T
E
L
K
O
M
NIKA
T
elec
o
mm
un
ica
t
io
n Co
m
pu
t
i
ng
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
,
p
p
.
206
~
218
I
SS
N:
1
6
9
3
-
6
9
3
0
,
DOI
: 1
0
.
1
2
9
2
8
/
T
E
L
KOM
NI
K
A
.
v
24
i
1
.
27401
206
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Disto
rted
born i
te
ra
tive
m
ethod
rec
o
nstruction in
hig
h
-
no
ise
env
iro
n
m
en
ts usi
ng
K
NN
-
ba
sed
ma
chine learning
d
eno
ising
Ng
uy
en
Q
ua
ng
H
uy
,
Ng
uy
en
T
ruo
ng
T
ha
ng
I
n
st
i
t
u
t
e
o
f
I
n
f
o
r
mat
i
o
n
T
e
c
h
n
o
l
o
g
y
,
V
i
e
t
n
a
m
A
c
a
d
e
my
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
H
a
n
o
i
,
V
i
e
t
n
a
m
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
15
2025
R
ev
i
s
ed
Oct
23
,
2
0
2
5
A
cc
ep
ted
Dec
8
,
2
0
2
5
Ultras
o
u
n
d
t
o
m
o
g
ra
p
h
y
re
c
o
n
stru
c
ti
o
n
u
si
n
g
th
e
d
isto
r
ted
b
o
r
n
it
e
ra
ti
v
e
m
e
th
o
d
(DBIM)
is
se
n
siti
v
e
t
o
m
e
a
su
re
m
e
n
t
n
o
ise
,
w
h
ich
d
e
g
ra
d
e
s
im
a
g
e
f
id
e
li
t
y
a
n
d
slo
w
s
c
o
n
v
e
rg
e
n
c
e
.
W
e
p
ro
p
o
se
in
teg
ra
ti
n
g
a
k
-
n
e
a
re
st
n
e
ig
h
b
o
rs
(KN
N)
d
e
n
o
isin
g
ste
p
w
it
h
in
e
a
c
h
DBIM
it
e
ra
ti
o
n
t
o
su
p
p
re
ss
n
o
ise
a
d
a
p
ti
v
e
ly
w
h
il
e
p
re
se
rv
i
n
g
stru
c
tu
ra
l
e
d
g
e
s.
S
im
u
latio
n
s
w
it
h
a
c
ircu
lar
c
y
li
n
d
rica
l
targ
e
t
a
n
d
tran
sm
it
/rec
e
iv
e
g
e
o
m
e
tr
y
(1
2
×
1
2
)
w
e
re
c
o
n
d
u
c
ted
a
t
sig
n
a
l
-
to
-
n
o
ise
ra
ti
o
(S
NR)
lev
e
ls
o
f
6
d
B,
3
d
B,
a
n
d
1
d
B
.
Co
m
p
a
re
d
w
it
h
c
o
n
v
e
n
ti
o
n
a
l
DBIM
e
m
p
lo
y
in
g
T
ik
h
o
n
o
v
re
g
u
lariz
a
ti
o
n
,
th
e
KNN
-
f
il
ter
e
d
DBIM
re
d
u
c
e
s n
o
rm
a
li
z
e
d
re
c
o
n
stru
c
ti
o
n
e
rro
r
b
y
u
p
to
5
7
.
2
%
a
t
1
d
B
a
n
d
s
h
o
w
s
fa
ste
r
e
rro
r
d
e
c
a
y
o
v
e
r
su
c
c
e
ss
i
v
e
it
e
ra
ti
o
n
s.
T
h
e
m
e
th
o
d
is
train
i
n
g
-
f
re
e
,
c
o
m
p
u
tatio
n
a
ll
y
li
g
h
tw
e
ig
h
t,
a
n
d
p
re
se
rv
e
s
f
in
e
stru
c
tu
ra
l
d
e
tails.
T
h
e
se
p
ro
p
e
rti
e
s
m
a
k
e
KN
N
-
f
il
ter
e
d
DBIM
a
tt
ra
c
ti
v
e
fo
r
n
o
isy
o
r
re
so
u
rc
e
-
c
o
n
stra
in
e
d
im
a
g
in
g
e
n
v
iro
n
m
e
n
ts.
F
u
tu
re
w
o
rk
w
il
l
v
a
li
d
a
te
th
e
a
p
p
ro
a
c
h
o
n
e
x
p
e
rim
e
n
tal
d
a
ta an
d
e
x
p
lo
re
a
d
a
p
ti
v
e
K se
lec
ti
o
n
.
K
ey
w
o
r
d
s
:
Dis
to
r
ted
b
o
r
n
iter
ativ
e
m
eth
o
d
u
ltra
s
o
u
n
d
I
n
v
er
s
e
s
ca
tter
in
g
K
-
n
ea
r
est n
e
ig
h
b
o
r
s
T
ik
h
o
n
o
v
r
eg
u
lar
izatio
n
T
o
m
o
g
r
ap
h
y
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ng
u
y
e
n
Q
u
an
g
H
u
y
I
n
s
tit
u
te
o
f
I
n
f
o
r
m
atio
n
T
ec
h
n
o
lo
g
y
,
Viet
n
a
m
A
ca
d
e
m
y
o
f
S
cien
ce
an
d
T
ec
h
n
o
lo
g
y
1
8
Ho
an
g
Qu
o
c
Viet,
C
a
u
Gia
y
Dis
tr
ict,
Ha
n
o
i,
Viet
n
a
m
E
m
ail: q
u
a
n
g
h
u
y
7
8
8
9
@
g
m
ail
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Ultr
aso
u
n
d
i
m
ag
in
g
an
d
to
m
o
g
r
ap
h
y
ar
e
cr
itical
tech
n
iq
u
es
in
clin
ical
d
iag
n
o
s
tics
,
p
r
o
v
id
in
g
n
o
n
-
in
v
a
s
i
v
e
an
d
r
ea
l
-
ti
m
e
v
i
s
u
al
i
za
tio
n
o
f
in
ter
n
a
l
tis
s
u
es.
T
r
a
d
itio
n
al
u
ltra
s
o
u
n
d
i
m
a
g
e
ac
q
u
is
itio
n
p
r
i
m
ar
il
y
r
elies
o
n
th
e
p
u
ls
e
-
ec
h
o
m
et
h
o
d
,
w
h
er
e
r
ef
lecte
d
s
ig
n
al
s
f
r
o
m
ti
s
s
u
e
b
o
u
n
d
ar
ies
ar
e
u
s
e
d
to
r
ec
o
n
s
tr
u
ct
th
e
u
n
d
er
l
y
in
g
s
tr
u
ct
u
r
e
o
f
th
e
i
m
ag
ed
o
b
j
ec
t
[
1
]
.
Ho
w
e
v
er
,
t
h
i
s
ap
p
r
o
ac
h
h
as
in
h
er
en
t
li
m
it
atio
n
s
i
n
r
eso
l
u
tio
n
an
d
co
n
tr
ast,
p
ar
ticu
lar
l
y
i
n
h
ig
h
l
y
s
ca
tter
i
n
g
m
ed
ia.
T
o
ad
d
r
ess
th
ese
ch
a
llen
g
es,
in
v
er
s
e
s
ca
tter
in
g
tech
n
iq
u
es
h
a
v
e
b
ee
n
d
ev
elo
p
ed
,
allo
w
in
g
f
o
r
m
o
r
e
ac
cu
r
at
e
i
m
a
g
e
r
ec
o
n
s
tr
u
ctio
n
b
y
in
c
o
r
p
o
r
atin
g
m
u
ltip
le
v
ie
w
i
n
g
an
g
le
s
ar
o
u
n
d
t
h
e
o
b
j
ec
t
[
2
]
.
T
h
ese
m
et
h
o
d
s
en
a
b
le
i
m
p
r
o
v
ed
i
m
ag
in
g
q
u
alit
y
,
esp
ec
iall
y
u
n
d
er
s
tr
o
n
g
s
ca
tter
i
n
g
co
n
d
it
io
n
s
,
m
ak
in
g
t
h
e
m
s
u
itab
le
f
o
r
b
io
m
ed
ical
ap
p
licatio
n
s
s
u
c
h
as b
r
ea
s
t c
an
ce
r
d
etec
tio
n
an
d
s
o
f
t
tis
s
u
e
ch
ar
ac
ter
izatio
n
.
I
n
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
y
,
t
w
o
p
r
i
m
ar
y
i
m
a
g
i
n
g
m
o
d
ali
ties
ar
e
co
m
m
o
n
l
y
s
tu
d
ied
:
atten
u
atio
n
i
m
a
g
i
n
g
an
d
s
o
u
n
d
-
s
p
ee
d
i
m
ag
in
g
[
3
]
.
W
h
ile
atten
u
atio
n
i
m
a
g
e
s
p
r
o
v
id
e
v
alu
ab
le
in
f
o
r
m
atio
n
ab
o
u
t
tis
s
u
e
p
r
o
p
er
ties
,
s
o
u
n
d
-
s
p
ee
d
i
m
ag
i
n
g
g
en
er
all
y
o
f
f
er
s
s
u
p
er
io
r
r
eso
lu
tio
n
an
d
co
n
tr
ast
,
m
ak
in
g
it
a
p
r
ef
er
r
ed
ch
o
ice
f
o
r
h
ig
h
-
f
id
elit
y
to
m
o
g
r
ap
h
ic
r
ec
o
n
s
tr
u
ctio
n
s
.
Desp
ite
it
s
p
o
ten
tial,
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
y
h
a
s
n
o
t
b
ee
n
w
id
el
y
co
m
m
er
cialize
d
d
u
e
to
th
e
co
m
p
u
tatio
n
a
l
co
m
p
lex
it
y
a
n
d
li
m
ited
ef
f
icie
n
c
y
o
f
s
tate
-
of
-
th
e
-
ar
t
in
v
er
s
e
s
ca
t
ter
in
g
tec
h
n
iq
u
es.
T
h
e
b
o
r
n
iter
ativ
e
m
et
h
o
d
(
B
I
M)
an
d
its
ad
v
an
ce
d
v
ar
ian
t,
th
e
d
is
to
r
ted
b
o
r
n
iter
ativ
e
m
eth
o
d
(
DB
I
M)
,
ar
e
am
o
n
g
th
e
m
o
s
t
w
i
d
el
y
u
s
ed
r
ec
o
n
s
tr
u
ctio
n
alg
o
r
ith
m
s
in
d
if
f
r
ac
tio
n
to
m
o
g
r
ap
h
y
[
4
]
-
[
6
]
.
DB
I
M,
in
p
ar
ticu
lar
,
is
k
n
o
w
n
f
o
r
its
f
a
s
ter
co
n
v
er
g
e
n
ce
co
m
p
ar
ed
to
B
I
M
b
u
t
s
u
f
f
er
s
f
r
o
m
h
i
g
h
er
s
e
n
s
i
tiv
it
y
to
n
o
is
e
d
u
e
to
th
e
iter
ativ
e
n
at
u
r
e
o
f
th
e
f
o
r
w
ar
d
a
n
d
in
v
er
s
e
s
o
l
v
er
s
.
A
d
d
itio
n
al
l
y
,
DB
I
M
h
as
b
ee
n
s
u
cc
ess
f
u
ll
y
ap
p
lied
in
b
o
th
2
D
an
d
3
D
r
ec
o
n
s
tr
u
ctio
n
s
,
as
w
ell
as
in
la
y
er
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
Dis
to
r
ted
b
o
r
n
iter
a
tive
meth
o
d
r
ec
o
n
s
tr
u
ctio
n
in
h
ig
h
-
n
o
is
e
en
viro
n
men
ts
u
s
in
g
…
(
N
g
u
ye
n
Qu
a
n
g
Hu
y
)
207
m
ed
ia
an
d
lo
s
s
y
en
v
ir
o
n
m
e
n
ts
[
7
]
-
[
9
]
.
T
h
ese
s
tu
d
ies
d
em
o
n
s
tr
ate
t
h
e
f
lex
ib
il
it
y
o
f
DB
I
M,
th
o
u
g
h
t
h
e
co
m
p
u
tatio
n
al
b
u
r
d
en
r
e
m
ai
n
s
a
s
ig
n
i
f
ica
n
t
ch
alle
n
g
e,
as
e
ac
h
iter
atio
n
r
eq
u
ir
es
s
o
l
v
i
n
g
lar
g
e
-
s
ca
le
m
atr
i
x
eq
u
atio
n
s
,
m
ak
i
n
g
r
ea
l
-
ti
m
e
i
m
p
le
m
en
ta
tio
n
i
m
p
r
ac
tica
l
f
o
r
m
a
n
y
cli
n
ical
ap
p
licatio
n
s
[
1
0
]
.
