T
E
L
KO
M
NI
K
A
,
V
ol
.
14,
N
o.
3,
S
ept
em
ber
20
16,
pp.
10
0
9
~
101
5
I
S
S
N
:
1
693
-
6
930
,
ac
c
r
edi
t
ed
A
b
y
D
IK
T
I,
D
e
c
r
e
e
N
o
:
58/
D
I
K
T
I
/
K
ep/
2013
D
O
I
:
10.
12928/
T
E
LK
O
M
N
I
K
A
.
v
1
4
i
3
.
3412
10
09
R
ec
ei
v
ed
J
anu
ar
y
2
1
,
201
6
;
R
ev
i
s
ed
Ma
y
1
6
,
201
6
;
A
c
c
ept
ed
J
u
ne
3
,
2
01
6
G
P
U C
UD
A
A
c
c
eler
at
ed
Im
a
g
e I
n
p
a
i
nti
n
g u
s
i
n
g
Four
th
O
rd
e
r PD
E
E
qu
a
ti
o
n
E
d
w
i
n
P
r
a
n
a
n
ta
*
1
,
P
r
a
n
o
w
o
2
,
D
j
o
k
o
B
u
d
i
a
n
to
3
D
epar
t
e
m
ent
of
I
nf
or
m
at
i
on T
e
c
hno
l
ogy
,
F
ac
u
l
t
y
of
I
ndu
s
t
r
i
a
l
T
ec
hnol
o
gy
,
U
ni
v
er
s
i
t
y
A
t
m
a
J
ay
a Y
ogy
ak
a
r
t
a,
J
l
n B
a
bar
s
ar
i
no 4
4,
Y
ogy
ak
ar
t
a,
D
I
Y
,
I
ndo
ne
s
i
a
*
C
or
r
es
po
ndi
ng a
ut
hor
,
e
-
ma
i
l
:
edw
i
npr
ana
nt
a@
g
m
ai
l
.
c
om
1
,
pr
an@
m
ai
l
.
uaj
y
.
a
c
.
i
d
2
,
dj
ok
o.
bdy
@
g
m
ai
l
.
c
o
m
3
A
b
st
r
act
T
hi
s
paper
de
s
c
r
i
bes
t
he
t
ec
hni
qu
e
t
o ac
c
el
er
at
e i
npa
i
nt
i
ng
pr
o
c
es
s
u
s
i
n
g f
our
t
h or
d
e
r
P
D
E
equat
i
on
u
s
i
ng
G
P
U
C
U
D
A
.
I
npai
n
t
i
n
g
i
s
t
he
pr
o
c
e
s
s
o
f
f
i
l
l
i
ng
i
n
m
i
s
s
i
ng
par
t
s
of
dam
age
d
i
m
ages
bas
ed
on i
nf
or
m
at
i
o
n gl
eane
d f
r
om
s
ur
r
ound
i
ng
ar
ea
s
.
I
t
us
e
s
t
h
e
G
P
U
c
om
put
at
i
on ad
v
a
nt
ag
e t
o pr
oc
es
s
P
D
E
equat
i
on i
nt
o
par
al
l
e
l
pr
oc
e
s
s
.
F
our
t
h
or
de
r
P
D
E
w
i
l
l
be
s
o
l
v
ed
us
i
ng p
ar
al
l
el
c
om
pu
t
at
i
o
n i
n G
P
U
.
T
hi
s
m
et
hod c
an
s
pe
ed u
p t
he
c
om
put
at
i
on t
i
m
e up
t
o 3
6x
us
i
ng
N
V
D
I
A
G
E
F
O
R
C
E
G
T
X
670.
Ke
y
w
o
rd
s
: In
p
a
i
n
ti
n
g
, P
D
E
, G
P
U
-
C
U
D
A
,
par
al
l
e
l
,
c
om
put
a
t
i
on
C
o
p
y
r
i
g
h
t
©
20
16 U
n
i
ver
si
t
a
s
A
h
mad
D
ah
l
an
.
A
l
l
r
i
g
h
t
s r
eser
ved
.
1
.
I
n
tr
o
d
u
c
ti
o
n
T
he
I
npai
n
t
i
n
g
pr
ob
l
em
i
s
t
he t
er
m
f
or
w
hat
r
es
ear
c
he
r
s
i
n i
m
age pr
oc
es
s
i
ng c
al
l
“
i
m
age
i
nt
er
p
ol
at
i
on”
.
D
i
g
i
t
a
l
I
m
age r
es
t
or
at
i
on t
ec
hn
i
qu
e us
i
ng i
nf
or
m
at
i
on g
l
e
ane
d f
r
om
s
ur
r
oundi
n
g
ar
eas
of
i
m
age
t
o
f
i
l
l
t
he
d
am
ages
or
m
i
s
s
i
ng
par
t
s
of
i
m
age
i
s
c
al
l
ed i
npa
nt
i
ng.
T
he
o
r
i
gi
n of
i
np
ai
n
t
i
n
g
w
as
beg
un i
n ar
t
w
or
l
d,
i
t
w
as
us
ed i
n r
es
t
or
at
i
o
n ol
d oi
l
p
ai
nt
i
ng.
T
hi
s
t
er
m
w
as
f
i
r
s
t
l
y
i
nt
r
od
uc
ed
i
nt
o d
i
g
i
t
a
l
i
m
age pr
oc
es
s
i
n
g i
n t
h
e
w
or
k
of
B
er
t
a
l
m
i
o
,
e
t
a
l.
, [
1
].
T
her
e ar
e
m
an
y
ap
pl
i
c
at
i
o
ns
of
i
m
age i
n
pai
nt
i
ng.
I
t
c
an be
us
ed
t
o r
em
ov
e s
c
r
at
c
hes
aut
om
at
i
c
al
l
y
f
r
om
di
gi
t
al
i
m
age
or
f
i
l
m
,
di
gi
t
al
r
es
t
or
at
i
o
n
of
anc
i
ent
p
ai
n
t
i
ng,
t
e
x
t
er
as
i
ng,
an
d
c
onnec
t
r
oa
ds
i
n s
at
e
l
l
i
t
e
i
m
age
[2
].
F
i
gur
e 1.
I
npa
i
nt
i
ng
P
r
o
bl
e
m
I
n 198
4,
G
em
an and G
em
an
us
ed
s
t
at
i
s
t
i
c
a
l
ap
pr
oac
h
f
or
i
m
age
r
es
t
or
at
i
o
n
t
as
k
s
[
3]
.
T
hi
s
appr
oac
h
w
as
t
o
r
e
m
ov
e noi
s
e
i
n t
he
i
m
age,
i
f
t
he
i
m
age s
t
r
uc
t
ur
e
w
a
s
m
i
s
s
i
ng,
t
hi
s
appr
o
ac
h
w
ou
l
d
b
e di
f
f
i
c
ul
t
t
o gi
v
e
f
ai
r
r
es
ul
t
.
T
he
w
o
r
k
of
B
er
t
al
m
i
o
,
e
t. a
l
.
