
e
dic
al
i
m
age
s
e
gm
e
nt
at
ion
ha
s
b
een
an
a
c
t
iv
e
f
iel
d
of
r
es
ea
rc
h
f
or
s
ev
e
ral
d
ec
a
de
s
,
t
her
e
is
no
aut
om
a
t
ic
pr
oc
es
s
c
an
b
e
a
pplied
t
o
all
im
aging
m
odalit
ie
s
and
anat
om
i
c
al
s
t
ruc
t
u
re
s
[
1]
,
[
2]
.
T
he
role
o
f
aut
om
at
ic
s
egm
ent
at
io
n
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
16
93
-6
930
T
E
L
K
O
M
N
I
K
A
V
ol.
8,
N
o.
2,
A
gus
t
us
2
0
10
:
175
–
1
86
176
really
c
rit
ic
al
i
n
c
om
p
ut
er
a
s
s
i
s
t
ed
di
agn
os
t
ic
,
s
i
nc
e
it
help
s
t
he
c
lini
c
ian
s
an
d
do
c
t
or
s
ex
t
ra
c
t
t
he
dif
f
erent
anat
om
ic
al
o
rg
an
f
orm
m
edic
a
l
im
ages
.
T
hi
s
s
e
gm
ent
at
i
on
t
as
k
is
v
e
ry
dif
f
ic
ult
t
o
be
perf
o
rm
ed
m
anually
d
ue
t
o
int
ra
-
an
d
i
nt
er
-
ope
rat
o
r
s
egm
e
nt
at
ion
r
es
ult
s
af
t
er
s
egm
ent
at
ion.
F
urt
he
rm
o
re
i
t
is
t
edious
a
nd
t
im
e
c
on
s
um
ing
f
or
t
he
oper
at
or.
I
n
gen
eral,
s
egm
ent
at
io
n
t
ec
hni
que
s
c
an
b
e
c
la
s
s
if
ied
in
t
w
o
m
ain
c
at
eg
or
ies
:
(
a)
s
egm
ent
at
ion
m
et
ho
d
s
t
h
at
allow
u
s
e
rs
t
o
ex
pli
c
it
ly
s
pe
c
if
y
t
he
de
s
ir
ed
f
e
at
ure,
a
nd
(
b)
algorit
h
m
s
w
here
t
he
s
p
ec
if
ic
at
ion
is
im
plic
it
.
T
he
f
irs
t
s
eg
m
ent
at
ion
c
l
as
s
c
o
ns
i
der
s
t
he
s
egm
ent
at
ion
as
a
real
-t
im
e
int
era
c
t
ion
p
r
oc
e
s
s
b
et
w
e
e
n
t
he
us
er
an
d
t
he
algo
rit
h
m
.
T
he
us
er
i
s
prov
ide
d
w
it
h
t
he
out
put
and
allow
ed
t
o
f
eed
it
bac
k
dire
c
t
ly
in
ord
er
t
o
m
odif
y
t
he
s
e
gm
ent
at
ion
unt
il
he
get
s
a
s
at
is
f
a
c
t
ory
res
ult
.
I
n
t
he
ex
t
rem
e
c
a
s
e,
t
his
f
ram
e
w
or
k
m
ig
ht
be
degen
er
at
e
d
t
o
be
a
m
anu
al
s
egm
ent
at
ion
w
it
h
t
he
us
e
r
f
orc
ing
t
he
re
s
ult
s
h
e
w
a
nt
s
.
W
e
pr
opo
s
e
our
c
ont
rib
ut
ion
in
t
his
pa
per,
a
n
ew
a
ppro
ac
h
f
o
r
B
S
pline
b
as
e
d
ex
t
er
nal
ener
gy
t
hat
u
nif
y
t
he
c
la
s
s
i
c
al,
w
hic
h
is
im
age
gra
die
nt
and
regi
on
bas
ed
e
ne
rg
y
.
I
m
age
ba
s
ed
ener
gy
hel
ps
our
def
o
rm
ab
le
m
od
el
pl
ac
e
t
he
f
in
al
c
o
nt
our
in
t
h
e
d
es
ir
ed
obje
c
t
c
or
re
c
t
ly
,
w
hil
e
regio
n
ba
s
ed
redu
c
e
t
he
m
odel
s
e
ns
it
iv
it
y
t
o
t
he
init
iali
z
at
ion
w
hic
h
is
a
real
p
robl
em
in
c
la
s
s
i
c
al
def
orm
able
m
odel.
W
e
al
s
o
pro
po
s
e
a
s
im
pl
e
u
nif
ic
at
ion
s
c
hem
e
w
hi
c
h
c
a
n
b
e
do
ne
int
uit
i
v
ely
t
o
perf
or
m
t
his
ene
rgy
c
o
m
binat
ion
.
T
his
pa
per
w
ill
be
or
gani
z
ed
a
s
f
oll
ow
s
,
in
s
ec
t
io
n
2
;
w
e
rev
ie
w
t
he
m
ain
c
o
n
c
ept
of
def
orm
able
m
odel
a
nd
it
s
a
pplic
at
ion
o
n
im
age
s
e
gm
e
nt
at
ion.
T
hi
s
pre
s
ent
at
io
n
aim
ed
t
o
giv
e
a
gene
ral
de
s
c
ript
ion
t
o
r
ea
der
a
m
et
hod
t
hat
w
e
f
ollo
w
t
o
dev
elop
our
ap
pr
oa
c
h.
S
pec
if
ic
all
y
in
s
e
c
t
ion
3,
w
e
pr
es
ent
in
d
et
ail
ou
r
pro
p
os
e
d
B
S
plin
e
S
na
k
e;
a
t
y
pe
of
def
orm
a
ble
m
odel
w
h
ic
h
rep
re
s
ent
it
s
c
o
nt
our
ex
plic
it
ly
u
s
ing
par
am
et
ri
c
c
ont
o
ur.
S
e
c
t
ion
4
w
ill
b
e
de
dic
at
ed
t
o
dem
on
s
t
r
at
e
t
he
pe
rf
or
m
an
c
e
of
t
h
e
m
od
el
in
dif
f
er
en
c
e
pa
ram
et
e
r
s
and
d
at
a
t
y
p
e.
T
hen,
f
inall
y
w
e
dr
aw
t
he
c
on
c
lu
s
io
n
ob
t
ained
f
rom
t
h
is
w
o
rk
in
s
e
c
t
ion
5.
2.
B
A
S
I
C
C
O
N
C
E
P
T
O
F
S
N
A
K
E
T
he
ba
s
i
c
id
ea
s
egm
e
nt
a
t
ion
us
i
ng
s
n
ak
e
i
s
t
o
em
bed
a
n
init
ial
c
ont
o
ur
(or
s
urf
a
c
e
i
n
t
he
t
hree
di
m
ens
i
onal
c
a
s
e)
int
o
t
he
i
m
age,
and
t
hen
let
it
ev
olv
e
w
hile
s
ubje
c
t
t
o
v
arious
c
on
s
t
rai
nt
s
re
lat
ed
t
o
t
he
i
m
age
an
d
c
o
nt
our
it
s
elf
.
I
n
orde
r
t
o
det
ec
t
objec
t
s
in
t
hat
im
age,
t
he
c
ont
o
ur
h
as
t
o
s
t
op
it
s
ev
olut
ion
o
n
t
he
bo
und
ary
of
t
he
o
bjec
t
of
int
e
res
t
.
T
h
e
im
ag
e
s
egm
ent
at
ion
t
as
k
is
t
he
n
perf
o
rm
ed
a
s
a
m
inim
iz
at
io
n
of
energy
.
A
lt
hough
t
he
t
erm
“
s
n
ak
e
”
init
ially
appea
red
in
t
he
c
la
s
s
i
c
al
w
o
rk
pr
es
e
nt
ed
by
K
as
s
[
3]
in
t
he
lat
e
eight
ies
,
t
he
id
ea
of
def
orm
i
ng
a
t
em
plat
e
f
or
ex
t
rac
t
i
ng
im
age
f
ea
t
ures
d
at
e
d
b
ac
k
m
uc
h
f
art
h
er,
w
it
h
t
he
w
or
k
of
F
i
s
c
hle
r
[
4]
w
ho
pr
op
os
e
d
s
pri
ng
-l
oade
d
t
em
pl
a
t
es
,
an
d
W
i
dr
ow
[
5]
applied
ru
bber
m
a
s
k
t
ec
hni
que.
I
n
im
age
pr
oc
es
s
ing
lit
er
at
ure
s
,
t
his
“s
na
k
e
”
are
als
o
k
n
ow
n
as
:
s
na
k
e
s
,
a
c
t
iv
e
c
ont
ou
r
s
o
r
s
u
rf
ac
es
,
ballo
on
s
,
def
orm
a
ble
m
od
el
or
def
orm
a
ble
c
ont
o
ur
s
or
s
urf
a
c
e
s
.
