T
E
L
KO
M
N
I
KA
T
e
lec
om
m
u
n
icat
ion
,
Com
p
u
t
i
n
g,
E
lec
t
r
on
ics
an
d
Cont
r
ol
Vol.
18
,
No.
5
,
Oc
tober
2020
,
pp
.
2391~2400
I
S
S
N:
1693
-
6930
,
a
c
c
r
e
dit
e
d
F
ir
s
t
G
r
a
de
by
Ke
me
nr
is
tekdikti
,
De
c
r
e
e
No:
21/E
/KP
T
/2018
DO
I
:
10.
12928/
T
E
L
KO
M
NI
KA
.
v18i5.
14321
2391
Jou
r
n
al
h
omepage
:
ht
tp:
//
jour
nal.
uad
.
ac
.
id/
index
.
php/T
E
L
K
OM
N
I
K
A
D
e
ve
lo
p
e
d
a
p
p
r
oac
h
f
o
r
p
h
as
e
-
b
as
e
d
E
u
le
r
ia
n
vi
d
e
o m
ag
n
ific
a
t
io
n
Hai
d
e
r
I
s
m
ae
l
S
h
ah
ad
i
1
,
Z
aid
Ja
b
b
ar
Al
-
a
ll
aq
2
,
Hayd
e
r
Jawad
Al
b
at
t
a
t
3
1
E
l
ec
t
ri
ca
l
an
d
E
l
ec
t
ro
n
i
c
E
n
g
i
n
eeri
n
g
D
e
p
art
me
n
t
,
U
n
i
v
ers
i
t
y
o
f
K
er
b
al
a,
Iraq
2
,3
Co
mmu
n
i
cat
i
o
n
T
ech
n
i
q
u
e
s
E
n
g
i
n
eer
i
n
g
D
ep
a
rt
me
n
t
,
A
l
-
Fu
rat
A
l
-
A
w
s
a
t
T
ec
h
n
i
cal
U
n
i
v
ers
i
t
y
,
Iraq
2
T
ech
n
i
ca
l
In
s
t
i
t
u
t
e
o
f
K
ar
b
al
a,
A
l
-
Fu
r
at
A
l
-
A
w
s
a
t
T
ec
h
n
i
ca
l
U
n
i
v
er
s
i
t
y
,
Iraq
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
Oc
t
13,
2019
R
e
vis
e
d
Apr
21,
2020
Ac
c
e
pted
M
a
y
1,
2020
T
h
i
s
p
ap
er
p
ro
p
o
s
es
a
m
o
d
i
fi
ca
t
i
o
n
a
p
p
r
o
ac
h
fo
r
p
h
a
s
ed
-
b
as
e
d
E
V
M
i
n
o
r
d
er
t
o
red
u
ce
t
h
e
p
ro
c
es
s
i
n
g
t
i
me
w
i
t
h
o
u
t
effect
t
h
e
q
u
a
l
i
t
y
o
f
t
h
e
mag
n
i
f
i
ed
v
i
d
e
o
.
T
h
e
p
ro
p
o
s
ed
ap
p
ro
ac
h
ap
p
l
i
es
a
res
i
zi
n
g
p
r
o
ces
s
o
n
t
h
e
i
n
p
u
t
v
i
d
e
o
u
s
i
n
g
L
an
czo
s
-
3
al
g
o
ri
t
h
m.
T
h
e
n
,
i
t
d
eco
m
p
o
s
es
v
i
d
e
o
frames
u
s
i
n
g
s
t
eerab
l
e
p
y
ram
i
d
t
o
o
b
t
ai
n
m
u
l
t
i
-
s
cal
e
frame
w
i
t
h
i
t
s
o
r
i
en
t
at
i
o
n
.
Su
b
s
e
q
u
en
t
l
y
,
t
h
e
res
u
l
t
e
d
frames
are
fi
l
t
ered
b
y
t
emp
o
ral
fi
l
t
ers
fo
r
s
p
eci
fi
c
b
a
n
d
s
an
d
t
h
e
fi
l
t
ere
d
frames
are
mu
l
t
i
p
l
i
ed
b
y
a
mag
n
i
fi
cat
i
o
n
fact
o
r.