DB
I
M,
in
p
ar
ticu
lar
,
is
k
n
o
w
n
f
o
r
its
f
a
s
ter
co
n
v
er
g
en
ce
co
m
p
ar
ed
to
B
I
M
b
u
t
s
u
f
f
er
s
f
r
o
m
h
i
g
h
er
s
en
s
iti
v
it
y
to
n
o
is
e
d
u
e
to
th
e
iter
ativ
e
n
atu
r
e
o
f
t
h
e
f
o
r
w
ar
d
an
d
in
v
er
s
e
s
o
lv
er
s
.
A
d
d
itio
n
al
l
y
,
t
h
e
co
m
p
u
ta
ti
o
n
al
b
u
r
d
en
o
f
th
e
s
e
m
et
h
o
d
s
r
e
m
ai
n
s
a
s
i
g
n
i
f
ica
n
t
ch
a
llen
g
e,
as
ea
c
h
iter
atio
n
r
eq
u
ir
es
s
o
l
v
i
n
g
lar
g
e
-
s
ca
l
e
m
atr
ix
eq
u
atio
n
s
,
m
ak
in
g
r
ea
l
-
ti
m
e
i
m
p
le
m
e
n
tat
i
o
n
i
m
p
r
ac
tical
f
o
r
m
an
y
cli
n
i
ca
l a
p
p
licatio
n
s
.
Sev
er
al
s
t
u
d
ies
h
a
v
e
atte
m
p
t
ed
to
m
iti
g
ate
th
e
s
e
co
m
p
u
ta
tio
n
al
an
d
n
o
is
e
-
r
elate
d
ch
all
en
g
e
s
.
Fo
r
in
s
ta
n
ce
,
ed
g
e
d
etec
tio
n
m
et
h
o
d
s
w
er
e
in
co
r
p
o
r
ated
in
to
D
B
I
M
to
en
h
an
ce
co
n
v
er
g
e
n
ce
s
p
ee
d
an
d
im
p
r
o
v
e
r
ec
o
n
s
tr
u
c
tio
n
q
u
alit
y
[
1
]
.
Ho
w
e
v
er
,
th
is
ap
p
r
o
ac
h
d
o
es
n
o
t
f
u
ll
y
ad
d
r
ess
th
e
is
s
u
e
o
f
n
o
is
e
s
en
s
it
iv
i
t
y
an
d
m
a
y
in
tr
o
d
u
ce
ar
ti
f
ac
ts
in
h
i
g
h
l
y
s
ca
tter
in
g
e
n
v
ir
o
n
m
e
n
t
s
.
An
o
th
er
n
o
tab
le
ad
v
an
ce
m
e
n
t
is
t
h
e
u
s
e
o
f
th
e
m
u
lti
-
le
v
el
f
ast
m
u
l
ti
-
p
o
le
alg
o
r
ith
m
(
ML
FM
A
)
as
a
f
o
r
w
ar
d
s
o
lv
er
to
ac
ce
ler
ate
th
e
r
ec
o
n
s
tr
u
ctio
n
p
r
o
ce
s
s
[
1
1
]
.
W
h
ile
ML
FM
A
ef
f
ec
tiv
el
y
r
ed
u
ce
s
co
m
p
u
tatio
n
ti
m
e,
it
in
cu
r
s
a
h
ig
h
s
et
u
p
co
s
t
an
d
d
em
an
d
s
ex
ten
s
i
v
e
p
r
e
-
p
r
o
ce
s
s
in
g
,
m
a
k
i
n
g
p
r
ac
tical
i
m
p
le
m
en
ta
tio
n
d
if
f
ic
u
lt.
T
o
s
tab
ilize
DB
I
M
in
th
e
p
r
esen
ce
o
f
n
o
is
e
,
T
ik
h
o
n
o
v
r
e
g
u
lar
izat
io
n
h
as
tr
ad
itio
n
all
y
b
ee
n
e
m
p
lo
y
ed
to
s
o
lv
e
t
h
e
i
n
v
er
s
e
p
r
o
b
le
m
b
y
i
n
co
r
p
o
r
atin
g
li
n
ea
r
m
ea
s
u
r
e
m
e
n
t
s
o
f
p
r
ess
u
r
e
s
i
g
n
als [
1
2
]
.
W
h
i
l
e
T
ik
h
o
n
o
v
r
eg
u
l
a
r
i
z
a
ti
o
n
m
it
ig
at
e
s
s
o
m
e
i
l
l
-
p
o
s
e
d
n
ess
,
i
t
d
o
e
s
n
o
t
e
f
f
e
ct
iv
e
ly
s
u
p
p
r
e
s
s
n
o
is
e
,
o
f
te
n
l
e
a
d
in
g
t
o
d
eg
r
a
d
e
d
r
e
c
o
n
s
t
r
u
c
t
i
o
n
q
u
al
i
ty
in
n
o
is
y
en
v
i
r
o
n
m
en
ts
.
S
ev
e
r
a
l
m
a
ch
in
e
l
e
a
r
n
in
g
te
ch
n
i
q
u
es
h
av
e
b
e
en
ex
p
l
o
r
e
d
f
o
r
u
l
t
r
aso
u
n
d
t
o
m
o
g
r
a
p
h
y
.
F
o
r
i
n
s
ta
n
ce
,
C
h
en
g
e
t
a
l
.
[
1
3
]
p
r
o
p
o
s
e
d
a
d
e
e
p
l
e
a
r
n
in
g
m
eth
o
d
f
o
r
lim
i
t
e
d
-
an
g
le
p
r
o
s
t
a
te
im
ag
i
n
g
,
w
h
i
l
e
S
h
i
e
t
a
l
.
[
1
4
]
f
o
cu
s
e
d
o
n
t
im
e
o
f
f
li
g
h
t
(
T
O
F
)
ex
t
r
ac
t
i
o
n
i
n
b
o
n
e
u
l
t
r
as
o
u
n
d
t
o
m
o
g
r
a
p
h
y
.
T
h
ese
m
e
th
o
d
s
h
i
g
h
lig
h
t
th
e
p
o
te
n
tial
o
f
d
ata
-
d
r
i
v
e
n
ap
p
r
o
ac
h
es
in
i
m
p
r
o
v
i
n
g
i
m
ag
e
q
u
alit
y
.
Fu
r
t
h
er
d
ev
e
lo
p
m
e
n
t
s
in
d
ee
p
lear
n
in
g
-
b
ased
to
m
o
g
r
ap
h
ic
r
ec
o
n
s
tr
u
ctio
n
h
av
e
b
ee
n
r
ep
o
r
ted
in
[
1
5
]
,
[
1
6
]
,
w
h
er
e
s
p
ar
s
e
s
a
m
p
lin
g
an
d
g
en
er
al
to
m
o
g
r
ap
h
ic
in
v
er
s
io
n
w
er
e
e
n
h
a
n
ce
d
u
s
i
n
g
co
n
v
o
lu
tio
n
a
l
ar
ch
i
tectu
r
e
s
.
A
d
d
iti
o
n
al
s
t
u
d
ies
d
e
m
o
n
s
tr
ated
f
ast
lear
n
in
g
-
b
ased
ap
p
r
o
ac
h
es
f
o
r
u
ltra
s
o
u
n
d
s
p
ee
d
m
ap
p
in
g
[
1
7
]
,
[
1
8
]
.
Desp
ite
th
es
e
ad
v
an
ce
s
,
d
ee
p
lear
n
in
g
m
o
d
el
s
o
f
te
n
r
eq
u
ir
e
ex
ten
s
i
v
e
tr
ain
i
n
g
d
atasets
a
n
d
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
,
m
a
k
i
n
g
th
eir
r
ea
l
-
ti
m
e
d
ep
lo
y
m
e
n
t
ch
alle
n
g
i
n
g
.
Desp
ite
th
e
s
e
ad
v
an
ce
s
,
d
ee
p
lear
n
in
g
m
o
d
e
ls
o
f
ten
r
eq
u
ir
e
ex
ten
s
i
v
e
tr
ain
i
n
g
d
ataset
s
an
d
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
,
m
a
k
in
g
t
h
eir
r
ea
l
-
ti
m
e
d
ep
lo
y
m
en
t
ch
a
llen
g
i
n
g
.
T
o
ad
d
r
ess
th
is
,
w
e
p
r
o
p
o
s
e
an
alter
n
ati
v
e
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
-
b
ased
m
ac
h
i
n
e
lear
n
in
g
d
en
o
is
in
g
ap
p
r
o
ac
h
to
en
h
an
ce
th
e
r
o
b
u
s
tn
e
s
s
o
f
DB
I
M
r
ec
o
n
s
tr
u
ctio
n
s
.
Un
lik
e
d
ee
p
lear
n
in
g
m
eth
o
d
s
t
h
at
r
e
l
y
o
n
d
ata
-
d
r
iv
e
n
f
ea
t
u
r
e
ex
tr
ac
tio
n
,
KNN
d
en
o
is
in
g
le
v
er
ag
e
s
lo
ca
l
n
ei
g
h
b
o
r
h
o
o
d
in
f
o
r
m
atio
n
to
s
u
p
p
r
ess
n
o
is
e
w
h
ile
p
r
eser
v
i
n
g
s
tr
u
ct
u
r
al
d
etails,
m
ak
in
g
i
t
p
ar
ticu
lar
l
y
s
u
itab
le
f
o
r
iter
ati
v
e
r
ec
o
n
s
tr
u
ctio
n
f
r
a
m
e
w
o
r
k
s
li
k
e
DB
I
M.
T
h
e
s
i
m
p
licit
y
a
n
d
ef
f
ici
en
c
y
o
f
KNN
m
ak
e
it
a
n
attr
ac
tiv
e
ch
o
ice
f
o
r
r
ea
l
-
ti
m
e
to
m
o
g
r
ap
h
ic
i
m
a
g
i
n
g
ap
p
licatio
n
s
,
p
ar
ticu
lar
l
y
in
s
ce
n
ar
io
s
w
ith
li
m
ited
tr
ain
in
g
d
ata
o
r
co
m
p
u
tatio
n
al
co
n
s
tr
ain
t
s
.
Ultr
aso
u
n
d
to
m
o
g
r
ap
h
y
i
s
a
s
ig
n
i
f
ican
t
i
m
a
g
in
g
m
o
d
alit
y
,
p
ar
ticu
lar
l
y
i
n
b
r
ea
s
t
ca
n
c
er
d
etec
tio
n
,
s
o
f
t
t
is
s
u
e
i
m
a
g
i
n
g
,
a
n
d
n
o
n
-
d
estru
cti
v
e
tes
tin
g
(
NDT
)
,
w
h
er
e
h
i
g
h
-
r
e
s
o
lu
tio
n
an
d
lo
w
-
co
s
t
s
o
l
u
tio
n
s
ar
e
v
i
tal.
T
h
e
p
r
o
p
o
s
ed
KNN
-
f
ilter
e
d
DB
I
M
ad
d
r
ess
es
k
e
y
li
m
itatio
n
s
in
co
n
v
en
tio
n
al
m
et
h
o
d
s
b
y
o
f
f
er
in
g
a
lo
w
-
co
m
p
le
x
it
y
,
tr
ain
in
g
-
f
r
ee
d
en
o
i
s
in
g
s
tr
ate
g
y
s
u
itab
le
f
o
r
clin
ical
an
d
r
ea
l
-
ti
m
e
ap
p
licatio
n
s
.
T
h
is
w
o
r
k
p
r
esen
t
s
a
n
o
v
el
i
n
te
g
r
ati
o
n
o
f
KNN
-
b
ased
d
en
o
is
in
g
w
i
th
i
n
t
h
e
DB
I
M
f
r
a
m
e
w
o
r
k
,
w
h
ic
h
,
to
t
h
e
b
est
o
f
o
u
r
k
n
o
w
led
g
e,
h
a
s
n
o
t
b
ee
n
p
r
ev
io
u
s
l
y
r
ep
o
r
ted
in
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
y
.
W
h
ile
D
B
I
M
an
d
KNN
ar
e
in
d
iv
id
u
al
l
y
w
e
ll
-
es
tab
lis
h
ed
,
th
e
m
et
h
o
d
o
lo
g
ical
i
n
n
o
v
atio
n
li
es
i
n
t
h
e
ad
ap
tiv
e
KNN
f
ilte
r
in
g
ap
p
lied
in
ea
c
h
iter
ati
v
e
u
p
d
ate
o
f
DB
I
M,
ef
f
ec
tiv
e
l
y
en
h
a
n
ci
n
g
n
o
i
s
e
s
u
p
p
r
ess
io
n
w
it
h
o
u
t
co
m
p
r
o
m
i
s
in
g
s
tr
u
c
tu
r
al
r
eso
lu
t
io
n
.