,
i
nt
r
oduc
ed i
m
age
i
np
ai
n
t
i
n
g as
a ne
w
r
es
ea
r
c
h ar
ea of
di
gi
t
a
l
i
m
age pr
oc
es
s
i
ng
[
1]
.
T
hi
s
w
or
k
not
on
l
y
c
an
r
em
ov
e noi
s
e,
b
ut
al
s
o c
a
n r
epai
r
s
t
r
uc
t
ur
e of
t
he i
m
age.
T
he
m
odel
i
s
bas
e
d on n
onl
i
ne
ar
P
D
E
.
W
h
en t
he dam
age ar
ea
h
as
l
ar
ge d
i
s
t
a
nc
es
and c
o
m
pl
i
c
at
ed t
ex
t
ur
e
,
t
hi
s
m
et
hod
w
i
l
l
ha
v
e
di
f
f
i
c
ul
t
i
es
t
o r
ep
ai
r
i
m
age.
A
di
f
f
er
ent
appr
o
ac
h t
o i
n
p
ai
nt
i
n
g w
as
pr
opos
e
d b
y
C
han an
d S
h
en
[
4]
.
T
he
y
i
nt
r
oduc
ed
t
he
i
dea t
hat
w
e
l
l
-
k
now
n v
ar
i
at
i
on
al
i
m
age den
oi
s
i
ng
and s
egm
ent
at
i
on m
odel
s
c
an be
eas
i
l
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
14
,
N
o
.
3,
S
ept
em
ber
2016
:
10
09
–
1
015
1010
adap
t
ed
t
o
t
he
i
n
pai
nt
i
ng
t
as
k
b
y
a
s
i
m
pl
e
m
odi
f
i
c
at
i
o
n.
T
he
m
odel
c
an
s
uc
c
es
s
f
ul
l
y
pr
o
pag
at
e
s
har
p
edg
es
i
nt
o
t
h
e
dam
aged
dom
ai
n.
H
o
w
e
v
er
,
t
h
i
s
m
odel
ex
ac
t
s
a
pena
l
t
y
o
n
t
he
l
e
ngt
h
of
edges
,
t
h
i
s
t
ec
hni
qu
e c
an
not
c
onn
ec
t
c
ont
our
s
ac
r
os
s
l
ar
ge di
s
t
anc
es
.
O
ne
y
ear
af
t
er
t
hat
,
2002
,
C
han
,
K
a
ng,
a
nd
S
he
n i
nt
r
o
duc
e
d a ne
w
v
ar
i
at
i
ona
l
i
m
age i
np
ai
nt
i
ng m
odel
t
hat
addr
es
s
ed t
he s
hor
t
c
om
i
ngs
of
t
he t
he t
ot
al
v
ar
i
at
i
o
n
[
5]
.
T
hi
s
m
et
hod
m
a
k
es
i
m
pr
ov
em
ent
i
n
i
np
ai
n
t
i
n
g l
ong
er
di
s
t
a
nc
es
t
han
bef
or
e.
E
s
e
dog
l
u a
nd
S
he
n ad
apt
e
d i
m
age s
egm
ent
at
i
o
n t
o t
h
e
i
np
ai
n
t
i
n
g
pr
ob
l
em
[
6]
.
T
hi
s
m
odel
c
an
be
s
o
l
v
ed
r
at
her
qu
i
c
k
l
y
b
ut
t
h
i
s
m
odel
c
annot
c
on
nec
t
i
m
age ac
r
os
s
l
ar
ge
di
s
t
a
nc
e.
I
n 2
007
,
m
odi
f
i
ed C
ahn
H
i
l
l
i
ar
d
be
i
n
g us
e
d f
or
i
n
pa
i
nt
i
ng.
I
t
s
ho
w
n t
h
at
t
hi
s
m
et
ho
d g
i
v
e
f
as
t
i
npai
n
t
i
n
g
of
bi
nar
y
i
m
ages
[
2]
.
I
n
2009,
S
c
h
önl
i
eb
,
m
a
k
e
r
es
ear
c
h
i
n
m
oder
n
P
D
E
in
p
a
in
t
in
g
[
7]
.
T
hi
s
r
es
ear
c
h s
how
us
t
he ad
v
a
nt
ag
e of
f
our
t
h
-
or
der
i
npa
i
nt
i
n
g
m
odel
s
ov
e
r
m
odel
s
of
s
ec
ond di
f
f
er
ent
i
al
or
der
i
s
i
n t
he s
m
oot
h c
ont
i
n
uat
i
on of
i
m
age c
ont
ent
s
ev
en ac
r
os
s
l
ar
ge d
i
s
t
anc
e
i
n t
h
e i
m
age
.
T
he r
es
ear
c
h em
phas
i
s
e t
o m
a
k
e
m
or
e r
e
s
ear
c
h abo
ut
f
as
t
er
num
er
i
c
al
s
ol
v
er
s
f
or
i
np
ai
n
t
i
ng
w
i
t
h h
i
gh
er
-
or
der
e
qua
t
i
ons
t
han bef
or
e
.
T
he
m
os
t
r
ec
ent
appr
oac
h
t
o i
np
ai
nt
i
ng i
s
bas
ed o
n f
our
t
h or
der
P
D
E
m
et
hod
[
8]
.
T
her
e
ar
e
l
o
t
s
of
r
es
ear
c
hes
t
r
y
i
n
g
t
o
i
m
pr
ov
e
t
he
qu
al
i
t
y
an
d
t
o
r
ed
uc
e
pr
oc
es
s
i
ng
t
i
m
e
of
i
np
ai
n
t
i
n
g
pr
oc
es
s
us
i
ng h
i
gh
or
der
P
D
E
m
et
hod
[
5,
8]
.
A
nu
m
ber
o
f
m
et
hods
w
er
e
s
u
g
ges
t
ed
t
o
s
ol
v
e
i
np
ai
n
t
i
n
g pr
o
bl
em
,
m
an
y
o
f
w
hi
c
h
ar
e
bas
ed
on
ad
v
a
n
c
ed m
at
hem
at
i
c
al
t
ec
hn
i
qu
es
[1
, 2
], [4
-
6]
,
8,
9]
.
T
hi
s
r
es
ear
c
h
i
s
t
r
y
i
ng t
o ac
c
el
er
at
i
ng
i
n
p
a
i
nt
i
ng pr
oc
es
s
us
i
ng
d
i
f
f
er
ent
appr
o
ac
h t
han
adv
anc
e m
at
hem
at
i
c
al
t
ec
h
ni
q
ues
.
H
i
gh
per
f
or
m
anc
e
c
o
m
put
i
ng
o
n
G
P
U
us
i
n
g
C
U
D
A
l
i
br
ar
y
w
i
l
l
b
e
us
e
d
t
o
s
o
l
v
e
i
m
age
I
npa
i
nt
i
ng pr
ob
l
em
.
P
ar
al
l
e
l
i
z
ed ex
ec
ut
i
on of
hi
gh or
d
er
P
D
E
i
n i
m
age
I
np
ai
nt
i
n
g w
ou
l
d be
a
f
ea
s
i
bl
e
opt
i
m
i
z
at
i
on m
et
h
od.