N
o
w
,
b
o
t
h
t
h
e
ma
g
n
i
fi
e
d
re
g
i
o
n
s
an
d
t
h
e
u
n
mag
n
i
fi
e
d
reg
i
o
n
s
fo
r
eac
h
frame
are
ad
d
e
d
t
o
g
e
t
h
er.
F
i
n
a
l
l
y
,
reco
n
s
t
ru
c
t
i
n
g
t
h
e
p
r
o
d
u
ced
mag
n
i
fi
e
d
mu
l
t
i
-
s
ca
l
e
frames
u
s
i
n
g
t
h
e
i
n
v
ers
e
s
t
eera
b
l
e
p
y
rami
d
.
T
h
e
e
x
p
er
i
men
t
al
res
u
l
t
s
s
h
o
w
t
h
a
t
s
u
p
eri
o
ri
t
y
o
f
t
h
e
p
r
o
p
o
s
e
d
ap
p
r
o
ach
co
m
p
are
s
t
o
t
h
e
co
n
v
e
n
t
i
o
n
al
p
h
a
s
e
-
b
a
s
ed
E
V
M
i
n
p
ro
ce
s
s
i
n
g
t
i
me,
w
h
ere
t
h
e
p
ro
ce
s
s
i
n
g
t
i
me
r
ed
u
ct
i
o
n
ab
o
u
t
6
0
-
6
5
%
.
Fu
r
t
h
ermo
re,
t
h
i
s
ap
p
ro
ac
h
d
o
e
s
n
o
t
affec
t
o
n
t
h
e
v
i
d
eo
q
u
al
i
t
y
,
w
h
i
ch
mai
n
t
a
i
n
i
t
i
n
t
h
e
b
o
u
n
d
ary
o
f
t
h
e
c
o
n
v
en
t
i
o
n
a
l
Ph
as
e
-
b
as
e
d
E
V
M.
K
e
y
w
o
r
d
s
:
E
uler
ian
video
magnif
ica
ti
on
L
a
nc
z
o
s
-
3
a
lgor
it
hm
L
inea
r
-
b
a
s
e
d
M
oti
on
magnif
ica
ti
on
P
ha
s
e
-
b
a
s
e
d
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e
.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Ha
ider
I
s
mae
l
S
ha
ha
di
,
E
lec
tr
ica
l
a
nd
E
lec
tr
onic
E
nginee
r
ing
De
p
a
r
t
ment
,
Unive
r
s
it
y
of
Ke
r
ba
la
,
56001,
Ka
r
ba
la,
I
r
a
q
.
E
mail:
c
or
r
e
s
p
-
a
uthor
@ma
il
.
c
om
1.
I
NT
RODU
C
T
I
ON
Na
tur
a
l
li
mi
tations
of
human
vis
ion
s
ys
tem
(
HV
S
)
to
pick
up
im
po
r
tant
a
nd
us
e
f
ul
s
igns
make
the
video
magni
f
ica
ti
on
tec
hniques
a
s
a
powe
r
f
ul
ke
y
f
or
s
e
ve
r
a
l
ne
e
ds
in
he
a
lt
h
c
a
r
e
[1
-
6]
,
phys
ica
l
va
r
iations
[
7,
8]
,
mate
r
ial
be
ha
viour
de
tec
ti
on
[
8
]
,
e
tc.
T
he
s
e
ns
ing
of
ve
r
y
s
low
mot
ion
or
ve
r
y
s
mall
c
ha
nge
in
c
olor
is
ha
r
d
to
be
obs
e
r
ve
d
by
na
ke
d
e
ye
of
human.
L
iu
e
t
.
a
l
.
[
7]
ha
ve
int
r
oduc
e
d
a
mot
ion
a
nd
c
olor
magnif
ica
ti
on
tec
hniques
de
pe
nd
on
de
ter
mi
ning
of
the
va
r
iations
in
mot
ion
o
r
c
olor
via
a
r
obus
t
inve
s
ti
ga
ti
on
of
tr
a
jec
tor
ies
.