T
h
is
n
o
v
el
in
te
g
r
a
tio
n
ad
d
r
ess
es
th
e
cr
itical
li
m
itatio
n
s
o
f
co
n
v
en
tio
n
al
r
eg
u
lar
izat
io
n
,
p
ar
ticu
lar
l
y
u
n
d
er
h
ig
h
-
n
o
is
e
s
ce
n
ar
io
s
,
an
d
th
u
s
r
ep
r
esen
t
s
a
s
u
b
s
tan
t
ial
a
d
v
an
ce
m
en
t b
e
y
o
n
d
ex
i
s
ti
n
g
ap
p
r
o
ac
h
es.
T
h
e
r
em
ai
n
d
er
o
f
th
is
p
ap
er
is
o
r
g
an
ized
as
f
o
llo
w
s
.
Sect
io
n
2
d
escr
ib
es
th
e
th
eo
r
etica
l
b
ac
k
g
r
o
u
n
d
o
f
th
e
DB
I
M
an
d
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
in
teg
r
atio
n
o
f
KNN
-
b
ased
d
en
o
i
s
i
n
g
i
n
to
th
e
r
ec
o
n
s
tr
u
ctio
n
f
r
a
m
e
w
o
r
k
.
Sectio
n
3
p
r
o
v
id
es
s
i
m
u
la
tio
n
r
es
u
lts
to
ev
al
u
at
e
th
e
ef
f
ec
ti
v
en
e
s
s
o
f
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
u
n
d
er
v
ar
io
u
s
n
o
i
s
e
co
n
d
i
tio
n
s
.
Sec
tio
n
4
d
is
cu
s
s
es
t
h
e
k
e
y
f
i
n
d
in
g
s
,
co
m
p
ar
es
th
e
m
w
i
th
ex
is
ti
n
g
ap
p
r
o
ac
h
es,
h
ig
h
li
g
h
ts
t
h
e
i
m
p
licatio
n
s
an
d
li
m
itatio
n
s
,
an
d
o
u
tli
n
es
p
o
ten
tial
d
ir
ec
tio
n
s
f
o
r
f
u
t
u
r
e
r
esear
ch
.
Fin
all
y
,
Sectio
n
5
co
n
clu
d
e
s
th
e
p
ap
er
b
y
s
u
m
m
ar
izi
n
g
t
h
e
m
ain
co
n
t
r
ib
u
tio
n
s
a
n
d
r
esu
lt
s
.
2.
M
E
T
H
O
D
2
.
1
.
D
B
I
M
T
h
e
r
eg
io
n
o
f
in
ter
est
(
R
OI
)
en
co
m
p
as
s
es
th
e
r
ec
o
n
s
tr
u
cted
o
b
j
ec
t,
w
h
ic
h
is
ce
n
ter
ed
at
th
e
o
r
ig
in
o
f
a
t
w
o
-
d
i
m
en
s
io
n
al
s
p
ac
e
an
d
d
is
cr
etize
d
in
to
a
n
N×
N
g
r
id
o
f
s
q
u
ar
e
p
i
x
els,
ea
c
h
w
i
th
a
s
id
e
len
g
th
o
f
h
.
T
h
e
s
y
s
te
m
i
n
cl
u
d
es
N
t
tr
an
s
m
i
tt
er
s
an
d
N
r
r
ec
eiv
er
s
.
Fig
u
r
e
1
p
r
o
v
id
es
a
s
ch
e
m
at
ic
r
ep
r
esen
tatio
n
o
f
t
h
e
g
eo
m
etr
ical
a
n
d
ac
o
u
s
tic
s
etu
p
o
f
th
e
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
y
s
y
s
te
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
2
0
6
-
2
1
8
208
Fig
u
r
e
1
.
Ultr
aso
u
n
d
to
m
o
g
r
a
p
h
y
i
m
ag
in
g
s
y
s
te
m
Giv
e
n
t
h
e
cir
cu
lar
s
ca
tter
in
g
r
eg
io
n
ill
u
s
tr
ated
in
F
ig
u
r
e
1
,
th
e
o
b
j
ec
t
f
u
n
ct
io
n
ca
n
b
e
d
eter
m
i
n
ed
u
s
i
n
g
:
(
)
=
{
1
(
)
2
−
2
2
=
2
(
1
1
2
−
1
2
2
)
≤
0
>
(
1
)
w
h
er
e
c
1
an
d
c
2
r
ep
r
esen
t
t
h
e
s
p
e
ed
o
f
s
o
u
n
d
w
it
h
i
n
t
h
e
o
b
j
ec
t
an
d
in
w
ater
,
r
esp
ec
tiv
e
l
y
.
T
h
e
u
ltra
s
o
u
n
d
f
r
eq
u
en
c
y
is
d
en
o
ted
b
y
f
,
w
h
i
le
ω
co
r
r
esp
o
n
d
s
to
th
e
an
g
u
l
ar
f
r
eq
u
en
c
y
,
g
i
v
e
n
b
y
ω
=
2π
f
.
A
d
d
itio
n
all
y
,
R
s
ig
n
i
f
ies t
h
e
r
ad
iu
s
o
f
th
e
o
b
j
ec
t
.
T
o
ac
q
u
ir
e
th
e
s
ca
tter
ed
d
ata
,
w
e
estab
li
s
h
a
m
ea
s
u
r
e
m
e
n
t
co
n
f
ig
u
r
atio
n
f
o
r
th
e
tr
an
s
m
itter
s
a
n
d
r
ec
eiv
er
s
.
A
t
a
n
y
g
iv
e
n
m
o
m
en
t,
o
n
l
y
a
s
i
n
g
le
tr
a
n
s
m
itter
an
d
a
s
in
g
le
r
ec
ei
v
er
ar
e
ac
tiv
e,
co
r
r
esp
o
n
d
in
g
to
o
n
e
m
ea
s
u
r
ed
d
ata
p
o
in
t.
Ass
u
m
in
g
t
h
at
d
en
s
it
y
v
ar
iatio
n
s
ar
e
n
eg
li
g
ib
le,
th
e
in
h
o
m
o
g
en
eo
u
s
w
a
v
e
eq
u
atio
n
is
ex
p
r
ess
ed
as
:
(
2
+
0
2
(
)
)
(
)
=
−
(
)
(
)
(
2
)
w
h
er
e
k
0
=
ω
/
c
0
r
ep
r
esen
ts
th
e
w
a
v
e
n
u
m
b
er
in
t
h
e
r
ef
er
e
n
ce
m
ed
i
u
m
(
i.e
.
,
w
a
ter
)
,
an
d
p
(
r
)
d
en
o
tes
th
e
to
tal
p
r
ess
u
r
e
f
ie
ld
.
B
y
s
o
l
v
i
n
g
(
2
)
,
th
e
s
ca
tter
ed
p
r
ess
u
r
e
ca
n
b
e
ex
p
r
ess
ed
i
n
an
i
n
teg
r
al
f
o
r
m
u
s
i
n
g
th
e
Gr
ee
n
’
s
f
u
n
ctio
n
as
:
(
)
=
(
)
−
(
)
=
∫
∫
(
)
(
)
(
׀
r
−
r
’
׀
)
(
3
)
w
h
er
e
(
)
r
ep
r
esen
ts
t
h
e
in
c
id
en
t
p
r
ess
u
r
e,
an
d
G
d
en
o
tes
t
h
e
f
r
ee
-
s
p
ac
e
Gr
ee
n
’
s
f
u
n
ctio
n
.
I
n
(
3
)
ca
n
b
e
s
o
lv
ed
u
s
in
g
t
h
e
m
et
h
o
d
o
f
m
o
m
e
n
t
s
,
e
m
p
lo
y
in
g
s
i
n
c
b
asis
f
u
n
ctio
n
s
a
n
d
d
elta
f
u
n
ctio
n
s
[
1
9
]
.
T
h
e
p
r
ess
u
r
e
at
th
e
g
r
id
p
o
in
ts
ca
n
b
e
r
ep
r
esen
ted
as a
n
N
2
×
1
v
ec
t
o
r
,
=
+
∗
(
∗
)
∗
(
4
)
an
d
th
e
s
ca
t
ter
ed
p
r
ess
u
r
e
ca
n
also
b
e
d
eter
m
i
n
ed
as a
s
ca
lar
v
alu
e.
=
∗
(
∗
)
∗
(
5
)
w
h
er
e
∗
is
a
1
×
N
2
v
ec
to
r
d
e
r
iv
ed
f
r
o
m
a
m
atr
ix
co
n
s
tr
u
cted
u
s
in
g
t
h
e
Gr
ee
n
’
s
co
ef
f
icie
n
t
G
0
(
r
,
r
’
)
f
o
r
ea
c
h
p
ix
el
to
th
e
i
th
r
ec
eiv
er
.
T
h
e
m
a
tr
ix
̅
is
an
N
2
×
N
2
m
atr
i
x
f
o
r
m
ed
b
y
Gr
ee
n
’
s
co
ef
f
icien
t
a
m
o
n
g
all
p
ix
els
in
t
h
e
m
e
s
h
in
g
ar
ea
.
T
h
e
o
p
er
ato
r
D
(
⋅
)
co
n
v
er
t
s
a
v
ec
to
r
in
to
a
d
iag
o
n
al
m
atr
ix
.
Deta
iled
co
m
p
u
tatio
n
s
o
f
∗
an
d
∗
ca
n
b
e
f
o
u
n
d
i
n
[
1
9
]
.
I
f
N
t
tr
an
s
m
itter
s
a
n
d
N
r
r
ec
eiv
er
s
ar
e
e
m
p
lo
y
ed
,
th
e
s
ca
tter
ed
p
r
ess
u
r
e
s
i
g
n
a
l
ca
n
b
e
r
ep
r
ese
n
t
ed
as
a
v
ec
to
r
o
f
s
ize
N
t
N
r
×
1
,
w
h
ic
h
is
d
er
i
v
ed
f
r
o
m
(
5
)
as
:
=
∗
(
∗
)
∗
=
∗
(
6
)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
Dis
to
r
ted
b
o
r
n
iter
a
tive
meth
o
d
r
ec
o
n
s
tr
u
ctio
n
in
h
ig
h
-
n
o
is
e
en
viro
n
men
ts
u
s
in
g
…
(
N
g
u
ye
n
Qu
a
n
g
Hu
y
)
209
w
h
er
e
M
=
BD
(
p
∗
)
is
t
h
e
m
atr
i
x
w
h
o
s
e
s
iz
e
is
N
t
N
r
×
N
2
Sin
ce
t
h
e
Gr
ee
n
’
s
f
u
n
ctio
n
s
at
is
f
ie
s
t
h
e
s
a
m
e
d
i
f
f
er
en
tial
eq
u
atio
n
as
t
h
e
p
r
ess
u
r
e
f
ield
,
th
e
f
o
r
w
ar
d
s
o
lv
er
is
u
tili
ze
d
to
co
m
p
u
te
th
e
Gr
ee
n
’
s
f
u
n
c
tio
n
f
o
r
an
ar
b
itra
r
y
r
ef
er
e
n
ce
b
ac
k
g
r
o
u
n
d
.
Giv
en
a
n
in
itial
v
alu
e
O
0
*
an
d
th
e
co
r
r
esp
o
n
d
in
g
r
e
f
er
en
ce
b
ac
k
g
r
o
u
n
d
at
t
i
m
e
s
tep
k
,
th
e
o
b
j
ec
t
f
u
n
ct
io
n
at
s
tep
k
+
1
is
u
p
d
ated
as:
+
1
∗
=
∗
+
∆
(
7
)
w
h
er
e
∆
O
is
t
h
e
u
p
d
ate
o
f
th
e
o
b
j
ec
t f
u
n
ctio
n
w
h
ich
ca
n
b
e
d
ed
u
ce
d
f
r
o
m
(
6
)
as:
∆
∗
=
∆
(
8
)
Ob
v
io
u
s
l
y
,
th
er
e
is
an
iter
ati
v
e
p
r
o
ce
s
s
in
DB
I
M
in
o
r
d
er
t
o
esti
m
ate
th
e
o
b
j
ec
t
f
u
n
ctio
n
O
∗
.
Mo
r
eo
v
er
,
in
ea
ch
DB
I
M
s
tep
,
w
e
n
ee
d
an
o
th
er
iter
ativ
e
p
r
o
ce
s
s
m
e
n
tio
n
ed
in
th
e
n
e
x
t
s
u
b
s
ec
t
io
n
.
2.
2
.
I
nv
er
s
e
p
ro
ble
m
I
t
is
w
ell
k
n
o
w
n
t
h
at
th
e
m
a
tr
ix
M
is
ill
-
co
n
d
it
io
n
ed
,
m
ea
n
in
g
th
at
s
m
al
l
m
ea
s
u
r
e
m
e
n
t
er
r
o
r
s
in
th
e
s
u
r
f
ac
e
d
ata
ca
n
r
es
u
lt i
n
s
ig
n
i
f
ica
n
t p
er
tu
r
b
atio
n
s
in
t
h
e
r
ec
o
n
s
tr
u
ctio
n
o
u
tco
m
e.