A
n
ex
t
ens
iv
e
rev
i
ew
of
t
he
c
u
rr
ent
res
ea
rc
h
in
t
his
are
a
c
a
n
be
f
ind
in
[
6]
.
I
n
t
heir
c
la
s
s
ic
al
w
o
rk
,
K
a
s
s
pr
opo
s
e
d
an
ea
rly
k
i
nd
of
s
n
ak
e
w
h
ic
h
rep
re
s
ent
ed
t
he
c
ont
o
ur
u
s
ing
a
num
be
r
of
dis
c
ret
e
p
oint
s
.
T
hen,
it
s
b
ehav
ior
of
t
h
e
m
odel
w
as
f
o
rm
ulat
e
d
as
a
t
ot
al
energy
o
f
a
w
eight
ed
l
inear
c
om
bin
at
ion
of
:
int
er
nal
ene
rg
y
c
a
c
ulat
e
d
f
rom
t
he
c
o
nt
our
t
h
at
im
pos
es
t
he
regul
arit
y
of
t
he
c
urv
e
in
s
egm
ent
at
io
n
;
ex
t
e
rnal
en
e
rgy
t
hat
at
t
r
a
c
t
s
t
h
e
c
ont
o
ur
t
ow
ar
d
t
he
s
i
gnif
ic
a
nt
f
eat
ure
s
in
t
h
e
i
m
age;
a
nd
s
om
e
a
ddit
ion
al
us
er
ene
r
g
ies
c
o
ns
t
rai
nt
s
allow
in
g
ope
r
at
or
t
o
bet
t
er
int
era
c
t
t
o
t
he
m
odel
.
T
hu
s
,
t
he
s
na
k
e
en
ergy
c
a
n
be
f
orm
ul
at
ed
a
s
:
=
(
)
+
+
(1)
and
t
he
s
egm
ent
at
ion
re
s
ul
t
c
an
be
obt
ai
ned
f
rom
o
pt
im
al
c
u
rv
e
par
am
et
er
s
w
hic
h
is
:
=
m
i
n
(
)
(2)
S
o,
regar
ding
t
o
t
he
equat
ion
t
o
be
m
ini
m
iz
ed,
it
is
o
bv
ious
t
hat
t
he
perf
o
rm
a
n
c
e
an
d
qualit
y
of
s
eg
m
ent
at
ion
r
es
ult
is
s
t
ro
ngly
depe
nd
on
t
h
e
def
init
ion
of
s
na
k
e’
s
ene
r
gy
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
I
K
A
I
S
S
N
:
1693
-6
930
■
A
U
nif
ied
I
mage
E
ner
gy
A
p
proa
c
h
f
o
r
S
egment
at
ion
us
ing
B
S
plin
e
S
nak
e
(A
gun
g
A
lf
ians
y
ah
)
177
K
as
s
et
.
al.
r
epre
s
e
nt
s
t
he
ir
s
na
k
e
in
t
h
e
s
im
ple
s
t
w
ay
t
o
repre
s
e
nt
t
he
m
odel:
a
s
et
of
dis
c
ret
e
p
oint
s
as
s
na
k
e
el
em
ent
s
(s
nax
els
)
(
(
)
).
U
s
in
g
t
his
rep
re
s
e
nt
at
ion,
c
lo
s
e
d
c
ont
o
ur
s
c
an
b
e
f
orm
e
d
by
c
onn
ec
t
i
ng
t
he
las
t
s
n
ax
el
t
o
t
he
f
irs
t
one.
C
o
n
to
u
r
E
n
e
r
g
y
.
A
pply
ing
dis
c
r
et
point
s
r
ep
res
es
ent
at
ion,
t
he
c
on
t
our
e
ner
gy
,
c
an
be
appr
ox
im
at
ed
by
ac
c
om
o
dat
ing
t
he
e
las
t
ic
it
y
(
rep
r
es
e
nt
lengt
h
of
t
he
c
urv
e)
a
nd
rigidi
t
y
(r
e
pre
s
e
nt
t
he
int
egral
of
t
he
s
q
uar
e
of
t
he
c
u
rv
e
alon
g
t
he
c
ont
o
ur
),
s
o
:
=
(
)
+
(
)
=
(
)
‖
‖
+
(
)
‖
‖
(3)
w
he
re
t
he
s
u
bs
c
ript
denot
es
dif
f
ere
nt
iat
ion
w
it
h
re
s
p
ec
t
t
o
t
he
c
ur
v
e
param
et
e
r
.
T
he
m
odel
behav
i
or
i
s
c
ont
roll
e
d
by
c
ons
t
a
nt
s
and
,
res
p
e
c
t
iv
ely
w
eig
ht
ing
t
he
c
u
rv
e
elas
t
i
c
it
y
and
ri
gidit
y
.
T
his
e
ner
gy
de
f
init
ion
ea
s
ily
c
an
be
di
s
c
re
t
iz
ed
u
s
in
g
f
init
e
dif
f
erent
m
et
hod
I
n
v
al
i
d
so
u
r
ce
s
p
eci
fi
ed
.
as
:
E
≈
‖
v
−
v
‖
E
=
x
−
x
+
y
−
y
(4)
T
his
t
e
rm
w
il
l
m
inim
iz
e
t
h
e
dis
t
an
c
e
b
et
w
ee
n
t
he
point
s
in
t
h
e
s
na
k
e,
c
au
s
ing
t
he
s
hri
nk
i
ng
d
uri
ng
t
he
o
pt
im
iz
at
io
n
ene
rgy
pro
c
e
s
s
in
t
h
e
ab
s
en
c
e
of
an
im
ag
e
ex
t
ernal
ene
rgy
.
I
n
a
s
im
ila
r
w
ay
,
t
he
rigidit
y
t
erm
is
di
s
c
ret
iz
ed
as
:
E
≈
‖
v
−
2
v
+
v
‖
E
=
x
−
x
−
x
+
y
−
y
−
y
(5)
I
t
w
as
not
ed
by
W
illiam
and
S
hah
in
[
7]
t
hat
t
he
elas
t
ic
it
y
def
init
ion
us
in
g
f
init
e
dif
f
eren
c
e
s
d
is
c
ret
iz
at
ion
s
c
hem
e
i
s
v
alid
at
t
he
c
ondit
ion
t
h
at
m
odel'
s
s
na
x
els
ar
e
ev
e
nly
s
pa
c
e
d.
I
n
ot
her
c
a
s
e
s
,
t
hey
prop
os
ed
t
o
def
ine
a
c
ont
inuit
y
t
er
m
t
hat
s
ubt
r
ac
t
t
he
av
e
ra
ge
dis
t
an
c
e
of
t
he
s
n
ax
els
.
O
t
her
w
is
e
t
h
e
ener
gy
v
alue
w
ill
be
la
r
ger
f
or
p
oint
s
w
hic
h
ar
e
f
art
her
apart
.
T
hi
s
c
ons
t
r
aint
f
or
c
es
t
he
point
s
t
o
be
m
or
e
ev
enly
s
pa
c
ed,
a
nd
av
oids
a
po
s
s
i
ble
c
ont
r
ac
t
io
n
of
t
he
s
na
k
e.
I
m
ag
e
E
n
er
g
y
.
F
or
t
his
t
erm
,
K
as
s
pr
op
os
e
d
a
w
eig
h
t
ed
s
um
of
t
he
f
ollow
ing
e
nergi
es
t
erm
s
t
o
det
e
c
t
im
age
s
f
ea
t
ure:
E
!
(
v
s
=
α
"
∙
E
"
+
α
∙
E
+
α
#
∙
E
#
(6)
T
he
m
o
s
t
c
o
m
m
on
im
ag
e
f
unc
t
ion
al
in
t
his
m
o
del
is
us
in
g
t
he
im
a
ge
int
en
s
it
y
f
unc
t
io
n
.
T
his
t
erm
w
ill
s
im
ply
at
t
rac
t
t
he
c
o
nt
our
t
o
low
er
or
highe
r
in
t
ens
it
y
v
alue
s
in
im
age
s
depe
nding
on
t
he
v
alue.
Large
po
s
it
iv
e
v
alues
of
α
"
t
end
t
o
m
a
k
e
t
he
s
n
ak
e
alig
n
it
s
elf
w
it
h
dar
k
re
g
ions
in
t
he
i
m
age,
I
(
s
)
,
w
her
eas
l
a
rge
ne
gat
iv
e
v
alues
of
α
"
t
end
t
o
m
ak
e
t
he
s
na
k
e
ali
gn
it
s
elf
w
it
h
bri
gh
t
region
s
in
t
he
im
age.