T
he
y
ga
ther
e
d
thes
e
tr
a
jec
tor
ies
to
de
r
ive
a
mpl
if
ied
mot
ion
o
r
c
olo
r
t
r
a
c
e
s
f
or
a
ll
pixe
ls
r
e
late
d
to
the
r
e
gis
ter
e
d
r
e
f
e
r
e
nc
e
f
r
a
me
.
T
his
tec
hnique
r
e
quir
e
s
a
long
pr
oc
e
s
s
ing
ti
me
a
nd
it
r
e
s
ult
s
a
lar
ge
no
is
e
in
the
magnif
ied
video
f
r
a
mes
.
T
he
method
in
[
7
]
is
ba
s
e
d
on
L
a
gr
a
ngian
a
nd
is
c
omput
a
ti
ona
ll
y
e
x
pe
ns
ive,
with
long
e
xe
c
uti
on
ti
me
(
10
hour
s
)
be
c
a
us
e
it
r
e
l
ied
on
opti
c
a
l
f
low
c
a
lcula
ti
ons
[
9]
a
nd
a
f
e
a
tur
e
tr
a
c
king
a
lgor
it
hm.
S
o
that
Ha
o
e
t.
Al
[
8]
ha
ve
p
r
opos
e
d
E
uler
ian
-
ba
s
e
d
video
magnif
ica
ti
on
(
E
VM
)
that
is
n
ot
f
o
ll
ow
the
pixels
mot
ion
s
uc
h
in
L
a
gr
a
ngian
a
ppr
oa
c
h.
T
h
e
r
e
f
or
e
,
it
is
f
a
s
ter
than
the
L
a
gr
a
ngian
-
ba
s
e
d
method.
T
he
f
ir
s
t
ve
r
s
ion
o
f
E
VM
uti
li
z
e
s
T
a
ylor
s
e
r
ies
e
xpa
ns
ion
that
ha
s
a
li
ne
a
r
pr
ope
r
ti
e
s
,
ther
e
f
o
r
e
,
it
is
c
a
ll
e
d
li
ne
a
r
-
ba
s
e
d
E
VM
a
lgor
it
hm
[
8]
.
T
he
a
lgor
it
hm
is
r
e
latively
f
a
s
t,
but
it
doe
s
not
s
uppor
t
e
d
high
magnif
ica
ti
on
f
a
c
tor
a
s
we
ll
nois
y
magni
f
ied
vide
o.
T
he
r
e
a
s
on
be
hind
the
high
nois
e
is
due
to
the
l
inea
r
it
y,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
5
,
Oc
tober
2020:
2391
-
2400
2392
whe
r
e
the
nois
e
is
a
mpl
if
ied
by
the
s
a
me
r
a
ti
o
o
f
the
magnif
ica
ti
on
f
a
c
tor
of
the
video
f
r
a
mes
.
I
n
or
de
r
to
r
e
duc
e
thes
e
dr
a
wba
c
ks
,
W
a
dhwa
e
t
a
l
.
[
10]
ha
ve
pr
opos
e
d
a
ne
w
E
uler
ian
a
ppr
oa
c
h
ba
s
e
d
on
c
ompl
e
x
s
tee
r
a
ble
pyr
a
mi
ds
[
11,
12]
,
whic
h
is
s
ti
mul
a
ted
b
y
pha
s
e
-
ba
s
e
d
opti
c
a
l
f
low
a
lgo
r
it
hms
[
13
]
.
T
he
i
r
pr
opos
e
d
a
ppr
oa
c
h
ha
s
f
e
we
r
a
r
te
f
a
c
ts
a
nd
le
s
s
nois
e
a
s
we
ll
a
s
s
uppor
ts
lar
ge
r
magnif
ica
ti
on
f
a
c
tor
s
c
o
mpar
e
d
to
li
ne
r
-
ba
s
e
d
E
VM
(
L
B
-
E
VM
)
.
Als
o,
it
c
a
n
a
tt
e
nua
te
unwa
nted
mot
ion
in
c
a
s
e
of
c
olor
magnif
ica
ti
on.