T
h
e
in
v
er
s
e
s
o
lv
er
m
atr
i
x
M
is
w
ea
k
l
y
d
iag
o
n
al
an
d
ill
-
co
n
d
itio
n
ed
d
u
e
t
o
th
e
p
lac
e
m
en
t
o
f
d
etec
to
r
s
o
u
ts
id
e
th
e
m
e
s
h
in
g
ar
ea
.
C
o
n
s
eq
u
en
tl
y
,
it i
s
o
f
te
n
r
ef
o
r
m
u
lated
as a
least
-
s
q
u
ar
es p
r
o
b
le
m
:
|
|
∆
∗
−
∆
∗
|
|
2
(
9
)
w
h
er
e
th
e
s
y
m
b
o
l
|
|
|
|
2
r
ep
r
esen
ts
t
h
e
E
u
clid
ea
n
n
o
r
m
o
f
v
ec
to
r
s
p
ac
e.
I
n
co
n
v
e
n
tio
n
al
m
eth
o
d
s
[
2
0
]
,
th
e
esti
m
atio
n
o
f
∆
O
∗
at
th
e
ti
m
e
s
tep
k
ca
n
b
e
p
er
f
o
r
m
ed
b
y
u
s
i
n
g
T
ik
h
o
n
o
v
r
eg
u
lar
izatio
n
[
1
2
]
:
∆
∗
=
|
|
∆
∗
−
∆
∗
|
|
2
2
+
|
|
∆
∗
|
|
2
2
(
1
0
)
w
h
er
e
∆
p
∗
sc
r
ep
r
esen
ts
th
e
d
i
f
f
er
en
ce
b
et
w
ee
n
t
h
e
p
r
ed
icted
an
d
m
ea
s
u
r
ed
s
ca
tter
ed
fi
eld
s
an
d
γ
is
t
h
e
r
eg
u
lar
izatio
n
p
ar
a
m
eter
.
T
h
e
r
eg
u
lar
ized
s
o
l
u
tio
n
i
s
ex
p
r
ess
ed
as:
∆
=
∑
/
[
(
+
)
]
=
1
(
1
1
)
w
h
er
e
u
i
an
d
v
i
ar
e
th
e
co
m
p
o
n
en
t o
f
th
e
t
w
o
m
atr
i
x
es
U
an
d
V
th
at
s
at
is
f
y
:
=
=
∑
=
1
(
1
2
)
w
h
er
e
Σ
=
(
σ
1
,
σ
2
,
.
.
σ
r
)
w
it
h
1
≥
2
≥
⋯
≥
>
0
an
d
=
=
.
W
e
ass
u
m
e
th
at
is
th
e
s
m
al
lest
n
o
n
ze
r
o
s
in
g
u
lar
v
al
u
e
th
at
w
e
w
is
h
to
r
etain
.
Fro
m
(
11
)
,
it
is
ev
id
en
t
th
at
t
h
e
r
eg
u
lar
izatio
n
p
ar
a
m
eter
γ
m
u
s
t
b
e
ca
r
ef
u
ll
y
s
elec
ted
,
as
it
p
lay
s
a
cr
u
cial
r
o
le
in
m
ai
n
tai
n
in
g
t
h
e
s
tab
ilit
y
o
f
t
h
e
s
y
s
te
m
[
2
0
]
.
A
lar
g
e
γ
v
alu
e
r
esu
l
ts
i
n
a
r
o
u
g
h
r
ec
o
n
s
tr
u
cted
i
m
a
g
e,
w
h
er
ea
s
a
s
m
a
ll
γ
in
cr
e
ases
co
m
p
u
tatio
n
al
co
m
p
le
x
it
y
.
T
h
ese
d
ata
ar
e
p
r
o
ce
s
s
ed
u
s
in
g
t
h
e
DB
I
M
to
r
ec
o
n
s
tr
u
ct
th
e
s
p
ee
d
o
f
s
o
u
n
d
co
n
tr
ast.
T
h
is
ap
p
r
o
ac
h
en
ab
les
th
e
d
etec
tio
n
o
f
tis
s
u
e
p
r
esen
ce
w
ith
in
th
e
m
ed
iu
m
.
DB
I
M
r
elies
o
n
th
e
B
o
r
n
ap
p
r
o
x
i
m
atio
n
to
ite
r
ativ
el
y
s
o
lv
e
th
e
n
o
n
l
in
ea
r
in
v
er
s
e
s
ca
tter
i
n
g
p
r
o
b
lem
.
A
lg
o
r
it
h
m
1
p
r
esen
ts
th
e
ca
lcu
la
tio
n
p
r
o
ce
s
s
o
f
t
h
e
DB
I
M
as f
o
llo
w
s
:
A
l
g
o
r
ith
m
1
.
T
h
e
d
is
to
r
ted
b
o
r
n
iter
ativ
e
m
et
h
o
d
Select
th
e
p
o
s
itio
n
s
o
f
t
h
e
tr
an
s
m
i
tter
s
an
d
d
etec
to
r
s
as ill
u
s
t
r
ated
in
Fig
u
r
e
1
.
C
h
o
o
s
e
i
n
itial
v
al
u
es: O
0
*
=0
a
n
d
p
0
*
=p
*inc
as sh
o
w
n
in
(
1
5
)
Whil
e
n
<
N
m
a
x
o
r
R
R
E
<ε
1.
C
alcu
late
B
*
an
d
C
*
2.
C
alcu
late
p
*
an
d
p
*sc
co
r
r
esp
o
n
d
s
to
O
n
*
u
s
i
n
g
(
4
)
an
d
(
5
)
3.
C
alcu
late
R
R
E
co
r
r
esp
o
n
d
s
to
O
n
*
4.
C
alcu
late
th
e
∆
O
n
*
b
y
s
o
lv
in
g
(
1
0
)
5.
C
alcu
late
a
n
e
w
v
al
u
e
o
f
O
n+
1
*
b
y
u
s
i
n
g
(
7
)
6.
n
=n
+1
E
nd
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
2
0
6
-
2
1
8
210
I
n
Alg
o
r
it
h
m
1
,
th
e
r
elat
iv
e
r
e
s
id
u
al
er
r
o
r
is
d
ef
in
ed
b
y
:
=
|
|
M
∆
∗
−
∆
∗
|
|
/
|
|
∆
∗
|
|
(
1
3
)
T
h
e
r
elativ
e
r
esid
u
al
er
r
o
r
(
R
R
E
)
is
co
m
p
u
ted
at
ea
ch
iter
at
io
n
.
T
h
e
iter
ativ
e
p
r
o
ce
s
s
w
il
l
ter
m
i
n
ate
w
h
e
n
t
h
e
R
R
E
f
alls
b
elo
w
a
p
r
ed
ef
in
ed
to
ler
an
ce
o
r
w
h
e
n
t
h
e
n
u
m
b
er
o
f
iter
atio
n
s
r
ea
c
h
es t
h
e
m
a
x
i
m
u
m
li
m
i
t N
m
a
x
.
2
.
3
.
K
NN
f
o
r
deno
is
ing
in DB
I
M
T
h
e
KNN
d
en
o
is
in
g
ap
p
r
o
ac
h
is
a
m
ac
h
i
n
e
lear
n
i
n
g
-
b
ase
d
tech
n
iq
u
e
t
h
at
e
n
h
a
n
ce
s
t
h
e
DB
I
M
b
y
r
ed
u
cin
g
n
o
i
s
e
i
n
t
h
e
r
ec
o
n
s
tr
u
cted
i
m
ag
e
s
.
T
r
ad
itio
n
al
DB
I
M
s
u
f
f
er
s
f
r
o
m
h
i
g
h
n
o
is
e
s
e
n
s
it
iv
i
t
y
,
w
h
ic
h
ca
n
d
eg
r
ad
e
i
m
ag
e
q
u
a
lit
y
a
n
d
s
l
o
w
d
o
w
n
co
n
v
er
g
e
n
ce
.
B
y
in
t
eg
r
atin
g
K
NN
f
ilter
i
n
g
in
to
th
e
iter
ati
v
e
p
r
o
ce
s
s
,
n
o
is
e
ca
n
b
e
ad
ap
tiv
el
y
s
u
p
p
r
ess
ed
w
h
ile
p
r
eser
v
in
g
i
m
p
o
r
tan
t
s
tr
u
ct
u
r
al
d
etails.
KNN
d
en
o
is
i
n
g
le
v
er
ag
e
s
th
e
s
p
atial
co
r
r
elatio
n
a
m
o
n
g
n
eig
h
b
o
r
in
g
d
ata
p
o
in
ts
to
s
m
o
o
th
o
u
t
n
o
is
e
w
h
ile
m
ai
n
tai
n
in
g
t
h
e
i
n
te
g
r
it
y
o
f
th
e
s
i
g
n
al.
T
h
is
m
a
k
es
it
p
ar
ticu
lar
l
y
e
f
f
ec
ti
v
e
f
o
r
ill
-
co
n
d
itio
n
ed
in
v
er
s
e
p
r
o
b
lem
s
s
u
ch
as
DB
I
M
-
b
ased
to
m
o
g
r
ap
h
ic
r
ec
o
n
s
tr
u
ct
io
n
.
DB
I
M
is
w
id
el
y
u
s
ed
i
n
d
if
f
r
ac
tio
n
to
m
o
g
r
ap
h
y
d
u
e
to
its
ab
ilit
y
to
iter
ati
v
el
y
r
ef
in
e
r
ec
o
n
s
tr
u
ctio
n
s
o
f
s
m
all
-
s
ca
le
s
tr
u
c
t
u
r
es.
Ho
w
e
v
er
,
its
h
ig
h
s
e
n
s
i
tiv
it
y
to
n
o
is
e
r
em
ai
n
s
a
m
aj
o
r
li
m
ita
tio
n
,
p
ar
ticu
lar
l
y
i
n
s
ce
n
ar
io
s
w
h
er
e
th
e
i
n
p
u
t
d
ata
is
co
n
ta
m
i
n
ated
b
y
m
ea
s
u
r
e
m
e
n
t
er
r
o
r
s
o
r
s
y
s
te
m
in
s
tab
il
it
y
.
T
r
ad
itio
n
al
r
eg
u
la
r
izatio
n
is
co
m
m
o
n
l
y
u
s
ed
t
o
ad
d
r
ess
th
is
i
s
s
u
e,
b
u
t
it
d
o
es
n
o
t
ef
f
ec
ti
v
el
y
r
e
m
o
v
e
n
o
i
s
e,
esp
ec
iall
y
w
h
e
n
d
ea
lin
g
w
i
th
s
tr
o
n
g
s
ca
tter
i
n
g
en
v
ir
o
n
m
e
n
t
s
.
KNN
f
ilter
in
g
is
in
tr
o
d
u
ce
d
as a
n
alter
n
ati
v
e
ap
p
r
o
ac
h
to
im
p
r
o
v
e
n
o
is
e
r
o
b
u
s
tn
e
s
s
w
it
h
o
u
t
s
i
g
n
i
f
ica
n
tl
y
i
n
cr
ea
s
i
n
g
co
m
p
u
t
atio
n
al
co
m
p
le
x
it
y
.
T
h
e
k
e
y
ad
v
a
n
ta
g
es
o
f
u
s
i
n
g
KNN
in
DB
I
M
in
cl
u
d
e:
(
a)
p
r
eser
v
atio
n
o
f
s
tr
u
ct
u
r
al
d
etail
s
:
Un
li
k
e
tr
ad
itio
n
al
s
m
o
o
th
in
g
f
i
lter
s
,
KNN
d
o
es
n
o
t
b
lu
r
ed
g
es
o
r
d
is
to
r
t
f
i
n
e
f
ea
t
u
r
es
in
t
h
e
r
ec
o
n
s
tr
u
cted
i
m
a
g
es;
(
b
)
a
d
ap
tab
ilit
y
to
n
o
n
lin
ea
r
d
ata:
KNN
o
p
er
ates
b
ased
o
n
s
i
m
i
lar
it
y
m
etr
ics
r
at
h
er
t
h
an
f
i
x
e
d
tr
an
s
f
o
r
m
atio
n
s
,
m
ak
in
g
it
s
u
itab
le
f
o
r
co
m
p
le
x
b
io
m
ed
ical
i
m
a
g
i
n
g
s
ce
n
ar
i
o
s
;
an
d
(
c)
c
o
m
p
u
tatio
n
al
ef
f
i
cien
c
y
:
KNN
d
o
es
n
o
t
r
eq
u
ir
e
p
r
e
-
tr
ain
i
n
g
,
m
ak
in
g
i
t
a
lig
h
t
w
ei
g
h
t
an
d
ea
s
y
-
to
-
i
n
te
g
r
ate
s
o
lu
tio
n
w
it
h
i
n
t
h
e
iter
ativ
e
DB
I
M
f
r
a
m
e
w
o
r
k
.