T
he
edg
e
en
ergy
t
hat
at
t
rac
t
t
he
c
ont
o
ur
t
ow
a
rd
s
hi
gh
gra
dient
v
alue
s
c
an
b
e
c
al
c
ulat
e
d
as
s
q
uar
ed
t
o
narr
ow
t
he
edge
s
gra
die
nt
res
po
ns
e.
A
nd
s
im
ilarly
,
large
pos
it
iv
e
v
alues
of
t
end
t
o
m
a
k
e
t
he
s
n
ak
e
ali
gn
it
s
elf
w
it
h
s
ha
rp
edge
s
i
n
t
he
im
a
ge
w
he
rea
s
la
rg
e
neg
at
iv
e
v
alue
s
of
m
ak
e
t
h
e
s
n
ak
e
av
oi
d
t
he
ed
ge
s
.
is
def
in
ed
t
o
f
ind
t
he
t
erm
in
at
ion
s
of
line
s
egm
ent
s
a
n
d
c
or
ne
rs
.
K
as
s
prop
os
e
d
t
o
us
e
t
he
c
urv
at
u
re
of
is
o
-
c
ont
o
ur
s
in
a
G
aus
s
i
an
s
m
oot
h
ed
im
age
t
o
at
t
ra
c
t
t
he
c
ont
o
ur
s
t
ow
a
rd
s
line
t
erm
in
at
ion.
C
on
s
t
rai
nt
e
nergy
is
ap
plied
t
o
i
nt
er
a
c
t
iv
ely
guid
e
t
he
s
na
k
e
t
ow
a
rd
s
o
r
a
w
ay
f
r
om
part
ic
ula
r
f
ea
t
ures
.
T
hi
s
e
nergy
h
el
ps
t
he
c
o
nt
our
t
o
ov
erc
o
m
e
t
h
e
init
ializ
at
ion
proble
m
or
t
he
s
en
s
it
iv
it
y
t
o
nois
e.
A
c
on
s
t
raint
ene
rgy
w
a
s
pr
opo
s
e
d
f
or
c
la
s
s
i
c
al
s
na
k
e
by
allo
w
ing
t
h
e
u
s
er
t
o
at
t
ac
h
s
p
ring
s
bet
w
ee
n
po
int
s
of
t
he
c
o
nt
our
and
f
ix
t
heir
p
os
it
ion
in
t
he
im
a
ge
plane.
K
a
s
s
[
3]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
16
93
-6
930
T
E
L
K
O
M
N
I
K
A
V
ol.
8,
N
o.
2,
A
gus
t
us
2
0
10
:
175
–
1
86
178
def
ine
a
n
en
ergy
k
no
w
n
as
s
p
ring
(t
o
at
t
rac
t
t
he
s
n
ak
e
t
o
s
pe
c
if
ied
poi
nt
s
)
an
d
v
olc
a
noe
s
(t
o
repul
s
e
f
r
om
s
pe
c
if
ied
p
oin
t
s
)
w
it
hin
t
he
i
m
age.
T
hi
s
e
nergy
is
d
ef
in
ed
by
:
F
$
"
=
α
%
"
(
v
−
x
)
+
"
&
m
ax
(
pe
ak
'
,
1
r
'
)
"
&
(7)
T
he
s
p
ring
t
e
rm
at
t
rac
t
s
c
o
nt
our
point
v
t
o
a
point
v
in
t
h
e
im
age
plan
e,
w
it
h
a
c
ons
t
ant
α
%
"
as
t
he
s
prin
g
c
on
s
t
ant
.
T
he
ac
t
iv
e
c
ont
our
m
o
del
is
at
t
rac
t
ed
or
rep
elled
b
y
t
he
s
pring
depe
nding
o
n
α
%
"
s
ign
and
v
alue.
T
he
v
olc
a
no
t
erm
ac
t
s
a
s
a
repuls
i
on
f
orc
e
bet
w
een
a
point
on
t
he
im
age
at
a
d
is
t
an
c
e
f
rom
a
point
in
t
h
e
s
na
k
e.
T
he
larger
t
he
v
alue
of
r
'
,
t
he
s
t
ron
ge
r
t
he
r
epul
s
ion.
O
p
ti
m
i
z
a
ti
o
n
S
ch
em
e.
A
s
m
ent
ion
ed
p
rev
iou
s
ly
,
i
m
a
ge
s
eg
m
ent
at
ion
u
s
ing
s
na
k
e
c
a
n
be
f
orm
ul
at
e
d
as
a
p
ro
c
e
s
s
of
e
ner
gy
m
inim
iz
at
ion
t
hat
ev
olv
es
t
he
c
o
nt
ou
r.
T
his
m
inim
i
z
at
i
on
c
ont
r
ols
t
he
m
odel
def
o
r
m
at
ion
t
o
re
ac
h
t
he
de
s
ir
ed
s
eg
m
ent
a
t
ion
res
ult
.
T
he
t
erm
"
s
na
k
e"
c
om
e
s
f
ro
m
t
he
"
s
lip
an
d
s
lide"
m
ov
em
e
nt
of
t
h
e
c
ont
our
du
rin
g
t
his
m
inim
i
z
at
io
n
pro
c
e
s
s
.
O
rigin
ally
,
K
a
s
s
pr
opo
s
e
d
a
v
ariat
io
n
m
et
hod
t
o
s
olv
e
t
he
m
i
nim
i
z
at
ion
pr
oc
es
s
af
t
e
r
dis
c
ret
i
s
at
ion
us
i
ng
f
init
e
elem
ent
m
et
hod.
B
ut
t
hi
s
app
roa
c
h
d
oes
not
gua
r
ant
ee
t
h
e
gl
obal
m
inim
um
s
ol
ut
ion
an
d
r
eq
uire
s
e
s
t
i
m
at
i
on
of
hi
gh
of
f
er
d
eriv
at
iv
e
on
t
he
dis
c
ret
e
dat
a.
M
ore
ov
er
again,
M
or
eo
v
er,
hard
c
o
ns
t
rai
nt
s
,
w
hi
c
h
ar
e
re
s
t
ri
c
t
ion
on
t
he
rang
e
of
v
or
it
s
deriv
at
iv
e
s
,
c
an
not
be
di
r
ec
t
ly
enf
or
c
e
d.
G
iv
en
a
d
es
ir
ed
c
on
s
t
r
aint
t
erm
lik
e
a
m
ean
o
r
m
inim
um
s
na
x
el
s
pa
c
in
g
,
it
c
a
n
only
be
enf
orc
ed
by
in
c
r
eas
i
ng
t
he
a
s
s
oc
i
at
ed
w
ei
ght
ing
t
erm
,
w
hi
c
h
w
ill
f
o
r
c
e
m
ore
ef
f
e
c
t
on
t
his
c
o
ns
t
ra
int
,
but
at
t
he
c
o
s
t
of
ot
her
t
erm
s
.
M
any
ef
f
ort
s
w
er
e
deliv
e
re
d
af
t
erw
ard
t
o
s
olv
e
t
hi
s
m
inim
iz
at
ion
probl
em
.
O
ne
of
t
hem
w
a
s
G
re
edy
algorit
h
m
s
[
8]
w
hi
c
h
f
ind
t
he
s
olut
io
n
in
c
re
m
ent
ally
b
y
c
ho
os
i
ng
at
ea
c
h
s
t
e
p
t
h
e
dire
c
t
ion
w
hi
c
h
is
l
oc
ally
t
he
m
o
s
t
pr
om
i
s
ing
f
o
r
f
inal
r
es
ult
,
i.
e.
w
hi
c
h
p
rov
id
es
t
he
la
rge
r
en
e
rgy
dec
rea
s
e.
A
m
ini
pr
opo
s
e
d
[
9]
als
o
D
y
nam
ic
P
rog
r
am
m
ing
w
hi
c
h
en
s
u
res
a
globally
o
pt
im
al
s
olut
io
n
w
it
h
res
pe
c
t
t
o
t
h
e
s
ea
rc
h
s
pa
c
e,
a
nd
num
eric
al
s
t
abilit
y
by
m
ov
in
g
t
he
c
ont
o
ur
p
oint
s
on
a
dis
c
r
et
e
grid
w
it
ho
ut
any
de
riv
at
iv
e
num
eri
c
al
appr
ox
im
at
io
ns
.