H
owe
ve
r
,
it
is
s
o
c
ompl
e
x
a
nd
r
e
quir
e
s
longer
ti
me
to
e
xe
c
ut
e
than
the
L
B
-
E
VM
.
S
e
ve
r
a
l
w
or
ks
ha
ve
be
e
n
publi
s
he
d
r
e
c
e
ntl
y
in
or
de
r
to
e
nha
nc
e
E
VM
,
whe
ther
in
li
ne
a
r
ba
s
e
d
o
r
pha
s
e
-
b
a
s
e
d
[
14]
.
L
e
L
iu
e
t
a
l.
[
15]
p
r
opos
e
d
a
n
e
n
ha
nc
e
ment
to
E
VM
to
r
e
duc
e
the
nois
e
in
L
B
-
E
VM
.
S
o
that
a
uthor
s
ha
ve
us
e
d
a
pixel
-
leve
l
mot
ion
a
na
lyze
r
t
o
c
a
ptur
e
the
s
pa
ti
o
-
tempor
a
l
mot
ion
a
nd
then
a
mpl
if
y
i
t.
T
he
a
ppr
oa
c
h
of
mot
ion
uti
li
z
e
s
a
n
im
a
ge
wa
r
ping
i
s
a
ls
o
us
e
d
in
or
de
r
to
a
mpl
if
y
the
mot
ion
in
a
video
r
e
lyi
ng
s
pe
c
if
ic
f
or
mer
mot
ion
m
a
pping
.
Al
though
the
m
e
thod
im
pr
ove
s
nois
e
pe
r
f
or
manc
e
in
the
p
r
oc
e
s
s
e
d
video
f
r
a
mes
,
the
c
omput
a
ti
on
ti
me
is
a
ls
o
inc
r
e
a
s
e
d
a
nd
s
ome
s
pe
c
if
ica
ti
ons
of
magnif
ied
f
r
a
mes
may
be
lo
s
t
dur
ing
im
a
ge
wa
r
ping.
F
ur
ther
mo
r
e
,
the
im
p
r
ove
a
ppr
oa
c
h
of
L
B
-
E
VM
a
ls
o
s
ti
l
l
f
a
il
e
d
in
lar
ge
magnif
ica
ti
o
n
f
a
c
tor
.
I
n
[
16
]
,
the
a
uthor
s
ha
ve
p
r
opos
e
d
a
n
i
mpr
ove
d
method
of
pha
s
e
-
ba
s
e
d
E
VM
(
PB
-
E
VM
)
,
whic
h
is
c
a
ll
e
d
f
a
s
t
pha
s
e
-
b
a
s
e
d
E
uler
ian
video
magnif
ica
ti
on
(
F
P
B
-
E
VM
)
.
I
n
thi
s
method
,
a
n
im
a
ge
pyr
a
mi
d
de
c
ompos
it
ion
that
e
mpl
oys
R
ies
z
-
t
r
a
ns
f
or
m
(
R
-
T
)
ha
s
be
e
n
us
e
d
in
the
s
tage
of
mul
ti
-
s
c
a
le
s
pa
ti
a
l
de
c
ompos
it
ion.
A
video
f
r
a
mes
a
r
e
f
i
r
s
t
de
c
ompos
e
d
us
ing
R
ies
z
pyr
a
mi
ds
that
us
e
s
f
ini
te
dif
f
e
r
e
nc
e
f
il
ter
s
of
two
thr
e
e
-
tap.
S
ubs
e
que
ntl
y,
the
de
c
ompos
e
d
f
r
a
mes
a
r
e
f
il
ter
e
d
by
inver
ti
ble
oc
tave
b
a
ndwidt
h
f
il
ter
s
.
T
he
a
uthor
s
in
[
16]
c
laim
that
thei
r
a
ppr
oa
c
h
c
a
n
be
us
ing
f
or
r
e
a
l
-
ti
me
hidden
mot
ion
vis
ua
li
z
a
ti
on.