W
e w
i
l
l
t
ak
e t
he
ad
v
ant
a
ges
of
G
P
U
m
ul
t
i
c
or
e
pr
oc
es
s
or
t
o
i
nc
r
eas
e t
he a
pp
l
i
c
at
i
on
p
er
f
or
m
anc
e b
y
ex
ec
ut
i
ng
t
hem
i
n G
P
U
c
or
es
[
10
]
.
T
he r
es
ear
c
h of
G
uo
and
H
e,
i
s
on
e
of
t
he
ex
am
pl
e
t
hat
t
he
us
e
of
G
P
U
has
be
en
s
i
g
ni
f
i
c
an
t
l
y
i
nc
r
eas
e
t
he
t
h
e
l
at
t
anc
y
t
i
m
e
i
n
d
i
gi
t
al
i
m
age
pr
oc
es
s
i
n
g.
T
he
y
c
a
n
s
peed
up
f
r
ac
t
al
i
m
age
c
om
p
r
es
i
on
pr
oc
es
s
t
o
123
t
i
m
es
f
as
t
er
c
o
m
par
e
t
o
t
he
t
r
adi
t
i
o
na
l
m
et
hod
[
11]
.
T
hat
r
es
ear
c
h
gi
v
es
an
opp
or
t
un
i
t
y
anot
her
i
m
age pr
oc
es
s
i
ng
al
g
or
i
t
m
t
o us
e t
h
e po
w
er
o
f
G
P
U
t
o s
pee
dup
c
om
put
at
i
on
pr
oc
es
s
.
T
hi
s
r
es
ear
c
h w
i
l
l
a
ns
w
er
ho
w
s
i
gn
i
f
i
c
ant
t
he us
es
of
G
P
U
t
o i
nc
r
eas
e t
h
e l
at
t
ec
y
t
i
m
e
i
n d
i
g
i
t
a
l
i
m
age
i
np
ai
n
t
i
ng u
s
i
ng f
our
t
h
or
der
P
D
E
.
2
.
I
n
p
a
i
n
ti
n
g
M
e
th
o
d
I
npa
i
nt
i
ng
d
i
g
i
t
a
l
i
m
age
i
s
es
s
ent
i
a
l
l
y
i
nt
er
po
l
at
i
on
pr
oc
es
s
on
f
i
l
l
i
ng
m
i
s
s
i
ng
par
t
s
of
an
i
m
age bas
ed on s
ur
r
ou
nd
i
ng ar
eas
.
T
he s
ol
ut
i
on
of
t
hi
s
pr
ob
l
em
i
s
us
i
ng f
our
t
h or
der
P
D
E
equa
t
i
o
n.
T
he
i
np
ai
nt
i
ng
pr
obl
em
i
s
t
o
t
o
r
ec
ons
t
r
uc
t
t
he
or
i
gi
n
al
i
m
age
u
i
n
(
da
m
aged)
dom
ai
n
D
⊂
Ω
(F
i
g
ur
e
1)
.
I
m
age f
r
epr
es
ent
s
om
e
gi
v
en
i
m
age
d
ef
i
ned
o
n
an
i
m
age
dom
ai
n
Ω
.
T
he
di
s
c
r
et
i
z
at
i
on
i
n
s
pac
e
us
ed
f
i
ni
t
e
d
i
f
f
er
enc
e
and
s
pec
t
r
al
m
et
hods
,
i
.
e.
,
t
h
e
f
as
t
F
our
i
er
t
r
ans
f
or
m
,
t
o
s
i
m
pl
i
f
y
t
he
i
nv
er
s
i
on
of
t
he
Lap
l
ac
i
an
∆
f
or
t
he
c
om
put
at
i
o
n
of
+
1
.
T
he
f
our
t
h
or
der
P
D
E
i
np
ai
nt
i
ng equ
at
i
on:
=
∆
−
∆
+
1
′
(
)
+
(
−
)
,
Ω
,
(
1)
T
he
E
q
u
at
i
on
(
1
)
i
s
t
he
w
o
r
k
of
B
et
r
o
z
z
i
,
e
t
a
l.
,
us
i
n
g
t
he
f
our
t
h or
der
P
D
E
t
o
s
ol
v
e
i
np
ai
n
t
i
n
g
pr
obl
em
[
2]
.
B
as
ed
o
n
t
he
w
or
k
o
f
E
q
uat
i
on
(
1
)
,
S
c
hö
nl
i
eb
,
e
t
al
.,
[
9]
b
ui
l
t
an
ot
h
er
f
our
t
h
or
der
P
D
E
eq
uat
i
on:
+
1
−
∆
+
1
∆
∆
+
1
+
2
+
1
=
1
∆
∆
+
1
−
Δ
∇
.
∆
|
∆
|
+
2
+
(
−
)
(
2)
T
he c
ons
t
ant
s
C
1
and C
2
ar
e pos
i
t
i
v
e,
an
d ne
ed t
o
be c
hos
en l
ar
ge e
nou
gh t
o m
a
k
e t
hi
s
equa
t
i
o
n c
on
v
ex
.
T
he c
ons
t
ant
s
C
1
a
nd C
2
m
us
t
f
ul
f
i
l
l
t
he c
on
di
t
i
on
of
C
1
>
1/
ϵ
,
C
2
>
λ
_0.
T
hi
s
equat
i
o
n h
as
be
en
pr
o
v
ed
b
y
S
c
hön
l
i
eb
,
et
al
.,
[
9]
ha
v
e c
ons
i
s
t
e
nc
y
,
unc
o
ndi
t
i
on
a
l
s
t
abi
l
i
t
y
,
and c
o
nv
er
ge
nc
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
G
P
U
C
U
D
A
A
c
c
el
er
at
ed I
m
age I
npa
i
nt
i
n
g
us
i
ng F
o
ur
t
h
O
r
der
P
D
E
E
qua
t
i
on
(
E
dw
i
n P
r
a
nant
a
)
1011
3
.
G
P
U CUD
A
C
om
m
odi
t
y
c
om
put
er
gr
ap
hi
c
s
c
hi
ps
,
k
no
w
n
gen
er
i
c
a
l
l
y
as
G
r
ap
hi
c
s
P
r
oc
es
s
i
ng
U
ni
t
s
or
G
P
U
s
,
ar
e pr
oba
bl
y
t
oda
y
’
s
m
os
t
po
w
er
f
ul
c
om
put
at
i
on
al
har
d
w
ar
e f
or
t
he
do
l
l
ar
.
R
es
ear
c
her
s
and
de
v
e
l
op
er
s
hav
e
bec
om
e i
nt
er
es
t
ed i
n
har
n
es
s
i
ng
t
hi
s
po
w
er
f
or
gener
al
-
pur
pos
e c
om
put
i
ng,
an
ef
f
o
r
t
k
now
n c
o
l
l
ec
t
i
v
e
l
y
as
G
P
G
P
U
(
f
or
“
G
ener
al
-
P
ur
p
os
e
c
om
put
i
ng o
n
t
he G
P
U
”
)
[
10]
.