T
h
e
o
pt
im
iz
at
ion
pr
oc
es
s
c
an
be
v
iew
ed
a
s
a
di
s
c
ret
e
m
ult
i
-
s
t
a
ge
de
c
is
i
on
pro
c
e
s
s
an
d
is
s
olv
e
d
b
y
a
t
im
e
-del
ay
ed
dis
c
ret
e
dy
n
a
m
ic
pr
ogr
am
m
ing
algo
rit
h
m
.
D
y
nam
ic
prog
ram
m
ing
by
pas
s
e
s
loc
al
m
inim
a
as
it
is
em
bed
ding
t
h
e
m
inim
iz
at
io
n
probl
em
in
a
neigh
bor
ho
od
relat
e
d
pr
oblem
.
W
e
pre
s
e
nt
e
d
in
t
hi
s
s
e
c
t
ion
K
a
s
s
pr
o
pos
it
io
n
t
o
e
x
t
rac
t
t
he
i
m
a
ge
f
e
at
ure
f
rom
t
h
e
im
age
u
s
ing
def
orm
able
m
odel.
T
hi
s
s
c
hem
e
w
ill
be
m
odif
ied
i
n
ex
plic
it
c
u
r
v
e
repr
es
ent
at
ion
us
in
g
B
S
pline
.
3.
B
S
P
L
I
N
E
S
N
A
K
E
T
his
m
od
el
def
or
m
abl
e
,
rep
re
s
ent
t
he
ev
olv
e
d
c
urv
e
(o
r
s
u
rf
a
c
e)
f
or
im
age
s
egm
ent
at
ion
in
an
ex
pli
c
it
par
am
et
r
i
c
f
orm
.
T
hi
s
re
p
res
ent
at
ion
al
low
s
a
dir
ec
t
int
era
c
t
ion
an
d
giv
es
a
c
om
p
ac
t
r
ep
res
ent
at
ion
f
or
re
al
-
t
im
e
im
plem
e
nt
at
ion.
I
t
is
w
idely
k
no
w
n
t
hat
,
s
im
ilia
r
t
o
t
he
poi
nt
ba
s
ed
r
ep
re
s
ent
at
ion;
dr
aw
ba
c
k
of
t
hi
s
m
o
del
r
ep
re
s
ent
at
ion
c
om
e
s
f
rom
it
s
dif
f
ic
u
lt
y
t
o
adapt
t
o
t
opologi
c
al
c
hang
es
(e.
g.
obje
c
t
s
plit
t
ing
o
r
m
e
rgi
ng)
duri
ng
m
odel
ev
olut
ion.
P
aram
et
ric
d
ef
orm
a
ble
m
odel
s
ar
e
u
s
u
ally
t
oo
s
en
s
it
iv
e
t
o
t
heir
ini
t
ial
c
on
dit
ions
bec
au
s
e
of
t
he
non
c
onv
ex
it
y
of
t
he
ener
gy
f
unc
t
ional
and
t
he
c
ont
r
ac
t
ion
f
o
rc
e
w
hi
c
h
a
ris
e
s
f
rom
t
he
int
er
nal
ener
gy
t
erm
.
O
ur
m
et
ho
d
i
s
dif
f
e
rent
f
o
r
m
t
hat
w
hi
c
h
w
a
s
p
ro
po
s
e
d
o
riginally
by
K
as
s
not
onl
y
in
t
er
m
f
m
odel
re
pr
e
s
ent
at
io
n,
but
als
o
in
e
ner
gy
re
pre
s
e
nt
at
ion
a
nd
als
o
t
he
opt
im
i
z
at
ion
m
et
hod
.
I
n
gene
ral,
w
e
enha
nc
e
t
he
c
la
s
s
i
c
al
m
et
hod
by
p
ropo
s
ing
a
c
o
m
p
ac
t
m
odel
re
pre
s
ent
at
io
n
and
ener
gy
def
init
ion.
3.
1.
R
ep
r
e
se
n
ta
ti
o
n
B
-s
pli
ne
is
of
t
en
us
ed
as
a
rep
re
s
ent
at
io
n
of
param
et
r
ic
def
or
m
able
m
odel.
I
n
t
his
c
a
s
e,
t
he
def
orm
ab
le
m
odel
is
s
plit
int
o
s
om
e
s
egm
e
nt
s
by
k
not
point
s
[
10
]
-[
13]
.
E
ac
h
c
urv
e
s
eg
m
ent
=
{
,
(
)
}
is
app
rox
im
a
t
ed
by
a
piec
ew
i
s
e
poly
n
o
m
ial
f
unc
t
io
n,
w
hic
h
is
o
bt
ained
by
a
linear
c
om
bin
at
ion
of
bas
i
s
f
unc
t
ion
s
an
d
a
s
et
of
c
on
t
rol
point
s
=
{
,
}
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
I
K
A
I
S
S
N
:
1693
-6
930
■
A
U
nif
ied
I
mage
E
ner
gy
A
p
proa
c
h
f
o
r
S
egment
at
ion
us
ing
B
S
plin
e
S
nak
e
(A
gun
g
A
lf
ians
y
ah
)
179
=
∑
(
)
&
(
(8)
T
hen,
a
poi
nt
-ba
s
ed
d
ef
orm
a
ble
m
o
del
c
a
n
be
analy
z
ed
lik
e
a
s
pe
c
ial
c
as
e
of
para
m
et
ri
c
c
urv
e
r
epr
es
e
nt
at
ion
w
her
e
t
he
ba
s
is
f
u
nc
t
ion
s
a
re
u
nif
orm
t
ra
ns
l
at
es
of
a
B
-s
pline
of
deg
ree
z
e
ro.
T
h
us
,
a
p
aram
et
ric
ap
proa
c
h
es
u
s
i
ng
s
m
oot
h
b
as
i
s
f
un
c
t
ion
s
w
ill
t
end
t
o
t
he
point
-b
as
ed
s
c
hem
e
a
s
t
he
num
b
er
of
bas
i
s
f
u
nc
t
ion
s
in
c
r
e
as
e
s
.
I
n
ge
neral,
h
ow
ev
er,
rep
re
s
ent
at
io
ns
u
s
ing
s
m
oot
h
ba
s
is
f
unc
t
io
ns
req
uire
f
e
w
er
p
aram
et
er
s
t
h
an
point
-ba
s
ed
appr
oa
c
he
s
a
nd
t
hus
r
es
ult
in
f
as
t
er
opt
i
m
iz
at
ion
al
go
rit
hm
s
[
14
]
.
M
oreov
e
r,
s
u
c
h
c
urv
e
m
o
del
s
hav
e
in
her
ent
re
gula
rit
y
an
d
he
nc
e
d
o
n
ot
re
quir
e
ex
t
ra
c
o
ns
t
r
aint
s
t
o
e
ns
u
re
s
m
oot
hne
s
s
[
1
4]
,
[
15]
.
B
ot
h
point
-ba
s
ed
an
d
par
a
m
et
ric
s
na
k
e
rep
re
s
ent
t
he
m
odel
in
ex
plic
it
w
ay
,
hen
c
e
it
is
eas
i
er
t
o
int
egrat
e
a
pri
or
s
ha
pe
c
o
ns
t
rai
bt
t
o
t
he
d
ef
orm
ab
le
m
idel.
M
o
reov
e
r
t
he
u
s
er
int
era
c
t
ion
c
a
n
be
a
c
c
om
m
odat
e
s
t
rai
ght
f
orw
ar
d
b
y
allow
ing
t
h
e
us
er
t
o
s
p
ec
if
y
s
om
e
p
oint
s
t
rough
t
he
d
es
ir
ed
c
o
nt
o
ur
ev
ol
ut
ion.
B
ut
t
he
in
c
onv
enie
nt
of
t
his
m
odel
l
ies
on
t
h
eir
les
s
f
lex
ibilit
y
in
ac
c
ount
in
g
f
or
t
opologi
c
al
c
ha
nge
s
d
uri
ng
t
he
e
v
olut
ion,
but
s
ev
e
ral
ef
f
ort
s
h
a
v
e
been
d
one
t
o
ov
erc
one
t
hi
s
lim
it
at
ion.
3.
2.
C
o
n
to
u
r
E
n
er
g
y
S
im
ilar
t
o
dis
c
ret
e
point
b
as
e
d
s
n
ak
e,
int
ernal
e
ne
r
gy
is
re
s
po
n
s
ible
f
o
r
en
s
uring
t
h
e
s
m
oot
h
ne
s
s
of
t
he
c
ont
o
u
r.
A
c
t
ually
,
K
as
s
pr
opo
s
e
d
a
linear
c
om
binat
ion
of
t
h
e
lengt
h
of
t
h
e
c
ont
o
ur
a
nd
t
he
int
egr
al
of
t
he
s
qu
ar
e
of
t
he
c
u
r
v
at
ure
alo
ng
t
he
c
ont
o
ur.