Although
thi
s
a
ppr
oa
c
h
is
f
a
s
ter
than
c
omp
lex
s
tee
r
a
ble
in
c
onve
nti
ona
l
P
B
-
E
VM
,
the
main
dr
a
wba
c
k
is
t
he
R
ies
z
pyr
a
mi
d
f
a
il
s
to
maintain
the
powe
r
o
f
t
he
input
s
i
gna
l,
whic
h
c
a
n
c
a
us
e
low
qua
li
ty
of
the
magni
f
ied
video.
I
n
[
17]
,
a
uthor
s
ha
ve
pr
opos
e
d
a
method
to
s
olve
pr
oblems
of
int
e
ns
ive
pr
oc
e
s
s
ing
ti
me,
a
nd
low
video
q
ua
li
ty
in
E
VM
.
T
he
method
is
r
e
lyi
ng
wa
ve
let
de
c
ompos
it
ion,
a
nd
im
a
ge
de
-
nois
ing.
T
he
a
uthor
s
ha
ve
c
laimed
that
their
method
is
s
upe
r
ior
ove
r
pr
e
viou
s
methods
in
ter
ms
of
p
r
oc
e
s
s
ing
ti
me
r
e
duc
ti
on
a
nd
nois
e
e
li
mi
na
ti
on.
How
e
ve
r
,
a
ll
the
tes
ts
that
a
r
e
li
s
t
e
d
in
[
17]
ha
ve
a
c
hieve
d
f
o
r
s
mall
magnif
ica
ti
on
f
a
c
tor
s
.
T
his
is
be
c
a
us
e
wa
ve
let
s
e
pa
r
a
te
both
nois
e
a
nd
s
ubtl
e
mot
ions
in
de
tails
ba
nds
that
ha
ve
s
mall
e
ne
r
gy
s
o
that
by
a
mpl
if
ying
thes
e
ba
nds
nois
e
a
l
s
o
magnif
ied
by
the
s
a
me
r
a
ti
o.
Ac
c
or
ding
to
li
ter
a
tur
e
,
P
B
-
E
V
M
is
be
s
t
a
ppr
oa
c
h
in
ter
ms
of
nois
e
pe
r
f
or
manc
e
s
a
nd
lar
ge
magnif
ica
ti
on
f
a
c
tor
[
18]
.
T
he
main
pr
oblem
is
the
long
pr
oc
e
s
s
ing
ti
me.
T
his
pa
pe
r
pr
opos
e
s
a
n
e
nha
nc
e
ment
f
o
r
p
ha
s
e
-
ba
s
e
d
-
E
V
M
(
P
B
-
E
VM
)
in
or
de
r
to
s
igni
f
ica
ntl
y
r
e
duc
e
the
pr
oc
e
s
s
ing
ti
me
o
f
the
video
magnif
ic
a
ti
on
.
T
he
pr
opos
e
d
a
ppr
oa
c
h
r
e
duc
e
s
the
pr
oc
e
s
s
e
d
da
ta
by
r
e
s
izing
-
down
the
input
s
our
c
e
video
a
nd
then
r
e
s
izing
-
up
the
outpu
t
f
r
a
mes
of
P
B
-
E
VM
.
T
he
pr
opos
e
d
r
e
s
izing
uti
li
z
e
s
L
a
nc
z
os
-
3
a
lgor
it
hm
that
is
m
a
int
a
in
the
qua
li
ty
of
the
magnif
ied
video
a
s
it
is
in
the
c
onve
nti
ona
l
P
B
-
E
VM
[
19]
.
T
he
r
e
s
t
of
the
pa
p
e
r
is
or
ga
nize
d
a
s
:
s
e
c
ti
on
2
de
s
c
r
ibes
a
P
B
-
E
VM
a
ppr
oa
c
h.
S
e
c
ti
on
3
il
lus
tr
a
tes
the
L
a
nc
z
os
-
3
a
lgor
it
hm.
S
e
c
ti
on
4
e
xplains
the
p
r
opos
e
d
tec
hnique.