C
U
D
A
i
s
a
par
a
l
l
e
l
pr
ogr
a
m
m
i
ng m
odel
de
v
e
l
op
ed
b
y
N
V
I
D
I
A
w
as
s
t
ar
t
ed
at
2
0
06.
T
he
f
i
r
s
t
C
U
D
A
S
D
K
w
as
r
el
ea
s
ed i
n t
h
e ear
l
y
2
007.
A
p
ar
al
l
el
s
y
s
t
em
us
i
ng C
U
D
A
c
ons
i
s
t
s
of
a
hos
t
(
C
P
U
)
and
a
d
ev
i
c
e
(
G
P
U
)
.
T
he
c
om
put
at
i
on
of
t
as
k
s
i
s
done
i
n
G
P
U
b
y
a
s
et
of
t
hr
eads
t
hat
r
un i
n p
ar
al
l
e
l
.
T
he G
P
U
ar
c
hi
t
ec
t
ur
e f
or
t
hr
eads
c
ons
i
s
t
of
t
w
o
-
l
ev
el
h
i
er
ar
c
h
y
,
nam
el
y
bl
oc
k
and gr
i
d (
F
i
g
ur
e
2)
.
B
l
oc
k
i
s
a s
et
of
t
i
ght
l
y
c
o
up
l
ed t
hr
eads
w
her
e
eac
h t
hr
e
ad i
s
i
d
ent
i
f
i
ed
b
y
a
t
hr
ea
d
I
D
,
w
h
i
l
e
gr
i
d
i
s
a
s
et
of
l
oos
el
y
c
oup
l
e
d
of
bl
oc
k
s
w
i
t
h
s
i
m
i
l
ar
s
i
z
e
and
d
i
m
ens
i
on
[
10]
.
T
he us
age
of
G
P
U
s
h
o
w
s
i
m
pr
ov
em
ent
per
f
or
m
anc
e i
n
w
i
de
ar
e
a s
uc
h
as
P
h
y
s
i
c
a
l
bas
ed s
i
m
ul
at
i
on,
s
i
g
nal
p
r
oc
es
s
i
ng,
dat
a pr
oc
es
s
i
ng
,
i
m
age s
egm
ent
at
i
on
,
c
om
put
er
v
i
s
i
o
n
and
i
m
age pr
oc
es
s
i
ng
[
10]
.
A
n ex
am
pl
e
of
t
he us
a
ge
G
P
U
f
or
i
m
age pr
oc
es
s
i
ng
i
s
t
h
e
w
or
k
of
G
uo
a
nd
H
ei
,
t
h
e
y
w
or
k
ed
f
r
ac
t
al
i
m
age
c
om
pr
es
s
i
on
i
n
G
P
U
[
11]
.
T
hei
r
w
or
k
c
an
ac
c
e
l
er
at
e
123
t
i
m
es
f
as
t
er
t
han
bef
or
e.
N
o
t
o
nl
y
i
n
i
m
age
pr
oc
es
s
i
ng t
he
us
ag
es
of
G
P
U
b
e
c
o
m
e pop
ul
ar
,
but
al
s
o
i
n
d
at
a
pr
oc
es
s
i
ng
.
I
t
s
h
o
w
n
i
n
t
h
e
w
or
k
of
X
u
an
d
X
u,
t
h
e
y
w
or
k
ed
a
h
y
b
r
i
d
s
hor
t
i
n
g
al
g
or
i
t
m
w
i
t
h t
h
e po
w
er
of
G
P
U
an
d C
P
U
.
T
hi
s
w
or
k
c
an s
hor
t
o
ne b
i
l
l
i
on
32
-
b
i
t
f
l
oat
i
n no m
or
e
t
han
5 s
ec
on
ds
[
12]
.
F
i
gur
e 2.
C
U
D
A
A
r
c
hi
t
ec
t
ur
e
4
.
G
P
U
C
o
m
b
i
n
e
d
P
a
r
a
l
l
e
l
I
m
a
g
e
I
n
p
a
i
n
ti
n
g
T
he ac
c
el
er
at
i
o
n i
n
pai
nt
i
ng
pr
oc
es
s
i
s
a c
hal
l
ag
ne t
o
r
es
ear
c
her
s
.
T
he
y
ar
e t
r
y
i
ng t
o
us
e ad
v
a
nc
e
m
at
e
m
at
hi
c
al
t
eqn
i
qu
es
t
o
i
m
pr
ov
e t
he
c
om
put
at
i
on
t
i
m
e but
t
h
er
e
anot
her
w
a
y
t
o
ac
c
el
er
at
e
i
npa
i
nt
i
ng
pr
oc
e
s
s
us
i
ng
p
ar
al
l
e
l
ex
ec
ut
i
o
n
i
n
G
P
U
.
W
e
r
eal
i
z
e
t
hat
t
h
e
us
e
of
G
P
U
i
n
i
np
ai
n
t
i
n
g pr
o
bl
em
w
i
l
l
r
e
duc
e t
h
e c
om
put
at
i
on
l
at
a
n
c
y
t
i
m
e.
I
m
age s
our
c
e and
m
as
k
i
npa
i
nt
i
ng
s
hou
l
d
be
i
n
i
s
i
a
l
i
z
e
i
n t
he
C
P
U
m
e
m
or
y
.
I
n G
P
U
m
e
m
o
r
y
,
d
at
a
o
nl
y
c
an
be
pr
oc
es
s
i
n
one
d
i
m
ens
i
ona
l
ar
r
a
y
.
I
m
age
dat
a
s
hou
l
d
be
pr
oc
es
s
t
o
one
di
m
ens
i
on
al
ar
r
a
y
d
at
a.
W
e w
i
l
l
as
s
i
gn
i
m
age s
or
uc
e a
nd m
as
k
t
o
v
ar
i
abl
e t
h
at
c
a
n b
e
t
r
ans
f
er
t
o G
P
U
me
mo
r
y
.
T
he
t
hr
eads
of
a
bl
oc
k
ar
e
pr
oc
es
s
ed
i
n
t
he
s
am
e
s
t
r
eam
i
ng
pr
oc
es
s
or
(
S
M)
,
a
nd
t
he
par
al
l
e
l
i
z
at
i
o
n
d
epe
nds
o
n
t
he
num
ber
of
c
or
es
i
n
t
he
S
M.
T
he
G
P
U
ar
c
hi
t
ec
t
ur
e
has
a
l
i
m
i
t
i
n
t
he n
um
ber
of
t
hr
eads
p
er
bl
oc
k
.
I
n t
h
i
s
w
or
k
,
a N
V
I
D
I
A
G
eF
or
c
e G
T
X
-
670
c
ar
d w
i
t
h up t
o
1
,
02
4
t
hr
eads
p
er
b
l
oc
k
w
as
us
e
d
.
T
he c
o
m
put
at
i
ons
of
f
as
t
F
our
i
er
t
r
ans
f
or
m
us
e t
he C
U
D
A
l
i
br
ar
y
.