T
hus
in
ex
plic
it
c
ont
o
ur,
t
his
ener
gy
c
an
b
e
def
ined
a
s
:
=
(
)
+
)
)
*
(
+
)
)
)
−
)
′
′
(
)
(
)
+
)
(
)
)
+
*
(
(9)
w
he
re
t
he
s
e
c
on
d
t
erm
(
(
)
)
w
hi
c
h
i
s
c
urv
at
ure
o
n
poi
n
t
.
T
his
t
er
m
i
s
t
hen
c
a
n
be
s
im
plif
ied
a
s
:
|
|
=
1
"
(
|
)
)
|
+
|
)
)
|
)
*
(
*
(
(10
)
I
n
c
a
s
e
w
he
re
t
he
c
u
rv
e
is
pa
ram
et
e
riz
e
d
in
c
u
rv
ilinear
ab
s
c
i
s
s
a,
t
hen
"
c
a
n
be
rep
re
s
ent
e
d
as
:
"
=
*
#
$
(
)
+
)
)
*
(
%
(11
)
I
n
ot
her
c
as
e
w
he
n
k
not
s
i
n
pa
ram
et
e
ri
z
ed
c
urv
e
ar
e
not
in
t
he
c
ur
v
ilinear
ab
s
c
i
s
s
a
,
t
he
c
ont
o
ur
en
er
gy
c
an
be
m
o
dif
ied
as
f
ollo
w
s
:
=
$
|
|
)
t
|
,
−
"
|
*
(
(12
)
E
v
olv
ing
t
he
c
urv
e
w
it
h
s
u
c
h
a
t
e
rm
w
il
l
f
orc
e
t
he
c
urv
e
k
not
s
t
o
m
ov
e
on
t
a
ngent
ial
dire
c
t
ion
t
o
t
he
c
urv
e,
t
hu
s
bringi
ng
it
t
o
t
he
c
u
rv
ilinea
r
abs
c
is
s
a
p
os
it
ion
.
3.
2.
I
m
ag
e
E
n
er
g
y
T
he
im
age
e
nergy
def
init
i
on
pl
ay
t
he
c
rit
ic
al
rol
e
ov
erall
s
na
k
e
p
erf
orm
an
c
e
s
inc
e
t
his
t
erm
s
d
et
erm
i
ne
w
hi
c
h
p
ert
inen
c
e
f
eat
u
r
e
s
ho
uld
be
c
apt
ure
d
u
s
ing
t
he
def
orm
a
ble
m
od
el.
T
h
is
s
e
c
t
ion
w
ill
b
e
de
dic
at
ed
t
o
r
ev
iew
s
o
m
e
c
om
m
on
im
age
ene
rg
y
def
init
ion
a
nd
t
he
n
p
ro
p
os
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
16
93
-6
930
T
E
L
K
O
M
N
I
K
A
V
ol.
8,
N
o.
2,
A
gus
t
us
2
0
10
:
175
–
1
86
180
s
om
e
m
odif
ic
at
ions
w
hic
h
c
an
ov
e
rc
om
e
t
he
probl
em
.
W
e
als
o
p
re
s
ent
a
n
int
eg
r
at
iv
e
f
ram
e
w
ork
w
hi
c
h
unif
ie
s
t
he
gra
dient
-
bas
ed
an
d
re
gion
-
ba
s
ed
a
ppro
ac
he
s
in
ener
gy
def
in
it
ion
of
B
S
pline
s
na
k
e
.
T
he
m
os
t
c
o
m
m
on
im
age
ene
rgy
a
ppli
ed
f
or
t
he
s
n
ak
e
is
def
ine
d
a
s
t
he
i
nt
e
gral
of
t
he
s
qu
are
of
t
he
gradie
nt
m
agnit
ud
e
alo
n
g
t
he
c
urv
e
as
t
he
im
age
energy
.
T
he
m
ain
dra
w
b
ac
k
w
idely
k
no
w
n
on
u
s
ing
t
his
ene
rgy
is
t
he
lac
k
of
g
ra
dient
dir
ec
t
io
n.
T
his
i
nf
orm
at
ion
c
an
be
u
s
e
t
o
det
ec
t
t
he
edge,
s
in
c
e
a
t
t
he
bounda
r
y
im
age
gradi
ent
is
us
u
ally
perpe
ndi
c
ula
r
t
o
t
he
c
urv
e
.
T
his
d
ir
ec
t
io
n
s
hould
b
e
inc
or
po
rat
ed
t
o
t
he
im
age
energy
t
o
brin
g
t
he
s
na
k
e
m
ore
r
obu
s
t
f
or
im
age
s
egm
e
nt
at
ion.
E
dge
ba
s
e
d
ener
gy
.
F
or
t
his
s
na
k
e
w
e
prop
os
e
t
o
apply
an
im
a
ge
en
ergy
d
e
f
ined
a
s
int
egral
of
s
c
alar
f
ield
de
ri
v
ed
f
rom
t
he
gradi
ent
v
ec
t
or
f
ield.
M
a
t
h
em
at
ic
ally
,
it
c
an
b
e
f
orm
ul
at
ed
as
f
ollo
w
s
:
=
−
&
'
∙
∇
-
(
×
=
−
∮
∇
-
(
∙
(
×
*
)
=
−
&
∇
-
(
∙
(
‖
‖
+
,
-
)
(13
)
w
he
re
k
is
t
he
unit
v
ec
t
or
t
h
at
ort
ogon
al
t
o
t
he
im
age
plane,
+
,
-
denot
es
t
he
unit
no
rm
al
t
o
t
he
c
urv
e
a
t
and
∇
(
r
is
t
he
gradi
e
nt
of
t
he
im
age
(
at
t
h
e
point
t
.
R
egi
on
ba
s
ed
en
erg
y
.
T
his
regi
o
n
ba
s
ed
e
nergy
rep
re
s
ent
s
t
he
s
t
at
is
t
ic
al
c
ha
ra
c
t
eri
s
t
ic
s
on
a
r
egio
n
in
t
he
c
ont
o
ur
a
nd
p
rov
id
e
s
n
ak
e
a
bo
unda
ry
inf
or
m
at
ion
t
hu
s
v
ery
helpf
ul
w
hen
t
he
c
ont
o
ur
i
s
f
ar
aw
ay
f
ro
m
t
he
real
c
o
nt
our
t
o
be
d
et
ec
t
ed.
F
o
r
t
his
pu
rpo
s
e
w
e
as
s
um
e
t
w
o
regio
ns
in
t
h
e
im
age
s
(
w
hic
h
c
an
b
e
ex
pand
ed
in
t
o
m
ore
nu
m
ber
)
w
it
h
dif
f
erent
prob
abilit
y
di
s
t
ribut
io
ns
.
E
ac
h
of
t
he
s
e
regio
ns
hav
e
dif
f
erent
m
e
ans
and
v
ari
anc
es
.
W
e
f
o
llow
S
t
aib’s
[
14
]
f
orm
ulat
io
n
t
o
det
erm
in
e
t
he
regio
n
lik
e
h
o
od
f
unc
t
io
n:
=
−
$
lo
g
.
(
/
|
∈
ℛ
0
.
−
$
lo
g
.
(
/
|
∈
ℛ′
0
.
)
(14
)
W
he
re
ℛ
and
ℛ′
denot
e
t
he
dif
f
erent
re
gi
ons
in
t
h
e
c
urv
e
an
d
0
an
d
0
′
indic
at
e
t
he
pos
it
io
n
in
s
id
e
or
out
s
i
de
t
he
r
egion
re
s
pec
t
iv
ely
.
T
h
e
ene
rgy
d
ef
ined
in
[
1
4
]
w
ill
be
m
ax
im
u
m
w
he
n
ℛ
=
0
and
ℛ′
=
0
′
.
T
hus
t
hi
s
ene
r
gy
c
an
be
r
ef
orm
ul
at
ed
a
s
:
=
−
lo
g
.
(
/
|
∈
ℛ
0
.
−
+
$
lo
g
.
(
/
|
∈
ℛ′
0
.
)
(15
)
w
h
e
re
–
=
+
$
lo
g
.
(
/
|
∈
ℛ′
0
.
∪
.
)
.
is
indep
ende
nt
f
rom
t
he
po
s
i
t
ion
of
t
he
c
urv
e,
s
o
it
c
an
be
rem
ov
ed
f
rom
t
he
c
o
s
t
f
unc
t
ion
c
alc
ul
at
ion.
T
his
s
im
plif
ic
at
ion
brin
gs
t
he
regio
nal
ba
s
e
d
ener
gy
int
o
new
f
or
m
ulat
i
on:
=
−
lo
g
1
.