KNN
d
en
o
is
i
n
g
is
ap
p
lied
w
it
h
i
n
th
e
DB
I
M
f
r
a
m
e
w
o
r
k
a
s
:
i)
n
o
i
s
e
id
en
ti
f
icatio
n
:
a
t
ea
ch
iter
atio
n
o
f
DB
I
M,
th
e
r
ec
o
n
s
tr
u
cted
f
ield
is
af
f
ec
ted
b
y
n
o
is
e,
w
h
ich
ca
n
ca
u
s
e
n
u
m
e
r
ical
in
s
tab
ili
t
y
a
n
d
i
m
a
g
e
d
eg
r
ad
at
io
n
;
ii)
l
o
ca
l
n
eig
h
b
o
r
h
o
o
d
s
elec
tio
n
:
f
o
r
ea
ch
p
ix
el
(
o
r
g
r
id
p
o
in
t)
in
th
e
r
ec
o
n
s
tr
u
cted
i
m
ag
e,
a
s
et
o
f
K
n
ea
r
es
t
n
ei
g
h
b
o
r
s
i
s
id
en
ti
f
ied
b
ased
o
n
E
u
clid
ea
n
d
is
ta
n
ce
;
ii
i)
w
eig
h
ted
a
v
er
ag
in
g
:
t
h
e
i
n
te
n
s
it
y
v
alu
e
o
f
th
e
tar
g
et
p
ix
el
is
r
ep
lace
d
w
it
h
th
e
w
ei
g
h
ted
m
ea
n
o
f
its
K
-
n
ea
r
es
t
n
ei
g
h
b
o
r
s
.
T
h
is
s
tep
s
m
o
o
t
h
s
o
u
t
n
o
is
e
w
h
ile
p
r
eser
v
i
n
g
h
i
g
h
-
c
o
n
tr
ast
f
ea
tu
r
es
;
an
d
iv
)
i
ter
at
iv
e
r
ef
i
n
e
m
en
t:
t
h
e
d
e
n
o
is
ed
i
m
a
g
e
is
f
ed
b
ac
k
in
to
t
h
e
n
e
x
t D
B
I
M
iter
atio
n
,
en
h
a
n
ci
n
g
s
tab
ilit
y
a
n
d
ac
ce
le
r
atin
g
co
n
v
er
g
e
n
ce
.
Ma
th
e
m
at
icall
y
,
t
h
e
d
en
o
i
s
ed
v
alu
e
f
o
r
a
p
ix
el
is
g
iv
e
n
b
y
:
′
=
1
∑
∈
(
)
(
1
4
)
w
h
er
e
(
)
r
ep
r
esen
ts
t
h
e
s
et
o
f
K
-
n
ea
r
est n
e
ig
h
b
o
r
s
o
f
.
T
h
e
KNN
f
ilter
i
n
th
i
s
s
t
u
d
y
w
as
i
m
p
le
m
e
n
ted
w
ith
=5
,
w
h
ic
h
p
r
o
v
id
ed
an
o
p
ti
m
al
t
r
ad
e
-
o
f
f
b
et
w
ee
n
n
o
is
e
s
u
p
p
r
ess
io
n
a
n
d
s
tr
u
ct
u
r
al
p
r
eser
v
atio
n
.
S
ev
er
al
v
al
u
es
o
f
(
3
,
5
,
an
d
7
)
w
er
e
test
ed
ex
p
er
i
m
e
n
tall
y
;
w
h
ile
s
m
alle
r
v
alu
e
s
led
to
i
n
s
u
f
f
icie
n
t
d
en
o
is
in
g
,
lar
g
er
v
al
u
es
ten
d
ed
to
o
v
er
s
m
o
o
th
o
b
j
ec
t
b
o
u
n
d
ar
ies.
T
h
e
E
u
clid
ea
n
d
is
tan
ce
m
etr
ic
w
as
u
s
e
d
to
d
eter
m
i
n
e
th
e
n
ea
r
est
n
eig
h
b
o
r
s
,
an
d
ea
ch
n
eig
h
b
o
r
’
s
co
n
tr
ib
u
tio
n
w
as
w
ei
g
h
ted
in
v
er
s
el
y
p
r
o
p
o
r
tio
n
al
to
its
d
is
tan
ce
f
r
o
m
t
h
e
tar
g
et
p
ix
el.
T
h
is
co
n
f
i
g
u
r
atio
n
y
ield
ed
th
e
m
o
s
t
s
tab
le
co
n
v
er
g
e
n
ce
b
eh
a
v
io
r
an
d
lo
w
est
r
ec
o
n
s
tr
u
ct
io
n
er
r
o
r
ac
r
o
s
s
d
if
f
er
e
n
t
n
o
is
e
le
v
els,
a
s
d
e
m
o
n
s
tr
ated
in
t
h
e
s
i
m
u
latio
n
r
es
u
lts
.
K
NN
d
en
o
is
i
n
g
is
i
n
te
g
r
ated
d
ir
ec
tl
y
a
f
ter
n
o
i
s
e
-
co
n
ta
m
i
n
ated
f
ield
es
ti
m
atio
n
a
n
d
b
ef
o
r
e
th
e
in
v
er
s
e
p
r
o
b
lem
is
s
o
lv
ed
i
n
ea
c
h
D
B
I
M
iter
atio
n
.
T
h
is
in
te
g
r
atio
n
en
s
u
r
es
t
h
at
DB
I
M
r
ec
eiv
es
a
clea
n
er
in
p
u
t
at
ea
ch
iter
atio
n
,
lead
in
g
to
f
ast
er
co
n
-
v
er
g
e
n
ce
an
d
i
m
p
r
o
v
ed
r
ec
o
n
s
tr
u
ctio
n
ac
cu
r
ac
y
.
A
lg
o
r
it
h
m
2
s
h
o
w
s
t
h
e
K
-
n
ea
r
est
n
ei
g
h
b
o
r
s
-
f
ilter
ed
d
is
to
r
ted
b
o
r
n
iter
ativ
e
m
eth
o
d
.
A
l
g
o
r
ith
m
2
.
T
h
e
KNN
-
f
ilte
r
e
d
DB
I
M
1.
I
n
itialize
DB
I
M
w
it
h
m
ea
s
u
r
e
d
s
ca
tter
in
g
d
ata.
2.
A
d
d
Gau
s
s
ia
n
n
o
i
s
e
to
s
i
m
u
la
t
e
r
ea
l
-
w
o
r
ld
m
ea
s
u
r
e
m
e
n
t
u
n
c
er
tain
ties
.
3.
A
p
p
l
y
K
NN
-
b
a
s
ed
f
ilter
i
n
g
o
n
th
e
s
ca
tter
ed
f
ield
to
r
e
m
o
v
e
h
ig
h
-
f
r
eq
u
en
c
y
n
o
is
e
co
m
p
o
n
en
ts
.
4.
Use th
e
d
en
o
i
s
ed
d
ata
to
u
p
d
ate
th
e
co
n
tr
ast
f
u
n
ct
io
n
u
s
i
n
g
t
h
e
DB
I
M
iter
ativ
e
p
r
o
ce
s
s
.
5.
R
ep
ea
t step
s
3
-
4
u
n
til co
n
v
er
g
en
ce
cr
iter
ia
ar
e
m
et.
Un
li
k
e
m
ed
ian
,
b
ilater
al,
o
r
w
a
v
elet
-
b
ased
d
en
o
is
er
s
t
h
at
eith
er
b
lu
r
ed
g
es
o
r
r
eq
u
ir
e
t
r
an
s
f
o
r
m
-
d
o
m
ai
n
tu
n
i
n
g
,
an
d
u
n
li
k
e
co
m
p
u
tatio
n
all
y
i
n
te
n
s
i
v
e
m
et
h
o
d
s
s
u
c
h
as
B
M3
D
o
r
NL
M,
o
u
r
KNN
-
b
ased
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
Dis
to
r
ted
b
o
r
n
iter
a
tive
meth
o
d
r
ec
o
n
s
tr
u
ctio
n
in
h
ig
h
-
n
o
is
e
en
viro
n
men
ts
u
s
in
g
…
(
N
g
u
ye
n
Qu
a
n
g
Hu
y
)
211
ap
p
r
o
ac
h
p
r
o
v
id
es
a
f
av
o
r
a
b
le
tr
ad
e
-
o
f
f
:
tr
ai
n
in
g
-
f
r
ee
,
co
m
p
u
tatio
n
all
y
li
g
h
t
w
ei
g
h
t,
lo
ca
ll
y
ad
ap
tiv
e
s
m
o
o
th
in
g
t
h
at
p
r
eser
v
es
s
tr
u
c
tu
r
al
ed
g
es a
n
d
is
s
tr
ai
g
h
tf
o
r
war
d
to
em
b
ed
in
to
ea
ch
DB
I
M
iter
atio
n
.
T
h
e
o
v
er
all
w
o
r
k
f
lo
w
o
f
th
e
p
r
o
p
o
s
ed
KNN
-
f
ilter
ed
DB
I
M
is
ill
u
s
tr
ated
i
n
Fi
g
u
r
e
2
.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
w
it
h
i
n
itia
lizi
n
g
th
e
D
B
I
M
u
s
i
n
g
m
ea
s
u
r
ed
s
ca
t
ter
ed
f
ield
d
ata,
f
o
llo
w
ed
b
y
th
e
ad
d
itio
n
o
f
Ga
u
s
s
ian
n
o
is
e
to
e
m
u
late
r
ea
li
s
tic
m
e
asu
r
e
m
en
t
u
n
ce
r
tain
ties
.
A
K
NN
-
b
ased
d
en
o
is
in
g
s
tep
i
s
t
h
en
ap
p
lied
to
th
e
s
ca
tter
ed
f
ield
to
s
u
p
p
r
ess
h
ig
h
-
f
r
eq
u
e
n
c
y
n
o
i
s
e
co
m
p
o
n
e
n
ts
w
h
ile
p
r
eser
v
in
g
s
tr
u
ct
u
r
al
in
f
o
r
m
at
io
n
.
T
h
e
d
en
o
is
ed
d
ata
ar
e
s
u
b
s
eq
u
en
t
l
y
u
s
ed
in
th
e
DB
I
M
iter
ativ
e
u
p
d
ate
to
r
ef
in
e
th
e
co
n
tr
a
s
t
f
u
n
c
tio
n
.
T
h
ese
f
ilter
i
n
g
an
d
u
p
d
atin
g
s
tep
s
ar
e
r
ep
ea
ted
u
n
til
th
e
co
n
v
er
g
en
ce
cr
iter
ia
ar
e
s
ati
s
f
ied
,
y
ield
i
n
g
th
e
f
in
a
l
r
ec
o
n
s
tr
u
cted
s
o
u
n
d
-
s
p
ee
d
d
is
tr
ib
u
tio
n
.
Fig
u
r
e
2
.
Flo
w
c
h
ar
t o
f
t
h
e
p
r
o
p
o
s
ed
KNN
-
f
i
lter
ed
DB
I
M
T
h
e
co
m
p
u
ta
tio
n
al
co
s
t
o
f
th
e
p
r
o
p
o
s
ed
KNN
-
f
i
lter
ed
DB
I
M
w
as
s
li
g
h
t
l
y
h
i
g
h
er
t
h
an
t
h
at
o
f
t
h
e
s
tan
d
ar
d
DB
I
M
d
u
e
to
t
h
e
ad
d
itio
n
al
d
en
o
i
s
i
n
g
s
tep
.
O
n
av
e
r
ag
e,
th
e
KNN
-
b
ased
f
ilter
in
g
in
cr
ea
s
ed
t
h
e
to
ta
l
co
m
p
u
tatio
n
ti
m
e
b
y
ap
p
r
o
x
i
m
atel
y
8
–
1
0
%
p
er
iter
atio
n
,
w
h
ic
h
is
n
e
g
li
g
ib
le
co
m
p
ar
ed
w
ith
t
h
e
o
v
er
all
DB
I
M
r
ec
o
n
s
tr
u
ctio
n
ti
m
e.
Un
li
k
e
d
ee
p
-
lear
n
i
n
g
-
b
ased
d
en
o
is
er
s
,
th
e
KN
N
o
p
er
atio
n
r
eq
u
ir
es
n
o
m
o
d
el
tr
ain
i
n
g
a
n
d
o
n
l
y
i
n
v
o
lv
es
s
i
m
p
le
d
i
s
tan
ce
co
m
p
u
tat
io
n
s
,
k
ee
p
in
g
t
h
e
m
eth
o
d
lig
h
t
w
ei
g
h
t
an
d
co
m
p
u
tatio
n
all
y
e
f
f
icie
n
t.