T
he s
t
ep t
o c
al
l
C
U
F
F
T
l
i
br
ar
y
i
s
:
c
r
eat
e p
l
an,
ex
ec
u
t
e
pl
a
n a
nd
des
t
r
o
y
pl
a
n.
T
he c
om
put
at
i
ons
nee
d d
oub
l
e
p
r
e
s
ic
io
n
s
o w
e w
i
l
l
us
e par
am
et
er
C
U
F
F
T
_
Z
2
Z
,
w
hi
c
h i
nd
i
c
at
e d
oub
l
e c
om
pl
ex
t
o d
oub
l
e
c
o
m
pl
ex
c
om
put
at
i
on
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
14
,
N
o
.
3,
S
ept
em
ber
2016
:
10
09
–
1
015
1012
S
t
ep
1
:
I
m
age and m
as
k
i
ni
s
i
a
l
i
z
at
i
on
S
t
ep
2
:
P
r
epr
oc
es
s
i
n
g d
at
a t
o t
r
ans
f
er
G
P
U
S
t
ep
3
:
C
om
put
at
i
o
n V
ar
i
a
bl
e
i
n
i
s
i
a
l
i
z
at
i
on
S
t
ep
4
:
T
r
ans
f
er
dat
a f
r
o
m
C
P
U
t
o
G
P
U
M
em
or
y
----
B
egi
n
par
a
l
l
el
dat
a c
o
mput
at
i
on
i
n G
P
U
----
S
t
ep
5
:
G
P
U
v
ar
i
a
bl
e an
d
w
or
k
l
oad
i
n
i
s
i
a
l
i
z
at
i
on
S
t
ep
6
:
I
ni
z
i
al
i
s
at
i
on C
al
c
u
l
at
i
on
P
a
r
am
et
er
S
t
ep
7
:
P
ar
a
l
l
e
l
i
m
age and m
as
k
F
F
t
r
ans
f
or
m
S
t
ep
8
:
P
ar
a
l
l
e
l
c
a
l
c
ul
at
i
on c
ur
v
e
S
t
ep
9 :
P
ar
a
l
el
c
ur
v
e
F
F
t
r
ans
f
or
m
S
t
ep
10
:
P
ar
a
l
l
e
l
F
o
ur
t
h O
r
der
P
D
E
I
npa
i
nt
i
ng
i
n s
p
ec
t
r
al
dom
ai
n
S
t
ep
11
P
ar
a
l
l
e
l
I
nv
er
s
e F
F
t
r
ans
f
or
m
o
f
i
npai
nt
i
ng r
es
u
l
t
S
t
ep
12
:
I
ni
s
i
al
i
s
at
i
on n
e
w
c
ur
v
e
St
e
p
9
-
12
w
i
l
l
b
e r
ep
eat
e
d u
nt
i
l
f
a
i
r
i
m
age c
o
ndi
t
i
o
n
S
t
ep
13
:
T
r
ans
f
er
dat
a f
r
o
m
G
P
U
t
o
C
P
U
Mem
or
y
S
t
ep
14
:
C
el
a
n G
P
U
Mem
or
y
----
E
nd par
al
l
el
dat
a c
om
p
ut
at
i
on
i
n G
P
U
----
S
t
ep
15
:
P
r
epr
oc
es
s
i
n
g d
at
a t
o C
P
U
S
t
ep
16
:
S
ho
w
an
d s
a
v
e
i
m
age
I
npa
i
nt
i
ng R
es
ul
t
F
i
gur
e 3.
A
l
g
or
i
t
m
par
al
l
e
l
I
npa
i
nt
i
ng f
or
t
h
or
der
P
D
E
i
n G
P
U
W
e
us
e
t
he
d
at
a
v
ar
i
ab
l
e
f
r
om
t
he
r
es
ear
c
h
of
S
c
hö
nl
i
eb
,
et
al
.,
[
9
]
.
T
he
C
1
v
ar
i
a
bl
e
i
s
s
et
t
o 2
00
and
C
2
v
ar
i
ab
l
e
i
s
100
.
W
e s
av
e t
hi
s
v
ar
i
ab
l
e
i
n r
e
gi
s
t
er
m
e
m
or
y
of
G
P
U
.
T
he c
ur
v
e
c
al
c
ul
a
t
i
o
n r
es
u
l
t
a
nd t
he f
o
ur
t
h or
d
er
i
n
pa
i
nt
i
ng r
es
ul
t
w
i
l
l
b
e s
av
ed
i
n
gl
oba
l
m
e
m
or
y
.
A
f
t
er
t
he
i
t
er
at
i
o
n
c
om
pl
et
e,
t
he G
P
U
w
i
l
l
b
egi
n
t
o t
r
a
ns
f
er
t
he dat
a t
o
C
P
U
.
T
he dat
a
w
i
l
l
b
e pr
oc
es
s
f
r
o
m
t
he on
e di
m
ens
i
on
al
ar
r
a
y
t
o t
w
o
di
m
ens
i
ona
l
i
m
age ar
r
a
y
.
5
.
R
e
s
u
l
t I
n
p
a
i
n
ti
n
g
F
o
u
r
t
h
O
r
d
e
r
G
P
U
T
he pr
opos
ed m
et
hod w
a
s
s
i
m
ul
at
ed
w
i
t
h an i
m
age of
w
hi
t
e r
ec
t
angu
l
ar
on
bl
ac
k
bac
k
gr
ound
ha
v
i
ng
t
h
e
da
m
age
i
n
t
he
m
i
ddl
e
of
t
he
i
m
age.
T
he
dam
age
ar
ea
h
ad
gr
a
y
c
o
l
our
.
T
hi
s
i
m
age r
es
ol
ut
i
on
w
as
100 x
100 p
i
x
el
s
.
T
he i
m
age t
es
t
c
an be s
een
i
n F
i
g
ur
e
3(
a)
w
hi
l
e t
h
e
r
equi
r
e
d
r
es
u
l
t
s
h
ou
l
d
be s
een
as
i
n F
i
g
ur
e
3(
f
)
.
T
hi
s
dam
age ar
ea
c
an
be
eas
i
l
y
s
een
i
n
F
ig
ur
e
3(
a)
w
i
t
h
gr
a
y
c
ol
our
i
n t
he
m
i
ddl
e of
t
he i
m
age.
T
hi
s
i
ni
t
i
al
t
es
t
i
m
age h
ad h
i
gh
M
S
E
(
9,
79
5.
40)
,
l
o
w
P
S
N
R
(
8.
22)
and
l
o
w
S
S
I
K
(
0.
81)
.
T
hes
e i
nd
i
c
at
or
s
i
ndi
c
at
ed t
h
at
t
he i
ni
t
i
a
l
i
m
age had hi
gh
dam
age and
s
m
al
l
s
i
m
i
l
ar
i
t
y
.
T
abl
e 1.
Lat
e
nc
y
I
np
ai
n
t
in
g
pr
oc
es
s
C
P
U
and G
P
U
I
t
er
at
i
on
s
Lat
enc
y
(
s
e
c
)
S
peedup
CP
U
GP
U
10,
000
155.