(
/
|
∈
ℛ
.
(
/
|
∈
ℛ′
2
0
.
(16
)
I
n
t
he
ab
s
en
c
e
o
f
p
rio
r
k
now
l
edge
of
t
he
pro
babilit
y
dis
t
ribut
io
n
s
.
(
/
|
∈
ℛ
and
.
(
/
|
∈
ℛ′
c
an
be
es
t
im
at
ed
f
rom
im
age
(
as
t
he
i
m
age
ev
olv
e
s
and
t
he
c
ur
rent
pos
it
io
n
of
t
he
c
ont
o
ur
c
an
be
a
s
s
um
ed
def
ine
d
t
he
regio
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
I
K
A
I
S
S
N
:
1693
-6
930
■
A
U
nif
ied
I
mage
E
ner
gy
A
p
proa
c
h
f
o
r
S
egment
at
ion
us
ing
B
S
plin
e
S
nak
e
(A
gun
g
A
lf
ians
y
ah
)
181
I
t
s
hould
b
e
not
ed
a
s
in
t
eres
t
in
g
p
ro
pert
y
of
t
his
regio
n
en
er
gy
def
init
ion
t
hat
t
he
ex
t
ens
io
n
of
t
his
def
init
ion
in
m
ult
i
-v
alu
ed
and
m
ult
i
-c
h
ann
el
im
a
ges
(e.
g.
c
olor
im
ag
e)
i
s
is
s
t
raig
ht
f
or
w
ar
d.
U
n
i
fi
ed
en
e
r
g
y
.
B
ot
h
of
t
hes
e
e
ner
gie
s
d
ef
init
ion
h
av
e
t
heir
o
w
n
s
t
r
ong
poi
nt
s
a
nd
w
ea
k
n
es
s
.
T
he
e
dg
e
-ba
s
ed
e
ner
gy
c
an
giv
e
a
g
ood
l
oc
aliz
at
ion
of
t
h
e
c
ont
our
ne
ar
t
he
boun
dari
es
.
U
nf
ort
u
nat
ely
,
it
ha
s
a
s
m
all
ba
s
in
of
a
t
t
rac
t
ion,
t
h
us
re
quiri
ng
a
g
ood
i
nit
ializ
at
ion
near
t
o
t
h
e
d
es
ir
ed
c
ont
ou
r
o
r
apply
ing
a
ball
oon
f
or
c
e
in
s
t
ead
[
16
]
,
[
17
]
.
O
n
t
h
e
ot
her
ha
nd,
t
he
regio
n
-
ba
s
ed
ener
gy
has
a
large
ba
s
in
of
at
t
rac
t
ion
a
nd
c
a
n
c
o
nv
e
rge
ev
e
n
if
ex
plic
it
edg
es
a
re
not
pr
es
ent
[
1
8
]
.
H
ow
ev
er,
i
t
does
not
giv
e
as
go
od
lo
c
aliz
at
io
n
a
s
t
h
e
edg
e
-
bas
ed
ene
rgy
at
t
he
im
age
b
oun
d
arie
s
.
M
ot
iv
at
ed
by
t
he
c
o
m
plem
e
nt
ary
f
eat
ure
s
of
t
hes
e
s
c
he
m
e
s
w
e
pr
opo
s
e
a
unif
ied
f
orm
o
f
im
age
ene
rg
y
.
T
his
ener
g
y
c
an
be
f
orm
ulat
ed
a
s
:
=
3
(
+
.
(
1
−
)
3
(
(
)
(17
)
w
he
re
f
unc
t
io
n
param
et
er
denot
e
s
t
he
c
ont
ribut
io
n
of
eac
h
ene
rgy
in
t
his
bs
pline
s
na
k
e.
T
his
para
m
et
er
al
ow
s
u
s
t
o
t
u
n
e
t
he
im
age
ener
gy
re
ga
r
ding
t
o
t
h
e
t
y
pe
an
d
q
ualit
y
of
t
he
im
a
g
e
w
e
w
ant
t
o
s
egm
ent
.
F
or
ex
a
m
ple
in
ult
r
as
ound
im
ag,
w
here
t
he
noi
s
e
is
v
ery
p
re
s
ent
an
d
g
radi
ent
w
ill
not
be
reliable
w
e
c
a
n
s
et
=
0,
s
o
t
he
s
na
k
e
be
c
am
e
purely
u
s
e
regi
on
bas
ed
ener
gy
.
F
or
les
s
nois
y
im
age
(C
T
or
M
R
I
)
it
c
an
be
c
om
bi
ne
s
us
i
ng
s
et
t
ing
v
alue
t
o
0.
5
t
o
m
ak
e
t
he
s
a
m
e
c
ont
ri
but
ion
b
et
w
ee
n
regi
o
n
and
gr
adie
n
t
bas
ed
e
ner
g
y
.
3.
3.
E
xter
n
al
C
o
n
s
tr
ai
n
t
E
n
er
g
y
S
t
ill
ins
pire
d
by
K
as
s
,
w
e
als
o
int
eg
rat
e
a
u
s
e
r
t
e
r
m
c
o
ns
t
raint
,
w
h
ere
t
he
u
s
er
m
ig
h
t
s
pe
c
if
y
a
f
ew
point
s
t
hat
s
hould
lie
o
n
t
he
c
o
nt
ou
r
t
o
be
det
e
c
t
ed.
W
e
c
on
s
t
rai
n
t
he
s
n
ak
e
by
addin
g
an
e
nergy
t
erm
w
hi
c
h
is
t
he
dis
t
anc
e
bet
w
ee
n
t
hes
e
point
s
an
d
t
he
c
o
rre
s
p
on
ding
c
lo
s
e
s
t
point
s
on
t
he
c
u
rv
e.
F
or
t
his
s
na
k
e
t
he
c
on
s
t
rai
nt
energy
is
g
iv
en
by
:
=
4
t
−
,
4
,
∈
(
(
,
*
)
&
(1
8
)
w
he
re
,
are
t
he
int
rodu
c
e
d
c
on
s
t
raint
s
in
t
he
B
S
pline
s
na
k
e
.
T
h
is
appr
oa
c
h
c
an
be
int
erp
ret
ed
f
o
rm
K
a
s
s
m
o
d
el
a
s
a
n
int
ro
duc
t
io
n
of
v
i
rt
ual
s
pri
ng
s
t
h
at
pull
s
t
he
c
urv
e
t
o
w
a
rd
s
t
he
des
i
red
p
oint
s
:
O
ne
en
d
of
t
he
s
pring
i
s
f
ix
ed
t
o
t
he
c
on
s
t
raint
p
oint
w
hile
t
he
ot
her
end
s
li
des
on
t
he
c
u
rv
e
.
3.
4.
O
p
ti
m
i
z
ati
o
n
S
ch
em
e
A
s
m
e
nt
ione
d
pr
ev
iou
s
ly
,
im
age
s
eg
m
ent
at
ion
i
s
f
inally
a
t
o
t
al
of
s
t
at
e
d
ene
rgy
m
inim
iz
at
ion
pro
c
e
s
s
t
h
at
w
i
ll
pla
c
e
a
r
egula
r
c
ont
o
ur
in
t
he
e
dg
e
of
t
he
obje
c
t
w
e
w
ant
t
o
det
ec
t
.
F
o
r
o
ur
c
as
e,
w
e
d
o
not
re
quir
e
global
o
pt
im
u
m
s
olut
io
n,
s
i
nc
e
t
h
e
init
ial
c
ont
o
ur
c
an
be
prov
ide
d
int
er
ac
t
iv
ely
by
us
er
t
o
obt
ain
a
rough
init
ial
c
ont
o
ur
ne
ar
t
o
t
he
edge.
E
v
en
t
hough
,
a
robu
s
t
opt
im
i
z
at
ion
s
c
hem
e
c
o
nv
erg
e
t
o
t
he
m
inim
um
s
olut
io
n
in
a
c
c
ept
abl
e
nu
m
ber
of
it
erat
ion
is
s
t
r
ongly
de
s
ire
d
,
t
o
m
ak
e
t
he
algorit
h
m
run
f
as
t
.
W
e
p
ro
po
s
e
t
o
apply
g
r
adient
d
es
c
e
nt
met
hod
in
t
his
s
pline
bas
ed
s
eg
m
ent
at
ion
t
ec
hni
que.