T
h
er
ef
o
r
e,
th
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
ac
h
iev
es
a
f
a
v
o
r
ab
le
b
alan
ce
b
et
w
ee
n
i
m
p
r
o
v
ed
r
ec
o
n
s
tr
u
ctio
n
q
u
alit
y
a
n
d
m
i
n
i
m
al
ad
d
ed
co
m
p
u
tatio
n
a
l o
v
e
r
h
ea
d
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
DB
I
M
im
p
le
m
en
ta
tio
n
an
d
o
u
r
p
r
o
p
o
s
ed
c
o
n
f
ig
u
r
atio
n
ar
e
v
alid
ated
u
s
in
g
s
i
m
u
lat
io
n
o
n
tar
g
et
w
it
h
m
o
d
er
ate
s
p
ee
d
co
n
tr
a
s
t
.
Si
m
u
lated
d
ata
w
er
e
g
e
n
er
ated
f
o
r
an
i
n
f
i
n
itel
y
lo
n
g
cir
cu
lar
c
y
li
n
d
er
,
d
is
cr
etize
d
in
to
an
N
×
N
=
12
×
12
p
ix
el
g
r
id
.
T
h
e
cy
li
n
d
er
h
as
a
r
ad
i
u
s
o
f
7
.
3
m
m
,
an
u
ltra
s
o
u
n
d
s
i
g
n
a
l
f
r
eq
u
en
c
y
o
f
0
.
5
MH
z,
an
d
a
s
o
u
n
d
s
p
ee
d
co
n
tr
ast
o
f
1
0
%.
T
h
e
s
y
s
te
m
in
cl
u
d
es
N
t
=1
2
tr
an
s
m
itter
s
a
n
d
N
r
=1
2
r
ec
eiv
er
s
.
T
h
e
tr
an
s
m
itt
er
s
ar
e
p
o
s
itio
n
ed
at
1
1
ev
en
l
y
s
p
ac
ed
lo
ca
tio
n
s
alo
n
g
a
cir
cu
lar
p
ath
s
u
r
r
o
u
n
d
in
g
th
e
o
b
j
ec
t.
Fo
r
e
ac
h
tr
an
s
m
i
tter
,
1
2
d
etec
to
r
s
ar
e
p
lace
d
o
n
th
e
o
p
p
o
s
ite
s
id
e
o
f
th
e
cir
cle
to
ca
p
tu
r
e
th
e
s
ca
t
ter
ed
s
ig
n
als.
T
h
e
in
cid
en
t
p
r
ess
u
r
e
f
o
r
a
ze
r
o
-
o
r
d
er
B
ess
el
b
ea
m
i
n
a
t
w
o
-
d
i
m
en
s
io
n
al
ca
s
e
i
s
g
iv
e
n
b
y
:
∗
=
0
(
0
׀
−
׀
)
(
1
5
)
w
h
er
e
0
is
th
e
0
th
o
r
d
e
r
B
ess
el
f
u
n
ctio
n
an
d
|
−
|
is
th
e
d
is
tan
ce
b
etw
ee
n
th
e
tr
an
s
m
itter
an
d
th
e
k
th
p
o
in
t
in
th
e
R
OI
.
Fig
u
r
e
3
is
th
e
id
ea
l
o
b
j
ec
t
f
u
n
ctio
n
co
r
r
esp
o
n
d
in
g
to
(
1
)
.
I
n
o
r
d
er
to
h
av
e
a
q
u
a
n
titati
v
e
co
m
p
ar
is
o
n
,
(
1
3
)
ca
n
b
e
u
s
ed
to
ca
lcu
late
th
e
R
R
E
f
o
r
ea
ch
ca
s
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
2
0
6
-
2
1
8
212
Fig
u
r
e
3.
I
d
ea
l o
b
j
ec
t f
u
n
ctio
n
(
o
n
e
o
b
j
ec
t in
th
e
r
eg
io
n
o
f
in
ter
est).
C
o
lo
r
b
ar
in
d
icate
s
p
er
ce
n
t o
f
s
o
u
n
d
-
s
p
ee
d
co
n
tr
ast (
%)
Fig
u
r
e
4
s
h
o
w
s
th
e
n
o
r
m
a
liz
ed
r
ec
o
n
s
tr
u
ctio
n
er
r
o
r
af
ter
th
e
f
ir
s
t
iter
atio
n
f
o
r
DB
I
M
an
d
KNN
-
f
ilter
ed
DB
I
M
at
s
i
g
n
al
-
to
-
n
o
i
s
e
r
atio
(
SNR
)
le
v
el
s
o
f
6
d
B
,
3
d
B
,
an
d
1
d
B
.
T
h
e
co
r
r
esp
o
n
d
in
g
r
ec
o
n
s
tr
u
cted
s
o
u
n
d
-
s
p
ee
d
m
ap
s
ar
e
d
is
p
la
y
ed
in
Fi
g
u
r
e
5
;
v
is
u
al
in
s
p
ec
tio
n
in
d
icate
s
i
m
p
r
o
v
ed
s
tr
u
ctu
r
al
p
r
eser
v
atio
n
w
it
h
KN
N
f
il
ter
in
g
.
Fig
u
r
e
4
.
No
r
m
alize
d
er
r
o
r
o
f
th
e
DB
I
M
an
d
KNN
-
f
i
lter
ed
DB
I
M
w
it
h
v
ar
io
u
s
s
i
g
n
al
-
to
-
n
o
is
e
r
atio
T
h
e
r
esu
lts
in
F
ig
u
r
e
4
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
KNN
-
b
ased
d
en
o
is
i
n
g
m
et
h
o
d
co
n
s
i
s
te
n
tl
y
r
ed
u
ce
s
er
r
o
r
ac
r
o
s
s
all
n
o
is
e
lev
els
,
d
e
m
o
n
s
tr
atin
g
s
u
p
er
io
r
r
o
b
u
s
tn
es
s
i
n
h
i
g
h
-
n
o
is
e
e
n
v
i
r
o
n
m
e
n
ts
.
A
t
SNR
=
6
d
B
,
as
s
h
o
w
n
i
n
Fig
u
r
e
6
,
th
e
KNN
-
f
ilter
ed
DB
I
M
a
ch
iev
e
s
a
4
9
.
5
%
r
ed
u
ctio
n
i
n
er
r
o
r
co
m
p
ar
ed
to
co
n
v
e
n
tio
n
al
DB
I
M
(
2
.
0
6
5
3
v
s
.
4
.
0
9
3
6
)
.
T
h
is
tr
en
d
co
n
ti
n
u
e
s
at
SN
R
=
3
d
B
as sh
o
w
n
i
n
Fig
u
r
e
7
,
w
h
er
e
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
r
ed
u
ce
s
th
e
er
r
o
r
b
y
49%
(
2
.
6
4
1
7
v
s
.
5
.
1
7
7
4
)
.
T
h
e
m
o
s
t
s
ig
n
i
f
ica
n
t
i
m
p
r
o
v
e
m
en
t
is
o
b
s
er
v
ed
at
SNR
=
1
d
B
,
w
h
er
e
th
e
er
r
o
r
is
r
ed
u
ce
d
b
y
5
7
.
2
%
(
3
.
3
9
9
8
v
s
.
7
.
9
4
3
1
)
,
h
ig
h
li
g
h
ti
n
g
t
h
e
ef
f
ec
tiv
e
n
e
s
s
o
f
KN
N
d
en
o
is
i
n
g
in
ex
tr
e
m
el
y
n
o
is
y
co
n
d
it
i
o
n
s
.
T
h
ese
f
i
n
d
in
g
s
s
u
g
g
es
t
t
h
at
in
te
g
r
ati
n
g
KNN
-
b
ased
d
en
o
is
in
g
i
n
to
DB
I
M
e
n
h
a
n
ce
s
co
n
v
er
g
e
n
ce
s
p
ee
d
b
y
p
r
o
v
id
in
g
a
clea
n
er
i
n
itial
es
ti
m
ate
(
a
s
s
h
o
w
n
i
n
Fig
u
r
e
5
)
,
r
e
d
u
cin
g
th
e
b
u
r
d
en
o
n
th
e
iter
ati
v
e
r
ec
o
n
s
tr
u
c
tio
n
p
r
o
ce
s
s
.
T
h
e
a
b
ilit
y
o
f
KNN
to
ad
ap
tiv
el
y
-1
0
1
-1
0
1
0
5
10
15
20
25
30
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
0
5
10
15
20
25
30
0
1
2
3
4
5
6
7
8
9
6
3
1
N
o
r
m
al
i
ze
d
e
r
r
o
r
S
i
g
n
al
-
to
-
n
o
i
se
r
at
i
o
(d
B
)
DB
IM
KNN
-
f
ilte
re
d
DBIM
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
Dis
to
r
ted
b
o
r
n
iter
a
tive
meth
o
d
r
ec
o
n
s
tr
u
ctio
n
in
h
ig
h
-
n
o
is
e
en
viro
n
men
ts
u
s
in
g
…
(
N
g
u
ye
n
Qu
a
n
g
Hu
y
)
213
s
m
o
o
th
n
o
is
e
w
h
ile
p
r
eser
v
i
n
g
s
tr
u
ct
u
r
al
d
etails
co
n
tr
ib
u
tes
to
m
o
r
e
s
tab
le
an
d
ac
c
u
r
ate
r
ec
o
n
s
tr
u
ctio
n
s
,
p
ar
ticu
lar
l
y
i
n
lo
w
-
SN
R
co
n
d
itio
n
s
.
C
o
n
s
eq
u
e
n
tl
y
,
th
e
KNN
-
f
il
ter
ed
DB
I
M
n
o
t
o
n
l
y
ac
ce
ler
ates
t
h
e
co
n
v
er
g
e
n
ce
r
ate
b
u
t
also
im
p
r
o
v
es
o
v
er
all
im
a
g
e
q
u
alit
y
,
m
ak
in
g
it
a
p
r
o
m
i
s
i
n
g
ap
p
r
o
a
ch
f
o
r
to
m
o
g
r
ap
h
ic
i
m
a
g
in
g
i
n
ch
al
len
g
i
n
g
e
n
v
ir
o
n
m
e
n
t
s
.
DB
I
M
K
NN
-
fil
te
r
e
d
DB
IM
S
NR
=
6
d
B
S
NR
=
3
d
B
S
NR
=
1
d
B
Fig
u
r
e
5
.
R
ec
o
n
s
tr
u
cted
o
b
j
ec
t
f
u
n
ctio
n
af
ter
t
h
e
f
ir
s
t iter
atio
n
,
u
s
in
g
t
h
e
co
n
v
e
n
tio
n
al
DB
I
M
an
d
p
r
o
p
o
s
ed
KNN
-
f
il
ter
ed
DB
I
M
m
et
h
o
d
s
in
ca
s
e
o
f
SNR
=
6
,
3
,
an
d
1
d
B
,
r
esp
ec
tiv
el
y
.
C
o
lo
r
b
ar
in
d
icate
s
p
er
ce
n
t o
f
s
o
u
n
d
-
s
p
ee
d
co
n
tr
ast (
%)
Fig
u
r
e
6
.
T
h
e
n
o
r
m
aliza
tio
n
er
r
o
r
af
ter
th
e
f
ir
s
t
th
r
ee
iter
atio
n
s
u
s
in
g
t
h
e
co
n
v
en
tio
n
al
DB
I
M
an
d
p
r
o
p
o
s
ed
KNN
-
f
ilter
ed
DB
I
M
m
eth
o
d
in
ca
s
e
o
f
SNR
=
6
d
B
Fig
u
r
e
7
.
T
h
e
n
o
r
m
aliza
tio
n
er
r
o
r
af
ter
th
e
f
ir
s
t t
h
r
ee
iter
atio
n
s
u
s
i
n
g
th
e
co
n
v
e
n
tio
n
al
DB
I
M
an
d
pr
o
p
o
s
ed
KNN
-
f
ilter
ed
DB
I
M
m
eth
o
d
in
ca
s
e
o
f
SNR
=
3
d
B
-1
0
1
-1
0
1
10
20
30
40
50
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
10
15
20
25
30
35
40
45
50
-1
0
1
-1
0
1
10
20
30
40
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
5
10
15
20
25
30
35
40
45
-1
0
1
-1
0
1
20
30
40
50
60
70
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
20
30
40
50
60
70
-1
0
1
-1
0
1
10
20
30
40
50
60
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
10
20
30
40
50
60
-1
0
1
-1
0
1
10
20
30
40
50
60
70
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
10
20
30
40
50
60
70
-1
0
1
-1
0
1
10
20
30
40
50
60
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
10
20
30
40
50
60
1
1
.
2
1
.
4
1
.
6
1
.
8
2
2
.
2
2
.
4
2
.
6
2
.
8
3
1
1
.
5
2
2
.
5
3
3
.
5
4
4
.
5
N
u
m
b
e
r
o
f
i
t
e
r
a
t
i
o
n
s
N
o
r
m
a
l
i
z
e
d
e
r
r
o
r
D
B
I
M
w
i
t
h
S
N
R
=
6
d
B
K
N
N
-
f
i
l
t
e
r
e
d
D
B
I
M
w
i
t
h
S
N
R
=
6
d
B
1
1
.