99
4.
31
36.
19
50,
000
780.
00
21.
65
36.
03
100,
000
1,
560.
16
43.
37
35.
98
200,
000
3,
120.
28
86.
50
36.
07
T
he
l
at
e
nc
y
of
i
npa
i
nt
i
n
g
p
r
oc
es
s
us
i
ng
f
our
t
h
or
der
P
D
E
has
bee
n
n
ot
ed
bo
t
h
i
n
C
P
U
and G
P
U
t
o r
epa
i
r
t
he da
m
age ar
ea.
A
s
s
how
n i
n
T
abl
e 1,
t
he C
P
U
c
om
put
at
i
o
n t
i
m
e w
as
s
l
o
w
er
t
h
an G
P
U
c
om
put
a
t
i
on t
i
m
e.
T
he c
o
m
put
at
i
o
n t
i
m
e f
or
10,
000 i
t
er
at
i
on
s
i
n C
P
U
w
as
155.
99s
an
d i
n G
P
U
w
as
4.
31s
.
T
hes
e
w
er
e
t
he
c
om
put
at
i
o
n t
i
m
e f
or
100,
00
0,
0
00 da
t
a (
1
00 x
100 x
10,
000
i
t
er
a
t
i
o
ns
)
.
T
h
e c
om
put
at
i
on t
i
m
e f
or
200,
000 i
t
er
at
i
ons
i
n C
P
U
w
as
3,
12
0.
28s
a
nd
i
n G
P
U
w
as
8
6.
5
0s
.
T
hes
e
w
er
e t
he c
om
put
at
i
on t
i
m
e
f
or
2,
000,
0
00,
000
dat
a
(
100 x
10
0 x
200,
000 i
t
er
at
i
ons
)
.
T
he par
al
l
el
c
om
put
at
i
on us
i
ng
G
P
U
w
as
s
ho
w
i
ng 36x
l
a
t
enc
y
s
pee
dup
t
han
us
i
n
g C
P
U
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
G
P
U
C
U
D
A
A
c
c
el
er
at
ed I
m
age I
npa
i
nt
i
n
g
us
i
ng F
o
ur
t
h
O
r
der
P
D
E
E
qua
t
i
on
(
E
dw
i
n P
r
a
nant
a
)
1013
(
a)
(
d)
(
b)
(
e)
(c
)
(f
)
F
i
gur
e 4.
(
a)
T
es
t
i
m
age w
i
t
h dam
age i
n t
h
e m
i
ddl
e
(
b)
.
I
npa
i
nt
i
n
g 10
.
00
0
i
t
er
at
i
ons
(
c
)
I
np
ai
nt
i
ng
50.
000
i
t
er
at
i
o
ns
(
d)
.
I
n
pai
nt
i
n
g 10
0.
00
0 i
t
er
at
i
ons
(
e)
.
I
npa
i
nt
i
ng
20
0.
00
0
i
t
er
at
i
on
s
(
f
)
T
es
t
i
m
age w
i
t
h
out
da
m
age
T
he
G
PU
hav
e
bee
n
ap
pl
i
ed
t
o
s
o
l
v
e
i
n
pa
i
nt
i
ng
pr
obl
em
us
i
ng
f
our
t
h
or
der
P
D
E
equa
t
i
o
n
i
n
E
q
uat
i
on
(
2
)
.
T
he r
es
u
l
t
of
t
he
10
,
000
i
t
er
at
i
o
ns
i
s
s
h
o
w
n
i
n F
i
g
ur
e
3
(
b)
.
T
hi
s
i
m
age
w
as
t
r
y
i
ng
t
o
r
ec
ons
t
r
uc
t
t
h
e
w
hi
t
e
r
ec
t
a
ngu
l
ar
i
n
t
he
m
i
ddl
e
of
t
he
i
m
age.
I
n
1
0,
000
i
t
er
at
i
o
ns
,
t
he r
ec
t
a
ng
ul
ar
al
m
os
t
c
onnec
t
e
d.
I
n 5
0
,
00
0
i
t
er
at
i
ons
as
s
ho
w
n
i
n F
i
g
ur
e
3(
c
)
,
t
he
w
hi
t
e
r
ec
t
angu
l
ar
a
l
r
ea
d
y
c
onn
e
c
t
ed an
d t
h
e b
l
ac
k
bac
k
g
r
ound
w
as
r
ec
ons
t
r
uc
t
e
d.
A
f
t
er
100,
000
i
t
er
at
i
ons
,
w
e
c
an s
e
e
onl
y
s
m
al
l
ar
ea
of
bl
ac
k
bac
k
gr
ound
t
ha
t
h
as
n’
t
r
ep
ai
r
e
d
l
i
k
e
in
F
i
g
ur
e
3(
d)
.
I
n
20
0,
00
0
i
t
er
at
i
o
ns
,
F
i
g
ur
e
3(
e)
a
l
r
ea
d
y
has
t
he
s
am
e s
t
r
uc
t
ur
e l
i
k
e i
n
F
i
g
u
re
3
(
f
)
.
T
h
is
is
pr
ov
e
d b
y
t
he
S
S
I
K
v
a
l
ue
o
f
0.
98.
F
i
gur
e 5.
L
i
n
e c
har
t
of
l
at
a
n
c
y
c
om
par
i
s
on G
P
U
and
C
P
U
T
abl
e 2.
I
m
age Q
u
al
i
t
y
Me
t
r
i
c
I
t
er
at
i
on
s
MS
E
PSN
R
S
SI
K
0
9,
795.
40
8.
22
0.
81
10,
000
7,
628.
24
9.
30
0.
74
50,
000
2,
663.
89
13.
88
0.
84
100,
000
663.
59
19.
91
0.
88
200,
000
29.
25
33.
47
0.
98
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
14
,
N
o
.
3,
S
ept
em
ber
2016
:
10
09
–
1
015
1014
T
he
out
c
om
e
of
i
npai
t
i
n
g
p
r
oc
es
s
us
i
ng
f
our
t
h
or
der
P
D
E
as
s
i
m
ul
at
ed
b
y
t
h
e
a
ut
hor
s
i
s
s
ho
w
n
i
n
F
i
g
ur
e
3(
e)
a
n
d
t
he
r
eq
ui
r
e
d
r
es
ul
t
w
as
s
ho
w
n
i
s
F
ig
ur
e
3(
f
)
.
I
t
w
as
f
ound
t
hat
t
he
i
t
er
at
i
ons
i
n
G
P
U
a
nd
t
he
qua
l
i
t
y
of
t
he
i
m
age
ar
e
c
or
r
el
at
ed.
I
f
w
e
i
nc
r
eas
e
t
h
e
i
t
er
at
i
on,
t
h
e
qua
l
i
t
y
of
t
he i
m
age w
i
l
l
i
nc
r
eas
e.
I
t
w
as
s
how
n i
n T
abl
e 2
,
t
hat
af
t
er
200,
0
00 i
t
er
a
t
i
ons
t
he M
S
E
dec
r
eas
ed
f
r
o
m
9,
795.
40
t
o 29
.