S
e
c
ti
o
n
5
a
nd
s
e
c
ti
on
6
e
xplains
the
s
im
ulation
r
e
s
ult
s
with
dis
c
us
s
ion
a
nd
c
ompar
is
on
with
the
r
e
late
d
wo
r
k
r
e
s
pe
c
ti
ve
ly.
F
inally,
c
onc
lus
ions
a
r
e
given
in
s
e
c
ti
on
7
.
2.
P
HAS
E
-
B
ASE
D
E
UL
E
RI
A
N
VI
DE
O
M
AGNI
F
I
CA
T
I
ON
T
he
P
B
-
E
VM
method
ha
s
be
e
n
de
ve
loped
by
W
a
dhwa
e
t
a
l.
[
10]
.
I
t
is
ba
s
e
d
on
c
ompl
e
x
s
tee
r
a
ble
tec
hnique
[
11
-
13,
20]
,
a
nd
pha
s
e
-
ba
s
e
d
opti
c
a
l
f
low
[
13]
.
T
he
method
a
tt
e
mpt
s
to
ove
r
c
ome
the
dr
a
wba
c
ks
of
LB
-
E
VM
s
uc
h
a
s
s
uppor
t
only
s
mall
magni
f
ica
ti
on
f
a
c
tor
s
a
nd
a
mpl
if
y
no
is
e
li
ne
a
r
ly
with
i
nc
r
e
a
s
ing
the
magnif
ica
ti
on
f
a
c
tor
.
T
he
main
dr
a
wba
c
k
in
thi
s
method
is
the
long
pr
oc
e
s
s
ing
ti
me.
T
he
P
B
-
E
VM
tec
hnique
s
uppor
ts
la
r
ge
magnif
i
c
a
ti
on
f
a
c
tor
s
with
s
igni
f
ica
ntl
y
les
s
nois
e
than
LB
-
E
VM
tec
hnique.
T
he
P
B
-
E
VM
method
doe
s
n
ot
incr
e
a
s
e
the
s
pa
ti
a
l
nois
e
be
c
a
us
e
the
method
m
a
gnif
ies
the
pha
s
e
ins
tea
d
of
a
mpl
it
ude
s
f
or
the
int
e
r
e
s
t
r
e
gions
.
As
a
r
e
s
ult
,
nois
e
is
tr
a
ns
late
d
r
a
ther
than
a
mpl
if
ied
whe
n
the
a
mpl
if
ica
ti
on
f
a
c
tor
is
incr
e
a
s
e
d.
T
h
e
r
e
f
or
e
,
th
is
method
incr
e
a
s
e
s
the
pha
s
e
va
r
iations
by
a
mul
ti
pli
c
a
ti
ve
f
a
c
to
r
to
a
mpl
if
y
the
s
ubtl
e
mot
ion
s
.
F
igur
e
1
s
hows
the
wor
king
mec
ha
nis
m
s
tage
s
of
the
P
B
-
E
VM
tec
hnique.
T
he
tec
hnique
s
tar
ts
by
de
c
ompos
ing
the
video
f
r
a
mes
int
o
mu
lt
ipl
e
s
c
a
les
a
nd
or
ienta
ti
on
f
or
dif
f
e
r
e
nt
s
pa
ti
a
l
-
f
r
e
que
nc
y
ba
nds
.
T
his
is
a
c
hieve
d
by
s
tee
r
a
ble
pyr
a
mi
ds
that
r
e
ly
on
F
our
ier
a
na
lys
is
to
a
na
lys
e
the
im
a
ge
int
o
s
ub
-
domain
s
a
nd
r
outi
ng.
T
he
ba
s
ic
f
unc
ti
ons
of
the
tr
a
ns
f
or
m
a
r
e
s
im
il
a
r
to
Ga
bor
wa
ve
lets
,
w
hich
a
r
e
s
inus
oids
windowe
d
by
a
Ga
us
s
ian
e
nve
lope
[
11
]
.
I
n
the
ne
xt
s
tage
,
a
tempor
a
l
f
i
lt
e
r
is
a
ppli
e
d
to
p
a
s
s
only
the
de
s
ir
e
d
f
r
e
que
nc
y
ba
nds
.