2
1
.
4
1
.
6
1
.
8
2
2
.
2
2
.
4
2
.
6
2
.
8
3
1
1
.
5
2
2
.
5
3
3
.
5
4
4
.
5
5
5
.
5
N
u
m
b
e
r
o
f
i
t
e
r
a
t
i
o
n
s
N
o
r
m
a
l
i
z
e
d
e
r
r
o
r
D
B
I
M
w
i
t
h
S
N
R
=
3
d
B
K
N
N
-
f
i
l
t
e
r
e
d
D
B
I
M
w
i
t
h
S
N
R
=
3
d
B
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
2
0
6
-
2
1
8
214
Fig
u
r
e
8
p
lo
ts
n
o
r
m
alize
d
er
r
o
r
ac
r
o
s
s
th
e
f
ir
s
t
t
h
r
ee
iter
atio
n
s
at
SNR
=
1
d
B
,
d
em
o
n
s
tr
atin
g
f
aster
er
r
o
r
r
ed
u
ctio
n
f
o
r
t
h
e
KNN
-
f
ilter
ed
s
ch
e
m
e.
T
h
e
r
es
u
lt
d
em
o
n
s
tr
ates
th
e
ef
f
ec
ti
v
e
n
es
s
o
f
t
h
e
K
NN
-
f
i
lter
ed
DB
I
M
ap
p
r
o
ac
h
in
r
ed
u
ci
n
g
r
ec
o
n
s
tr
u
ctio
n
er
r
o
r
s
co
m
p
ar
ed
to
co
n
v
en
tio
n
al
DB
I
M
u
n
d
er
h
ig
h
-
n
o
is
e
co
n
d
itio
n
s
.
O
v
er
th
r
ee
iter
ati
o
n
s
,
th
e
n
o
r
m
alize
d
er
r
o
r
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
r
e
m
ai
n
s
co
n
s
is
te
n
tl
y
lo
w
er
th
an
t
h
at
o
f
t
h
e
tr
ad
itio
n
a
l
DB
I
M.
I
n
t
h
e
f
ir
s
t
iter
atio
n
,
t
h
e
e
r
r
o
r
r
ed
u
ctio
n
is
p
ar
tic
u
lar
l
y
s
ig
n
if
ican
t,
w
it
h
t
h
e
KNN
-
f
il
ter
ed
DB
I
M
ac
h
iev
in
g
a
5
7
.
2
%
lo
w
er
er
r
o
r
th
an
t
h
e
co
n
v
e
n
tio
n
al
ap
p
r
o
ac
h
.
T
h
is
tr
e
n
d
co
n
tin
u
e
s
ac
r
o
s
s
s
u
b
s
eq
u
e
n
t
iter
atio
n
s
,
s
h
o
w
i
n
g
an
o
v
er
all
i
m
p
r
o
v
e
m
en
t
in
r
ec
o
n
s
tr
u
ct
io
n
ac
cu
r
a
c
y
.
T
h
e
r
esu
lts
in
d
icate
th
at
KNN
-
b
ased
d
en
o
is
in
g
ef
f
ec
ti
v
el
y
r
e
m
o
v
es
n
o
is
e
b
ef
o
r
e
th
e
iter
ativ
e
r
ec
o
n
s
tr
u
ctio
n
p
r
o
ce
s
s
,
lead
in
g
to
b
etter
in
it
ial
co
n
d
iti
o
n
s
an
d
i
m
p
r
o
v
ed
s
tab
ilit
y
.
W
e
f
u
r
th
er
ex
a
m
i
n
ed
a
m
o
r
e
co
m
p
le
x
s
ce
n
ar
io
co
n
s
is
ti
n
g
o
f
t
w
o
cir
c
u
lar
o
b
j
ec
ts
lo
ca
ted
w
it
h
i
n
th
e
r
eg
io
n
o
f
in
ter
e
s
t,
as
illu
s
tr
a
ted
b
y
th
e
id
ea
l
o
b
j
ec
t
f
u
n
ct
io
n
in
Fi
g
u
r
e
9
.
T
h
is
co
n
f
ig
u
r
atio
n
in
tr
o
d
u
ce
s
s
tr
o
n
g
er
m
u
l
tip
le
-
s
ca
tter
i
n
g
i
n
ter
ac
tio
n
s
an
d
o
v
er
lap
p
in
g
d
if
f
r
ac
tio
n
p
atter
n
s
,
p
o
s
in
g
a
m
o
r
e
ch
al
len
g
i
n
g
r
ec
o
n
s
tr
u
ctio
n
p
r
o
b
lem
co
m
p
ar
ed
w
it
h
th
e
s
in
g
le
-
o
b
j
ec
t
ca
s
e.
T
h
e
r
ec
o
n
s
tr
u
ctio
n
p
er
f
o
r
m
an
ce
o
f
th
e
co
n
v
e
n
tio
n
al
DB
I
M
an
d
th
e
p
r
o
p
o
s
ed
KNN
-
f
ilter
ed
DB
I
M
w
a
s
ev
al
u
ated
an
d
co
m
p
ar
ed
u
n
d
er
th
i
s
s
etti
n
g
at
an
in
p
u
t
S
NR
o
f
1
d
B
.
T
h
e
n
o
r
m
al
ized
r
ec
o
n
s
tr
u
ctio
n
er
r
o
r
s
f
o
r
th
r
ee
s
u
cc
e
s
s
iv
e
iter
atio
n
s
w
er
e
7
.
3
4
,
4
.
0
9
,
an
d
2
.
4
1
f
o
r
DB
I
M,
an
d
4
.
9
9
,
3
.
5
8
,
an
d
2
.
0
8
f
o
r
th
e
KNN
-
f
ilter
ed
DB
I
M,
r
esp
ec
tiv
el
y
.
T
h
e
r
esu
lts
clea
r
l
y
in
d
icate
t
h
at
t
h
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
ac
h
iev
es
f
a
s
ter
er
r
o
r
r
ed
u
ctio
n
an
d
s
u
p
er
io
r
r
ec
o
n
s
tr
u
ctio
n
f
id
elit
y
,
ef
f
ec
tiv
e
l
y
m
iti
g
ati
n
g
n
o
is
e
a
n
d
p
r
eser
v
in
g
t
h
e
s
tr
u
ct
u
r
al
b
o
u
n
d
ar
ies o
f
b
o
th
i
n
clu
s
io
n
s
.
F
ig
u
r
e
8
.
T
h
e
n
o
r
m
aliza
tio
n
er
r
o
r
af
ter
th
e
f
ir
s
t t
h
r
ee
iter
atio
n
s
u
s
in
g
t
h
e
co
n
v
e
n
tio
n
al
DB
I
M
an
d
p
r
o
p
o
s
ed
KNN
-
f
il
t
er
ed
DB
I
M
m
et
h
o
d
in
ca
s
e
o
f
S
NR
=
1
d
B
Fig
u
r
e
9
.
I
d
ea
l o
b
j
ec
t f
u
n
ctio
n
(
t
w
o
o
b
j
ec
ts
in
th
e
r
eg
io
n
o
f
in
ter
est).
C
o
lo
r
b
ar
in
d
icate
s
p
er
ce
n
t o
f
s
o
u
n
d
-
s
p
ee
d
co
n
tr
ast
(
%)
T
h
e
ef
f
ec
ti
v
e
n
ess
o
f
KNN
d
en
o
is
i
n
g
in
DB
I
M
d
ep
en
d
s
o
n
f
ac
to
r
s
s
u
c
h
as
n
o
is
e
le
v
el,
o
b
j
ec
t
co
m
p
le
x
it
y
,
an
d
s
p
atial
r
eso
lu
tio
n
.
I
t
is
p
ar
ticu
lar
ly
b
en
e
f
ici
al
w
h
e
n
th
e
d
ataset
h
a
s
m
o
d
er
ate
to
h
ig
h
lev
el
s
o
f
n
o
is
e,
w
h
er
e
tr
ad
itio
n
a
l
r
eg
u
l
ar
izatio
n
m
et
h
o
d
s
f
a
il
to
m
ai
n
tain
i
m
ag
e
q
u
alit
y
;
t
h
e
r
ec
o
n
s
tr
u
ctio
n
p
r
o
b
le
m
is
h
ig
h
l
y
il
l
-
co
n
d
itio
n
ed
,
r
eq
u
ir
in
g
r
o
b
u
s
t
d
en
o
is
i
n
g
tech
n
iq
u
es;
f
i
n
e
s
tr
u
ct
u
r
al
d
etails
n
ee
d
to
b
e
p
r
eser
v
ed
,
s
u
c
h
as
in
m
ed
ical
u
ltra
s
o
u
n
d
im
a
g
i
n
g
o
r
n
o
n
-
d
estr
u
cti
v
e
test
in
g
.
Si
m
u
latio
n
r
es
u
lts
c
o
n
f
ir
m
th
at
KN
N
d
en
o
is
in
g
co
n
s
is
te
n
tl
y
i
m
p
r
o
v
es
DB
I
M
p
er
f
o
r
m
an
ce
b
y
r
e
d
u
cin
g
r
ec
o
n
s
tr
u
c
tio
n
ar
ti
f
ac
ts
an
d
en
h
a
n
cin
g
SNR
.
T
h
e
p
r
o
p
o
s
ed
KNN
-
b
ased
DB
I
M
f
r
a
m
e
w
o
r
k
ca
n
b
e
ad
v
an
ta
g
eo
u
s
in
v
ar
io
u
s
b
io
m
e
d
ical
an
d
in
d
u
s
tr
ia
l
ap
p
licatio
n
s
,
in
cl
u
d
in
g
m
ed
ic
al
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
y
f
o
r
h
ig
h
-
r
eso
l
u
tio
n
tis
s
u
e
i
m
ag
in
g
;
b
r
ea
s
t
ca
n
ce
r
d
etec
t
io
n
,
w
h
er
e
i
m
p
r
o
v
ed
s
o
u
n
d
-
s
p
ee
d
m
ap
s
en
ab
le
b
etter
lesi
o
n
c
h
ar
ac
ter
izati
o
n
;
n
o
n
-
d
es
tr
u
cti
v
e
ev
alu
a
tio
n
(
NDE
)
o
f
m
ater
ial
s
u
s
i
n
g
u
ltra
s
o
u
n
d
-
b
ased
i
m
a
g
in
g
.
B
y
p
r
o
v
id
in
g
b
etter
n
o
is
e
s
u
p
p
r
ess
io
n
an
d
r
ec
o
n
s
tr
u
ctio
n
ac
c
u
r
ac
y
,
t
h
is
ap
p
r
o
ac
h
ca
n
m
a
k
e
u
ltra
s
o
u
n
d
t
o
m
o
g
r
ap
h
y
m
o
r
e
v
iab
l
e
f
o
r
cli
n
ical
a
n
d
in
d
u
s
tr
ial
ad
o
p
tio
n
.
B
y
ap
p
l
y
i
n
g
KNN
-
b
ased
m
ac
h
i
n
e
lear
n
in
g
d
en
o
is
i
n
g
i
n
DB
I
M,
t
h
is
s
tu
d
y
i
n
tr
o
d
u
ce
s
a
ef
f
icien
t
ap
p
r
o
ac
h
to
en
h
an
c
e
r
ec
o
n
s
tr
u
ctio
n
s
tab
ilit
y
,
n
o
is
e
r
o
b
u
s
tn
e
s
s
,
an
d
co
n
v
er
g
e
n
ce
s
p
ee
d
.
Un
lik
e
co
n
v
e
n
tio
n
al
r
eg
u
lar
izatio
n
tech
n
iq
u
es,
KNN
ef
f
ec
t
iv
el
y
ad
ap
ts
to
lo
ca
l
v
ar
iatio
n
s
in
th
e
d
ataset,
m
a
k
i
n
g
i
t
h
ig
h
l
y
e
f
f
ec
ti
v
e
f
o
r
in
v
er
s
e
s
ca
tter
in
g
p
r
o
b
lem
s
.
T
h
e
ex
p
er
im
e
n
tal
r
es
u
lts
d
e
m
o
n
s
tr
ate
s
ig
n
i
f
ica
n
t
i
m
p
r
o
v
e
m
en
t
s
in
i
m
a
g
e
q
u
a
lit
y
,
er
r
o
r
r
ed
u
ctio
n
,
an
d
iter
at
iv
e
ef
f
icie
n
c
y
,
co
n
f
ir
m
i
n
g
th
at
KNN
-
b
ased
d
en
o
is
in
g
i
s
a
p
r
o
m
i
s
in
g
e
n
h
a
n
ce
m
e
n
t
f
o
r
DB
I
M
in
p
r
ac
tical
b
io
m
ed
ical
i
m
a
g
i
n
g
ap
p
licati
o
n
s
.