2
5
,
P
S
N
R
i
nc
r
eas
ed
f
r
o
m
8.
22 t
o
33,
47
an
d
S
S
I
K
i
nc
r
eas
ed
f
r
o
m
0.
81 t
o 0,
98.
W
e
us
e
s
a
m
pl
e
i
m
age
t
o
t
es
t
t
he
r
eal
pr
o
bl
em
r
epar
i
ng
t
he
dam
age
pi
c
t
ur
e.
W
e
us
ed
t
he t
hr
ee
g
i
r
l
s
ol
d
gr
a
y
s
c
a
l
e
por
t
r
ai
t
t
h
at
ha
v
e
f
ol
d
ed
dam
age
l
i
k
e
i
n F
i
gur
e
6
(
a)
.
T
hi
s
pi
c
t
ur
e
d
i
m
ens
i
on
i
s
483
x
405
pi
x
el
s
.
W
e ar
e r
unni
n
g t
h
e a
l
g
or
i
t
m
ex
ec
ut
i
on
i
n
bot
h
C
P
U
and
G
P
U
.
T
abl
e 3
.
Lat
e
nc
y
I
n
pa
i
nt
i
ng
ol
d
por
t
r
a
i
t
l
at
anc
y
C
P
U
a
nd G
P
U
I
t
er
at
i
on
s
Lat
enc
y
(
s
e
c
)
S
peedup
CP
U
GP
U
500
124,
80
2,
58
48,
39
1.
000
249,
61
5,
12
48,
76
2.
000
499,
29
10,
21
48,
91
S
i
m
i
l
ar
f
r
om
t
he s
i
m
ul
at
i
on
bef
or
e
,
t
he
C
P
U
c
om
put
a
t
i
on
t
i
m
e w
as
s
l
o
w
er
t
h
an
G
P
U
c
o
m
put
at
i
on t
i
m
e.
T
he c
o
m
put
at
i
on t
i
m
e f
or
500
i
t
er
at
i
o
ns
i
n C
P
U
w
as
12
4,
80
s
and i
n G
P
U
w
as
2,
58
s
.
T
hes
e
w
er
e
t
h
e
c
om
put
at
i
on
t
i
m
e
f
or
97.
807.
5
00
d
at
a
(
4
83
x
4
05
x
500
i
t
er
at
i
o
ns
)
.
T
he
c
o
m
put
at
i
on
t
i
m
e
f
or
1.
00
0
i
t
er
at
i
ons
i
n
C
P
U
w
a
s
249,
6
1
s
an
d
i
n
G
P
U
w
a
s
5,
12
s
.
T
hes
e
w
er
e t
he c
om
put
at
i
o
n t
i
m
e
f
or
195.
615.
000
da
t
a (
483
x
405
x
1.
000 i
t
er
at
i
o
ns
)
.
T
he c
om
put
at
i
on
t
i
m
e
f
or
2.
000
i
t
er
at
i
ons
i
n
C
P
U
w
as
499,
29
s
a
nd
i
n
G
P
U
w
as
10,
21
s
.
T
h
es
e
w
er
e
t
he
c
o
m
put
at
i
on
t
i
m
e
f
or
391.
2
30.
0
00d
at
a
(
48
3
x
405
x
2.
000
i
t
er
at
i
ons
)
.
T
he
par
a
l
l
el
c
om
put
at
i
o
n
us
i
ng G
P
U
w
as
s
ho
w
i
ng
48
x
l
at
e
nc
y
s
pe
edu
p t
h
an
us
i
ng C
P
U
.
(
a)
(c
)
(
b)
(
d)
F
i
gur
e 6.
(
a)
D
am
age pi
c
t
ur
e
;
(
b)
I
n
pai
nt
i
ng 500 i
t
er
at
i
o
ns
;
(
c
)
I
npai
nt
i
ng
1.
00
0 i
t
er
at
i
o
ns
;
(
d)
I
npa
i
nt
i
ng
2.
0
00
i
t
er
a
t
i
o
ns
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
G
P
U
C
U
D
A
A
c
c
el
er
at
ed I
m
age I
npa
i
nt
i
n
g
us
i
ng F
o
ur
t
h
O
r
der
P
D
E
E
qua
t
i
on
(
E
dw
i
n P
r
a
nant
a
)
1015
T
he
G
PU
hav
e b
een a
ppl
i
ed t
o s
ol
v
e i
npa
i
nt
i
ng pr
o
bl
em
i
n F
i
gur
e 6(
a
)
us
i
n
g f
our
t
h
or
der
P
D
E
eq
uat
i
on
i
n
E
q
ua
t
i
on
(
2
)
.
T
he
r
es
u
l
t
of
t
he
500
it
e
r
a
t
io
n
s
wa
s
sh
o
wn
i
n
F
i
g
ur
e
6
(b
).
T
hi
s
i
m
age
w
as
t
r
y
i
n
g
t
o
r
e
pai
r
s
m
al
l
dam
age
i
n
t
h
e
b
od
y
an
d
t
h
e
bac
k
gr
ound,
b
ut
w
e
c
a
n
s
t
i
l
l
s
ee
t
he
dam
age
i
n
t
he
ar
e
a
ar
ound
t
h
e
gi
r
l
e
y
es
.
In
1
.
000
i
t
er
at
i
ons
,
t
he
d
am
age
i
n
t
he
gi
r
l
e
ye
r
educ
e t
hi
s
w
as
s
ho
w
n i
n F
i
gur
e
6(
c
)
.
B
ut
t
he
f
ac
e l
i
ne
of
t
he gi
r
i
l
i
s
not
c
on
nec
t
e
d
y
et
. In
2.
0
00
t
he
f
ac
e l
i
ne
ar
ou
nd t
he
gi
r
l
s
e
y
e
al
r
e
ad
y
c
on
nec
t
ed
l
i
k
e
w
as
s
ho
w
n i
n F
i
g
ur
e
6(
d)
.
6
.
C
o
n
c
l
u
s
io
n
I
n t
h
i
s
p
aper
,
t
he
aut
hor
s
pr
es
ent
ed
a t
ec
hni
que
f
or
ac
c
el
er
at
i
ng
i
npa
i
nt
i
n
g
p
r
obl
em
us
i
ng G
P
U
as
c
om
put
at
i
o
n
t
ool
s
.
T
hi
s
m
et
hod
ai
m
s
t
o r
educ
e t
he h
ea
v
y
c
om
put
at
i
on
pr
ob
l
em
of
f
our
t
h
or
der
P
D
E
a
nd
t
o
e
m
pl
o
y
t
he
us
e
of
G
P
U
i
n
p
ar
al
l
el
i
s
m
t
o
gai
n
c
om
put
at
i
on
s
pe
ed.
T
hi
s
i
m
pr
ov
em
ent
w
as
ac
c
om
pl
i
s
he
d b
y
us
i
ng N
V
D
I
A
G
E
F
O
R
C
E
G
T
X
670 t
o
c
a
l
c
ul
at
e
dat
a i
n
par
al
l
e
l
c
om
put
at
i
on.