T
he
n,
the
e
ne
r
gy
of
the
de
s
ir
e
d
f
r
e
que
nc
y
ba
nds
is
incr
e
a
s
e
d
by
mul
ti
plyi
ng
f
il
ter
e
d
f
r
a
me
by
a
n
a
pp
r
opr
iate
magnif
ica
ti
on
f
a
c
tor
.
T
he
n
,
the
magni
f
ied
ba
nds
a
r
e
c
oll
e
c
t
e
d
with
the
unmagnif
ied
ba
nds
f
or
e
a
c
h
f
r
a
me.
F
inally
,
the
magnif
ied
video
f
r
a
me
is
r
e
c
ons
tr
uc
ted
by
c
ompl
e
x
s
tee
r
a
ble
pyr
a
mi
d
r
e
c
ons
tr
uc
ti
on.
T
he
s
tee
r
a
ble
py
r
a
mi
ds
a
r
e
c
a
lcula
ted
thr
ough
2D
F
our
ier
tr
a
ns
f
or
m
(
DFT
)
on
a
ll
f
r
a
mes
in
the
s
our
c
e
video
a
nd
s
ubs
e
que
nt
a
pplyi
ng
the
s
pa
ti
a
l
f
il
te
r
s
,
with
va
r
ying
s
ize
a
nd
dir
e
c
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
De
v
e
loped
appr
oac
h
for
phas
e
-
bas
e
d
E
uler
ian
v
ideo
magnifi
c
ati
on
(
Haide
r
I
s
mae
l
Shahadi
)
2393
T
his
is
r
e
s
ult
e
d
a
mul
ti
-
dim
e
ns
io
na
l
a
nd
mul
ti
-
dir
e
c
ti
ona
l
li
ne
a
r
a
na
lys
is
.
E
a
c
h
leve
l
of
the
pyr
a
mi
d
is
a
s
e
t
of
c
ompl
e
x
number
s
.
S
o
that
pha
s
e
a
nd
a
mpl
i
tude
a
r
e
e
a
s
il
y
c
a
lcula
ted
f
or
e
a
c
h
r
e
s
ult
e
d
c
ompl
e
x
pa
ir
.
F
igur
e
1.
Ge
ne
r
a
l
s
tr
uc
tur
e
of
pha
s
e
-
ba
s
e
d
E
VM
[
10]
3.
L
AN
CZ
OS
-
3
I
NT
E
RP
OL
AT
I
ON
A
L
GO
RI
T
HM
T
he
L
a
nc
z
os
-
3
a
lgor
it
hm
ha
s
be
e
n
pr
opos
e
d
by
C
or
ne
li
us
L
a
nc
z
os
-
3.
I
t
is
a
n
int
e
r
po
lation
method
us
e
d
in
im
a
ge
pr
oc
e
s
s
ing
s
uc
h
a
s
medic
a
l
im
a
ging
[
21
]
a
nd
video
r
e
s
izing
[
22]
.
T
he
L
a
nc
z
os
-
3
int
e
r
polation
pr
ovides
a
s
moot
h
ing
to
th
e
i
mage
mo
r
e
than
the
bi
-
c
ubic
a
lgor
it
hm.
T
he
int
e
r
polation
ke
r
ne
l
is
c
ons
ider
e
d
in
the
a
lgor
it
hm
,
whic
h
is
a
mul
ti
pli
c
a
ti
on
of
tw
o
S
I
NC
f
unc
ti
ons
[
23]
.
F
igu
r
e
2
il
lus
tr
a
tes
the
L
a
nc
z
os
-
3
ke
r
ne
l
f
unc
ti
on
that
us
e
d
in
im
a
ge
r
e
s
a
mpl
ing.
T
h
e
main
f
o
r
mul
a
of
the
r
e
s
izing
f
il
ter
a
ppli
e
s
a
S
inc
f
unc
ti
on
a
s
s
hown
in
(
1
).
F
igur
e
2.