I
t
is
a
f
ir
s
t
o
r
der
opt
im
iz
at
io
n
algorit
h
m
w
hi
c
h
s
ee
k
s
t
o
f
ind
a
lo
c
al
m
i
nim
um
s
olut
i
on
of
an
ene
rgy
f
unc
t
ion
by
t
a
k
ing
s
t
e
p
p
ro
port
ion
al
t
o
t
he
n
egat
iv
e
of
t
he
e
s
t
im
a
t
ed
gr
adie
nt
(
i.
e.
f
irs
t
or
der
der
iv
at
iv
e)
of
t
he
f
unc
t
ion
at
t
h
e
c
u
rr
ent
poin
t
.
H
enc
e,
t
o
perf
orm
t
hi
s
m
et
hod
w
e
ne
e
d
abs
olut
ely
t
he
es
t
im
at
io
n
of
t
he
f
irs
t
d
eriv
at
iv
e
of
a
ll
t
he
def
ined
ener
gy
.
T
he
rev
iew
of
t
his
gradi
ent
es
t
i
m
at
ion
is
p
re
s
ent
e
d
in
t
his
f
ollow
ing
p
ar
t
.
P
ar
ti
al
d
er
i
v
ati
v
e
o
f
i
n
ter
n
al
en
er
g
y
.
D
if
f
erent
iat
in
g
t
he
ex
pres
s
ion
of
=
/
an
d
s
im
plif
y
ing
f
u
rt
her,
w
e
obt
ain
t
he
part
ial
deriv
at
iv
es
a
s
a
s
im
pl
e
m
ult
idim
en
s
ion
al
f
ilt
ering
of
t
he
s
c
aling
f
un
c
t
i
on
c
o
ef
f
ic
ient
s
.
T
hu
s
t
he
d
eriv
at
iv
e
of
c
an
be
c
om
put
ed
as
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
16
93
-6
930
T
E
L
K
O
M
N
I
K
A
V
ol.
8,
N
o.
2,
A
gus
t
us
2
0
10
:
175
–
1
86
182
5
5
,
,
0
=
"
,
0
1
"
,
0
1
"
,
0
1
|
|
,
|
|
,
|
|
34
ℎ
6
,
7
,
+
+
"
,
0
1
"
,
0
1
"
,
0
1
|
|
,
|
|
,
|
|
34
ℎ
6
,
7
,
+
+
4
"
,
0
1
|
|
34
ℎ
6
5
5
,
,
0
=
"
,
0
1
"
,
0
1
"
,
0
1
|
|
,
|
|
,
|
|
34
ℎ
6
,
7
,
+
+
"
,
0
1
"
,
0
1
"
,
0
1
|
|
,
|
|
,
|
|
34
ℎ
6
,
7
,
+
+
4
∑
"
,
0
1
|
|
34
ℎ
6
(19
)
w
he
re
ℎ
6
,
7
,
+
=
8
)
+
8
)
+
6
+
8
)
+
7
+
8
)
+
+
5
5
ℎ
6
=
8
)
+
8
)
+
6
5
5
T
his
m
ult
idi
m
ens
i
onal
f
ilt
ering
i
s
p
e
rf
orm
e
d
by
as
s
um
in
g
t
h
e
pe
riodi
c
b
ound
ary
c
on
dit
ion
s
.
T
he
c
om
put
at
i
onal
c
o
m
ple
x
it
y
is
s
m
all,
s
inc
e
t
he
s
um
de
pen
d
s
only
on
t
he
c
oef
f
ic
i
ent
s
e
quen
c
e
w
h
o
s
e
num
be
r
is
t
y
pic
ally
m
uc
h
les
s
er
t
han
t
he
num
b
er
of
c
urv
e
s
am
pl
es
.
T
hus
,
t
he
c
o
m
put
at
ional
c
om
plex
it
y
c
an
be
redu
c
e
d
u
s
ing
t
hat
f
orm
ulat
ion.
P
r
o
b
ab
i
l
i
t
y
d
i
str
i
b
u
ti
o
n
fu
n
c
ti
o
n
esti
m
ati
o
n
.
A
s
m
ent
ione
d
in
prev
iou
s
s
e
c
t
ion,
t
he
ev
aluat
ion
of
c
ur
re
nt
regio
n
ener
gy
nee
ds
s
p
ec
if
i
c
at
i
on
of
proba
bi
lit
y
dis
t
ribut
io
n
f
unc
t
ion.
T
his
m
eas
ur
e
c
an
be
es
t
im
at
e
d
als
o
f
ro
m
t
h
e
im
ag
e
d
at
a
in
c
ondit
io
n
t
hat
t
he
ev
olv
ed
s
na
k
e
pla
c
e
s
c
lo
s
e
t
o
t
he
det
ec
t
e
d
bou
ndary
.
I
n
t
his
s
na
k
e,
w
e
a
pply
t
he
G
au
s
s
i
an
dis
t
ribu
t
ion
as
den
s
i
t
y
bec
au
s
e
it
r
epre
s
e
nt
s
t
h
e
dat
a
u
s
ing
f
ew
pa
ram
e
t
ers
w
hic
h
a
re
m
ean
a
n
d
v
ari
an
c
e.
T
his
es
t
im
at
ion
re
quire
int
e
gr
at
ing
t
he
im
age
and
it
s
s
qua
r
e
in
t
he
regio
n
boun
ded
by
0
.
P
ar
ti
al
d
er
i
v
ati
v
e
o
f
co
n
str
ai
n
t
en
e
r
g
y
.
F
or
im
ple
m
ent
at
ion
p
u
rpo
s
e,
w
e
a
s
s
um
e
t
h
a
t
t
he
opt
im
al
p
aram
et
er
;
9
=
0
.
.
:
are
k
no
w
n.
F
inall
y
t
he
def
init
ion
of
c
on
s
t
rain
t
energy
:
=
∑
4
t
−
,
4
,
∈
(
(
,
*
)
&
(2
0
)
c
an
b
e
m
odif
i
ed
as
:
;
<
<
"
,
'
=
<
<
"
,
'
=
>
=
∑
?
@
,
,
A
−
B
CD
4
&
8
(
−
'
)
(2
1
)
U
s
i
ng
t
he
c
h
ara
c
t
eri
s
t
ic
of
t
he
s
c
ali
ng
f
unc
t
io
ns
,
w
e
c
an
lim
it
t
he
s
um
t
o
t
he
r
elev
ant
indic
es
w
e
n
eed
t
o
ev
alu
at
e
it
only
in
c
ert
ai
n
num
b
er
of
point
s
.
W
e
al
s
o
re
s
o
rt
t
o
a
t
w
o
-
s
t
ep
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
I
K
A
I
S
S
N
:
1693
-6
930
■
A
U
nif
ied
I
mage
E
ner
gy
A
p
proa
c
h
f
o
r
S
egment
at
ion
us
ing
B
S
plin
e
S
nak
e
(A
gun
g
A
lf
ians
y
ah
)
183
opt
im
iz
at
io
n
w
he
re
t
he
s
n
ak
e
is
f
ir
s
t
ev
olv
ed
us
ing
t
he
f
orm
ula
s
f
or
t
he
deri
v
at
iv
es
w
it
h
t
he
c
ur
re
nt
s
et
po
int
s
.
C
ur
v
e
’
s
lengt
h
and
ar
ea.
T
he
c
o
m
put
a
t
ion
of
t
he
i
nt
ernal
e
ner
g
y
als
o
req
uir
es
t
he
es
t
im
at
ion
of
t
he
c
ur
rent
l
engt
h
of
t
he
c
urv
e.
W
e
c
om
put
e
t
he
len
gt
h
as
a
di
s
c
r
et
e
appr
ox
im
at
io
n
s
u
c
h
a
s
:
E
F
+
ℎ
=
6
∑
G
′
#
6
%
+
′
#
6
%
*
6
(
(2
2
)
A
nd
t
he
are
a
us
in
g
G
re
en
t
heor
em
a
s
:
H
F
=
∑
∑
"
,
0
4
&
4
"
,
1
I
(
'
−
6
)
*
0
&
(
(2
3
)
w
he
re
q
m
=
$
φ
t
φ
)
(
t
−
m
)
5
5
.
I
t
s
hould
be
not
e
t
hat
t
he
area
c
o
m
put
e
d
us
in
g
t
his
f
orm
ulat
io
n
is
s
igne
d;
and
t
his
s
i
gn
c
an
b
e
us
e
t
o
det
e
r
m
ine
t
he
dir
e
c
t
ion
of
t
he
c
urv
e.
4.
R
E
S
U
L
T
A
N
D
D
I
S
C
U
S
S
I
O
N
I
n
t
his
s
ub
-
s
ec
t
ion
w
e
w
il
l
dem
on
s
t
rat
e
qualit
at
iv
el
y
t
he
perf
or
m
anc
e
of
t
he
propo
s
e
d
s
na
k
e
w
it
h
di
f
f
erent
par
am
et
ers
af
f
e
c
t
e
d
in.