1
1
.
2
1
.
4
1
.
6
1
.
8
2
2
.
2
2
.
4
2
.
6
2
.
8
3
1
2
3
4
5
6
7
8
N
u
m
b
e
r
o
f
i
t
e
r
a
t
i
o
n
s
N
o
r
m
a
l
i
z
e
d
e
r
r
o
r
D
B
I
M
w
i
t
h
S
N
R
=
1
d
B
K
N
N
-
f
i
l
t
e
r
e
d
D
B
I
M
w
i
t
h
S
N
R
=
1
d
B
-1
0
1
-1
0
1
0
5
10
15
20
25
30
p
e
r
c
e
n
t
o
f
t
h
e
s
o
u
n
d
c
o
n
t
r
a
s
t
0
5
10
15
20
25
30
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
Dis
to
r
ted
b
o
r
n
iter
a
tive
meth
o
d
r
ec
o
n
s
tr
u
ctio
n
in
h
ig
h
-
n
o
is
e
en
viro
n
men
ts
u
s
in
g
…
(
N
g
u
ye
n
Qu
a
n
g
Hu
y
)
215
T
o
f
u
r
th
er
v
alid
ate
th
e
r
o
b
u
s
t
n
es
s
o
f
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
,
w
e
ex
te
n
d
ed
th
e
an
al
y
s
i
s
b
y
i
n
tr
o
d
u
cin
g
an
o
th
er
li
g
h
t
w
ei
g
h
t
d
e
n
o
is
i
n
g
b
aseli
n
e
-
th
e
m
ed
ia
n
f
ilter
-
w
it
h
i
n
t
h
e
DB
I
M
iter
ativ
e
f
r
a
m
e
w
o
r
k
.
T
h
is
co
m
p
ar
is
o
n
ai
m
s
to
ex
a
m
i
n
e
w
h
et
h
er
t
h
e
i
m
p
r
o
v
e
m
e
n
t
a
ch
iev
ed
b
y
t
h
e
KNN
-
b
ased
d
en
o
is
in
g
o
r
ig
in
ate
s
f
r
o
m
its
ad
ap
tiv
e
n
eig
h
b
o
r
h
o
o
d
w
ei
g
h
ti
n
g
o
r
m
er
el
y
f
r
o
m
g
en
er
ic
lo
ca
l
s
m
o
o
t
h
i
n
g
.
Fi
g
u
r
e
1
0
illu
s
tr
ates
t
h
e
n
o
r
m
alize
d
r
ec
o
n
s
tr
u
ctio
n
er
r
o
r
s
ac
r
o
s
s
th
r
ee
s
u
cc
ess
iv
e
i
t
er
atio
n
s
f
o
r
co
n
v
e
n
tio
n
al
DB
I
M,
m
ed
ia
n
-
f
ilter
ed
DB
I
M,
an
d
th
e
p
r
o
p
o
s
ed
KN
N
-
f
ilter
ed
DB
I
M
at
an
in
p
u
t
SNR
o
f
1
d
B
.
T
h
e
q
u
an
titati
v
e
r
esu
lt
s
s
h
o
w
t
h
at
th
e
p
r
o
p
o
s
ed
KNN
-
f
ilte
r
ed
DB
I
M
ac
h
ie
v
es
t
h
e
lo
w
est
r
ec
o
n
s
tr
u
ctio
n
er
r
o
r
(
3
.
4
0
→1
.
6
5
)
co
m
p
ar
ed
w
it
h
m
ed
ia
n
-
f
ilter
ed
DB
I
M
(
6
.
7
7
→2
.
4
1
)
an
d
co
n
v
e
n
tio
n
al
DB
I
M
(
7
.
9
4
→3
.
2
3
)
.
T
h
is
r
ep
r
esen
ts
a
n
er
r
o
r
r
ed
u
ctio
n
o
f
ap
p
r
o
x
im
a
tel
y
4
9
%
r
elativ
e
to
DB
I
M
an
d
3
2
%
r
elativ
e
to
m
ed
ian
f
ilter
i
n
g
af
ter
th
e
t
h
ir
d
iter
atio
n
.
T
h
e
KNN
-
b
ased
ap
p
r
o
ac
h
co
n
v
er
g
es
f
a
s
ter
an
d
p
r
eser
v
es
s
tr
u
ct
u
r
al
b
o
u
n
d
ar
ies
m
o
r
e
ef
f
ec
ti
v
el
y
,
d
em
o
n
s
tr
ati
n
g
it
s
s
u
p
er
io
r
ab
ilit
y
to
s
u
p
p
r
ess
n
o
is
e
w
it
h
o
u
t
o
v
er
s
m
o
o
th
i
n
g
o
b
j
ec
t
ed
g
es.
Un
li
k
e
t
h
e
m
ed
ian
f
il
ter
,
w
h
ic
h
ap
p
lies
u
n
i
f
o
r
m
lo
ca
l
av
er
ag
in
g
an
d
ten
d
s
to
b
lu
r
f
in
e
d
etails,
th
e
KNN
f
ilter
ad
ap
tiv
el
y
weig
h
ts
n
ei
g
h
b
o
r
in
g
p
ix
els
b
ased
o
n
s
i
m
ilar
it
y
,
le
ad
in
g
to
e
n
h
a
n
ce
d
ed
g
e
f
id
el
it
y
a
n
d
s
tab
ilit
y
i
n
lo
w
-
SN
R
co
n
d
itio
n
s
.
T
h
ese
f
i
n
d
in
g
s
co
n
f
ir
m
th
at
t
h
e
p
r
o
p
o
s
ed
KNN
-
f
i
lter
ed
DB
I
M
p
r
o
v
id
es
a
m
o
r
e
ac
cu
r
ate
an
d
r
o
b
u
s
t
r
ec
o
n
s
tr
u
c
tio
n
f
r
a
m
e
w
o
r
k
w
h
i
le
m
ai
n
tai
n
i
n
g
co
m
p
u
tatio
n
al
ef
f
icie
n
c
y
co
m
p
ar
ab
le
to
o
th
er
lig
h
t
w
ei
g
h
t d
e
n
o
is
er
s
.
Fig
u
r
e
1
0
.
No
r
m
alize
d
r
ec
o
n
s
t
r
u
ctio
n
er
r
o
r
v
er
s
u
s
iter
atio
n
n
u
m
b
er
f
o
r
t
h
e
th
r
ee
m
eth
o
d
s
-
c
o
n
v
e
n
tio
n
al
DB
I
M,
m
ed
ia
n
-
f
il
ter
ed
DB
I
M
,
an
d
KNN
-
f
ilter
ed
DB
I
M
-
at
SNR
=
1
d
B
4.
RE
SU
L
T
AND
DI
SCUS
SI
O
N
T
h
e
ce
n
tr
al
co
n
tr
ib
u
tio
n
o
f
t
h
is
w
o
r
k
lies
i
n
th
e
i
n
teg
r
at
i
o
n
o
f
a
s
i
m
p
le
y
et
ef
f
ec
t
iv
e
KNN
-
b
ased
d
en
o
is
in
g
m
ec
h
a
n
is
m
i
n
to
th
e
DB
I
M
r
ec
o
n
s
tr
u
ctio
n
f
r
a
m
e
wo
r
k
.
Un
lik
e
co
n
v
e
n
tio
n
a
l
DB
I
M
ap
p
r
o
ac
h
es
th
at
r
el
y
s
o
lel
y
o
n
r
eg
u
lar
izatio
n
o
r
h
ea
v
y
d
ee
p
-
lear
n
in
g
-
b
ased
p
o
s
t
-
p
r
o
ce
s
s
i
n
g
,
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
e
m
b
ed
s
ad
ap
tiv
e,
d
ata
-
d
r
iv
en
n
o
is
e
s
u
p
p
r
ess
io
n
d
i
r
ec
tl
y
w
it
h
i
n
ea
ch
iter
atio
n
.
T
h
is
co
n
ce
p
tu
al
m
o
d
if
icatio
n
e
n
h
a
n
ce
s
co
n
v
er
g
e
n
ce
s
tab
ilit
y
a
n
d
r
ec
o
n
s
tr
u
ct
io
n
f
id
elit
y
u
n
d
er
s
tr
o
n
g
n
o
is
e
co
n
d
itio
n
s
w
h
ile
m
ai
n
tai
n
in
g
lo
w
co
m
p
u
tatio
n
al
co
m
p
le
x
it
y
.
T
h
e
ap
p
r
o
ac
h
d
em
o
n
s
tr
ate
s
th
at
class
ical,
n
o
n
-
p
ar
a
m
etr
ic
t
ec
h
n
iq
u
es
s
u
c
h
a
s
KNN
ca
n
b
e
s
u
cc
es
s
f
u
l
l
y
h
y
b
r
id
ized
w
it
h
iter
ati
v
e
to
m
o
g
r
ap
h
ic
alg
o
r
ith
m
s
to
ac
h
ie
v
e
r
o
b
u
s
t
an
d
ef
f
icie
n
t
u
ltra
s
o
u
n
d
i
m
a
g
e
r
ec
o
n
s
tr
u
c
tio
n
.
T
h
is
s
tu
d
y
ad
d
r
ess
es
th
e
c
h
alle
n
g
e
o
f
n
o
is
e
s
en
s
iti
v
it
y
an
d
s
lo
w
co
n
v
er
g
e
n
ce
i
n
co
n
v
e
n
tio
n
al
DB
I
M
r
ec
o
n
s
tr
u
ctio
n
s
,
p
ar
ticu
lar
l
y
u
n
d
er
SNR
co
n
d
itio
n
s
.
T
o
o
v
er
co
m
e
t
h
i
s
li
m
ita
tio
n
,
a
K
NN
-
b
a
s
ed
d
en
o
is
in
g
s
tep
w
a
s
i
n
te
g
r
ated
in
t
o
th
e
DB
I
M
iter
ati
v
e
lo
o
p
,
f
o
r
m
in
g
t
h
e
p
r
o
p
o
s
ed
KNN
-
f
il
ter
ed
DB
I
M
alg
o
r
ith
m
.
Si
m
u
latio
n
r
es
u
lt
s
d
e
m
o
n
s
tr
ated
th
at
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
s
ig
n
i
f
ica
n
tl
y
i
m
p
r
o
v
es
r
ec
o
n
s
tr
u
ct
io
n
f
id
el
it
y
a
n
d
co
n
v
er
g
e
n
ce
s
p
ee
d
co
m
p
ar
ed
w
it
h
b
o
th
s
tan
d
ar
d
DB
I
M
an
d
m
ed
ian
-
f
ilter
ed
DB
I
M,
w
h
ile
m
ai
n
tain
in
g
co
m
p
u
ta
tio
n
al
e
f
f
icien
c
y
.
T
h
ese
f
in
d
i
n
g
s
s
u
g
g
e
s
t t
h
at
i
n
co
r
p
o
r
atin
g
s
i
m
p
le,
ad
ap
tiv
e,
n
o
n
-
p
ar
a
m
etr
ic
f
i
lte
r
s
s
u
c
h
as
KNN
in
to
iter
ativ
e
i
m
ag
i
n
g
f
r
a
m
e
w
o
r
k
s
ca
n
o
f
f
er
a
p
r
ac
tical
an
d
lig
h
t
w
ei
g
h
t
alter
n
a
tiv
e
to
co
m
p
le
x
d
ee
p
-
lear
n
in
g
-
b
ased
d
en
o
is
in
g
ap
p
r
o
ac
h
es
f
o
r
r
o
b
u
s
t
to
m
o
g
r
ap
h
ic
u
ltra
s
o
u
n
d
r
ec
o
n
s
tr
u
ctio
n
.
T
h
is
s
t
u
d
y
h
as
d
e
m
o
n
s
tr
ated
th
e
e
f
f
ec
t
iv
e
n
es
s
o
f
in
co
r
p
o
r
at
in
g
a
KNN
-
b
ased
d
en
o
is
i
n
g
s
tr
ateg
y
in
to
th
e
DB
I
M
f
r
a
m
e
w
o
r
k
to
i
m
p
r
o
v
e
th
e
q
u
alit
y
a
n
d
r
o
b
u
s
tn
e
s
s
o
f
u
ltra
s
o
u
n
d
to
m
o
g
r
ap
h
ic
r
ec
o
n
s
tr
u
ct
io
n
u
n
d
er
7.9431
5.1155
3.2275
6.7668
4.8291
2.414
3.3998
2.9517
1.6482
0
1
2
3
4
5
6
7
8
9
1
2
3
N
o
r
m
al
i
ze
d
e
r
r
o
r
N
u
m
b
e
r
o
f i
te
r
ation
s
DBIM
Me
d
i
an
-fil
te
re
d
DBI
M
KN
N
-fil
te
re
d
D
B
IM
Evaluation Warning : The document was created with Spire.PDF for Python.