T
he
pr
opos
e
d
c
onc
e
pt
pr
ov
ed
a
s
pee
du
p
t
o
3
6x
i
n
t
h
e
s
i
m
ul
at
i
on
m
odel
and s
p
eed
up
t
o 4
8x
i
n t
h
e
s
i
m
ul
at
i
on
pi
c
t
ur
e.
I
t
w
as
a
l
s
o s
ho
w
n
t
hat
t
h
e us
e
of
G
P
U
C
U
D
A
c
a
n
be a
ppl
i
e
d i
n t
h
e i
m
age i
np
ai
nt
i
n
g ar
e
a.
R
ef
er
en
ces
[1
]
M
ar
c
el
o
B
er
t
al
m
i
o,
G
ui
l
l
er
m
o
S
api
r
o,
V
i
nc
e
nt
C
as
e
l
l
e
s
,
C
ol
om
a
B
a
l
l
e
s
t
er
.
I
m
ag
e
i
npa
i
nt
i
ng.
P
r
oc
ee
di
n
gs
of
t
he
27t
h
an
nu
al
c
onf
er
en
c
e
on
C
o
m
p
ut
er
gr
aphi
c
s
and
i
n
t
er
a
c
t
i
v
e
t
ec
hni
q
ues
.
200
0:
417
-
42
4.
[2
]
A
ndr
ea L B
er
t
oz
z
i
,
S
el
i
m
E
s
e
dogl
u
,
A
l
an G
i
l
l
et
t
e
.
I
npa
i
nt
i
ng
of
bi
nar
y
i
m
a
ges
u
s
i
ng t
he C
ahn
-
H
il
lia
r
d
equat
i
on.
I
E
E
E
T
r
a
ns
a
c
t
i
ons
o
n i
m
age
pr
oc
es
s
i
n
g
.
20
07;
16(
1)
:
285
-
291
.
[3
]
S
t
uar
t
G
em
an,
D
o
nal
d
G
e
m
an.
S
t
oc
has
t
i
c
r
el
ax
at
i
on,
G
i
bbs
d
i
s
t
r
i
bu
t
i
on
s
,
a
nd
t
he
B
ay
es
i
an
r
es
t
or
at
i
o
n
of
i
m
ag
es
.
P
at
t
er
n
A
nal
y
s
i
s
and
M
ac
hi
ne
I
nt
el
l
i
genc
e,
I
E
E
E
T
r
an
s
a
c
t
i
o
ns
on
.
198
4;
6:
721
-
74
1.
[4
]
T
ony
C
han,
J
S
hen.
M
at
hem
a
t
i
c
a
l
m
od
el
s
f
or
l
o
c
al
n
ont
ex
t
ur
e i
npa
i
nt
i
ngs
.
S
I
A
M
J
our
nal
o
n A
ppl
i
ed
M
at
hem
at
i
c
s
.
2
001;
62(
3
)
:
10
1
9
-
1043
.
[5
]
T
ony
F
C
ha
n,
S
ung
H
a
K
ang,
J
i
an
hon
g
S
hen.
E
ul
er
’
s
el
a
s
t
i
c
a
and
c
ur
v
at
ur
e
-
ba
s
ed
i
m
a
ge
i
npai
n
t
i
n
g.
S
I
A
M
J
our
nal
o
n A
pp
l
i
ed
M
at
hem
at
i
c
s
.
20
02;
63(
2)
:
564
-
592.
[6
]
S
el
i
m
E
s
ed
ogl
u,
J
i
anh
ong S
h
en.
D
i
gi
t
al
i
npa
i
nt
i
ng ba
s
ed
on
t
he M
um
f
or
d
–
S
h
ah
–
E
u
l
er
i
m
age m
odel
.
E
ur
opea
n J
our
n
a
l
of
A
pp
l
i
ed
M
at
hem
at
i
c
s
.
2
002;
13(
4
)
:
35
3
-
370.
[7
]
C
ar
ol
a
-
B
i
bi
a
ne S
c
h
önl
i
eb
.
M
oder
n pd
e t
ec
hni
que
s
f
or
i
m
age i
npa
i
nt
i
ng.
D
oc
t
or
a
l
di
s
s
er
t
at
i
on.
U
ni
v
er
s
i
t
y
of
C
am
br
i
dg
e;
20
0
9.
[8
]
J
es
s
i
c
a B
os
c
h,
D
av
i
d K
ay
,
M
ar
t
i
n S
t
ol
l
,
A
n
dr
ew
J
W
a
t
he
n.
F
as
t
S
ol
v
er
s
f
or
C
ahn
--
Hi
l
l
i
ar
d
I
npa
i
nt
i
ng
S
I
A
M
J
our
nal
o
n I
m
agi
ng S
c
i
e
nc
e
s
.
20
14;
7(
1)
:
67
-
9
7.
[9
]
C
ar
ol
a
-
b
i
bi
ane S
c
hön
l
i
eb
,
A
ndr
ea B
er
t
oz
z
i
.
U
nc
ond
i
t
i
o
n
al
l
y
s
t
ab
l
e s
c
he
m
e
s
f
or
hi
gher
or
der
i
npa
i
nt
i
ng.
C
om
m
uni
c
at
i
on
s
i
n
M
at
hem
at
i
c
al
S
c
i
en
c
e
s
.
2
011;
9(
2)
:
4
13
-
45
7.
[1
0
]
J
ohn
D
O
w
ens
,
et
al
.
A
S
ur
v
ey
of
g
ener
a
l
‐
pur
pos
e
c
om
pu
t
at
i
o
n on
gr
a
phi
c
s
har
dw
ar
e.
C
om
put
er
gr
aphi
c
s
f
or
um
.
2007
;
26(
1)
:
8
0
-
113.
[1
1
]
H
ui
G
uo,
J
i
e H
e.
A
F
a
s
t
F
r
a
c
t
al
I
m
age C
om
pr
es
s
i
o
n A
l
g
o
r
i
t
hm
C
om
bi
ne
d w
i
t
h G
r
aph
i
c
P
r
oc
e
s
s
or
U
n
i
t.
T
EL
KO
M
N
I
KA
T
el
ec
om
m
uni
c
at
i
on C
om
put
i
n
g
E
l
e
c
t
r
o
ni
c
s
an
d C
ont
r
ol
.
2015
;
13(
3)
:
1089
-
1
096.
[1
2
]
M
i
ng
X
u,
X
i
anbi
n X
u,
F
ang
Z
heng,
Y
uan
hua Y
ang,
M
engj
i
a Y
i
n.
A
H
y
br
i
d S
or
t
i
ng A
l
gor
i
t
h
m
on
H
et
er
oge
neo
us
A
r
c
hi
t
ec
t
ur
es
.
T
EL
KO
M
N
I
KA
T
el
ec
om
m
uni
c
at
i
o
n C
om
put
i
ng E
l
ec
t
r
oni
c
s
a
nd
C
ont
r
ol
.
201
5;
1
3(
4)
:
1399
-
1
40
7.
Evaluation Warning : The document was created with Spire.PDF for Python.