W
e
a
re
int
ere
s
t
ed
in
s
egm
ent
ing
m
edic
al
im
ag
es
f
rom
dif
f
ere
nt
m
odalit
ie
s
b
ec
a
us
e
it
is
v
ery
c
hall
eng
ing
due
t
o
l
o
w
c
ont
ra
s
t
a
nd
noi
s
e.
W
e
applie
d
t
he
m
et
hod
on
C
T
s
c
ann
er
an
d
M
R
I
im
age
s
as
an
ex
am
pl
e
of
c
lear
im
age
t
hat
eas
y
t
o
s
egm
ent
a
nd
ec
h
oc
ar
di
ogr
aph
im
age
s
w
he
re
a
s
t
ro
n
g
noi
s
e
p
res
e
nt
in
t
he
im
a
g
e
an
d
m
ak
e
t
he
s
egm
ent
at
ion
t
as
k
is
v
e
ry
dif
f
ic
ult
t
o
be
perf
orm
ed.
(a)
(b)
(c
)
(d)
F
igur
e
1.
C
o
m
pari
s
o
n
of
s
egm
ent
at
io
n
r
es
ult
of
a
C
T
s
c
an
dat
a.
(a
)
init
ial
c
ont
ou
r;
and
s
egm
ent
at
i
on
res
ult
u
s
ing
(
b)
edg
e
ba
s
e
d;
(c
)
regio
n
bas
ed
re
gion
only
;
(
d)
c
om
binat
ion
bet
w
e
en
50%
region
b
as
ed
and
50%
ed
ge
ba
s
ed;
int
egrat
in
g
t
he
r
egion
b
as
ed
ener
gy
allow
s
t
he
s
na
k
e
t
o
av
oid
loc
al
m
inim
um
s
olut
i
on
due
t
o
hig
h
gradi
ent
ar
ound
t
he
o
bje
c
t
and
c
apt
u
r
e
t
he
det
ail
in
t
he
im
age.
F
igur
e
1
d
em
ons
t
r
at
es
ou
r
pro
po
s
ed
un
if
ied
ene
rgy
f
or
B
S
pline
d
ef
orm
a
ble
m
odel
f
o
r
im
age
s
egm
e
nt
at
ion.
O
rigi
nal
m
od
el
ap
ply
ing
only
gr
adient
ba
s
e
d
ener
gy
w
ill
b
e
v
ery
s
en
s
it
i
v
e
t
o
t
he
pres
e
nt
of
loc
al
m
inim
um
ene
rg
y
in
t
he
im
a
ge
(noi
s
e,
s
p
ec
k
le
or
adj
a
c
ent
anat
o
m
i
c
al
orga
ns
).
I
n
ot
her
han
d,
r
e
gion
ba
s
ed
e
nergy
ov
er
c
o
m
e
t
he
pr
obl
em
but
it
ha
s
t
he
p
robl
em
in
c
apt
u
ring
t
h
e
objec
t
d
et
ail
in
t
he
im
age.
T
hes
e
pr
obl
em
s
c
an
b
e
s
olv
e
by
unif
y
ing
bot
h
e
ner
gy
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
16
93
-6
930
T
E
L
K
O
M
N
I
K
A
V
ol.
8,
N
o.
2,
A
gus
t
us
2
0
10
:
175
–
1
86
184
t
hus
m
i
nim
u
m
loc
al
s
olut
i
on
c
ould
be
av
oided;
b
ut
alw
ay
s
k
eep
t
he
obj
ec
t
d
e
t
ail
in
t
he
f
in
al
im
age
s
egm
e
nt
at
ion
re
s
ult
s
.
(a)
(b)
(c
)
(d)
F
igur
e
2.
I
nit
ializ
at
io
n
inde
pend
en
c
y
in
prop
os
ed
s
eg
m
ent
at
ion
m
e
t
hod
us
i
ng
re
gion
ba
s
e
d
ener
gy
.
(a)
ini
t
ial
c
ont
ou
r
f
or
s
eg
m
ent
at
io
n;
and
s
e
gm
e
nt
at
ion
re
s
ult
us
in
g
(b
)
pur
ely
gradie
nt
bas
ed
en
ergy
,
m
odel
def
or
m
able
t
ra
ppe
d
int
o
loc
al
o
pt
im
iz
at
ion;
(
c
)
r
egio
n
ba
s
ed
regi
on
onl
y
;
t
he
m
et
hod
r
eac
h
t
he
m
ini
m
um
s
ol
ut
ion
,
ev
en
w
it
h
init
ializ
at
ion
v
er
y
f
ar
f
rom
des
ired
c
ont
ou
r.
(d)
C
om
bi
nat
i
on
bet
w
een
5
0%
regio
n
ba
s
ed
a
nd
50%
edge
b
as
ed.
M
et
hod
init
ial
iz
at
ion
is
a
c
rit
ic
al
is
s
ue
in
c
las
s
ic
al
im
age
s
egm
ent
at
ion
us
i
n
g
m
odel
def
orm
able.
I
ns
t
ea
d
of
pro
pos
i
ng
a
s
m
a
rt
aut
om
at
ic
m
et
hod
t
o
pl
ac
e
a
n
init
ial
c
ont
o
ur
n
ear
t
o
t
he
des
i
red
s
olut
ion,
w
e
c
hoo
s
e
t
o
int
e
grat
e
t
he
r
egi
on
ba
s
ed
e
ne
rgy
t
o
s
olv
e
t
his
p
roble
m
.
T
he
ov
erall
r
ole
of
t
his
im
age
e
nergy
f
or
t
hi
s
purp
os
e
c
an
be
ob
s
e
rv
ed
i
n
F
igur
e
2.
U
s
ing
v
e
ry
rou
gh
init
ial
m
od
el
pre
s
ent
e
d
in
F
igur
e
2
(a
),
s
nak
e
w
it
h
s
ol
ely
im
age
ba
s
ed
en
ergy
f
a
ils
t
o
c
a
pt
ur
e
t
he
des
i
red
obj
e
c
t
s
due
t
o
t
he
p
re
s
ent
of
loc
al
m
ini
m
um
s
olut
io
n
re
pre
s
e
nt
e
d
by
lo
c
ally
high
gradi
ent
in
t
h
e
im
age.
I
nt
r
odu
c
ing
regio
n
bas
ed
e
ner
gy
helps
t
h
e
s
na
k
e
s
olv
e
t
he
pr
oblem
a
s
pre
s
ent
e
d
in
F
igur
e
2(
c
).
F
urt
he
rm
o
re,
u
nif
y
ing
t
his
en
ergy
enh
an
c
e
t
he
m
et
hod
p
erf
orm
an
c
e.
(a)
(b)
F
igur
e
3.
B
rai
n
s
t
ru
c
t
ur
e
(v
ent
ric
l
e)
s
eg
m
ent
at
ion,
(a
).
init
ial
c
ont
o
ur,
(b
).
r
es
ult
us
in
g
im
age
ener
gy
w
hi
c
h
is
c
om
binat
io
n
bet
w
e
en
25
%
region
b
as
ed
and
7
5%
edge
b
as
ed
.
F
igur
e
3
an
d
F
igur
e
4
illu
s
t
rat
e
pe
rf
or
m
a
n
c
e
of
t
he
m
et
hod
t
o
s
eg
m
ent
t
he
an
a
t
om
ic
al
obje
c
t
s
in
c
le
ar
a
nd
n
ois
y
i
m
age
s
.
T
he
s
e
f
igur
es
als
o
s
ho
w
how
t
h
e
ene
rgy
u
nif
ic
at
ion
c
o
uld
be
t
rim
m
ed
i
nt
uit
iv
ely
.
F
or
pr
a
c
t
ic
al
c
a
s
e,
t
o
s
eg
m
ent
a
c
l
ear
im
age
s
u
c
h
a
s
M
R
I
o
r
C
T
im
a
ge
s
,
w
e
c
an
s
et
a
hi
gh
pro
po
rt
i
on
of
im
age
bas
e
d
ene
rgy
.
I
n
ot
her
ha
nd,
f
or
nois
y
im
age
s
u
c
h
as
ult
ras
oun
d
an
d
ec
ho
c
a
rdio
grap
h
im
a
ge
s
,
regio
n
ba
s
e
d
ene
rgy
s
ho
uld
be
s
et
in
highe
r
p
riorit
y
t
o
redu
c
e
t
h
e
n
ois
e
s
e
ns
it
iv
it
y
.
B
ut
in
all
of
c
a
s
e,
unif
i
c
at
ion
of
bot
h
ene
rgie
s
w
ill
enha
nc
e
t
he
Evaluation Warning : The document was created with Spire.PDF for Python.