T
he
f
unc
ti
on
of
L
a
nc
z
os
-
3
ke
r
ne
l
f
il
ter
(
)
=
{
1
=
0
a
s
i
n
(
)
s
i
n
(
)
2
2
0
<
|
|
<
0
ℎ
(
1
)
whe
r
e
a
is
a
n
int
e
ge
r
that
r
e
pr
e
s
e
nts
the
f
i
lt
e
r
s
i
z
e
.
A
typi
c
a
l
s
ize
of
the
ke
r
ne
l
f
il
ter
is
2
or
3.
C
ons
i
de
r
a
one
-
dim
e
ns
ion
ve
c
tor
whic
h
is
ha
ving
s
a
mpl
e
s
S
i
,
a
nd
let
S
(
x
)
be
the
int
e
r
polate
d
va
lue
a
t
a
r
a
ndom
a
r
gument
x
,
then
the
r
e
s
a
mpl
e
d
va
lue
of
L
a
nc
z
os
ke
r
ne
l
is
given
by
(
2)
.
(
)
=
∑
|
|
+
=
|
|
−
+
1
(
−
)
(
2)
I
n
the
2D
matr
ix
,
the
r
e
s
a
mpl
e
d
va
lues
of
L
a
nc
z
o
s
ke
r
ne
l
a
r
e
c
a
lcula
ted
by
the
pr
oduc
t
of
two
1D
Ke
r
ne
l
f
il
ter
s
a
s
s
hown
in
(
3
)
;
(
,
)
=
(
)
.
(
)
(
3)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
5
,
Oc
tober
2020:
2391
-
2400
2394
C
ons
ider
a
two
-
dim
e
n
s
ion
S
ij
whic
h
is
de
f
in
e
d
a
t
r
ows
a
nd
c
olum
n
indi
c
e
s
i
a
nd
j
r
e
s
pe
c
ti
ve
ly.
T
he
r
e
c
on
s
tr
uc
ted
or
int
e
r
polate
d
matr
ix
is
given
b
y
(
4)
;
(
,
)
=
∑
∑
(
−
)
.
(
−
)
|
|
+
=
|
|
−
+
1
|
|
+
=
|
|
−
+
1
(
4)
4.
P
ROP
OS
E
D
A
P
P
ROAC
H
T
he
pr
opos
e
d
a
ppr
oa
c
h
us
e
s
the
s
a
me
a
lgor
it
hm
of
t
he
c
onve
nti
ona
l
P
B
-
E
VM
.
How
e
ve
r
,
a
n
im
por
tant
pr
e
-
pr
oc
e
s
s
ing
a
nd
pos
t
-
pr
oc
e
s
s
ing
s
tag
e
s
a
r
e
a
dde
d
in
s
uc
h
a
wa
y
to
ove
r
c
ome
the
pr
oblem
of
i
ntens
ive
e
xe
c
uti
on
ti
me.
M
or
e
ove
r
,
the
qua
li
ty
o
f
the
magnif
ied
video
is
maintaine
d
wi
thout
e
f
f
e
c
ti
ng.
F
igur
e
3
s
hows
the
main
s
teps
of
the
pr
opos
e
d
a
ppr
oa
c
h.
F
igur
e
3.
T
he
block
diagr
a
m
of
the
pr
opos
e
d
a
ppr
o
a
c
h
T
he
s
teps
of
the
a
ppr
oa
c
h
a
r
e
a
s
f
o
ll
ows
:
−
T
he
video
f
il
e
is
r
e
a
d
a
s
AV
I
f
or
mat,
a
nd
then
the
f
r
a
mes
o
f
the
video
is
r
e
s
izing
-
down
by
50
%
of
the
or
igi
na
l
he
ight
a
nd
width
dim
e
ns
ion.
I
n
thi
s
s
tep,
we
e
mpl
oy
L
a
nc
z
os
r
e
s
a
mpl
ing
method
[
24]
.
T
he
r
e
s
ize
d
f
r
a
mes
’
s
ize
of
the
r
e
s
ult
e
d
video
is
25
%
of
the
s
our
c
e
video
.
−
T
he
ne
xt
s
tep
a
f
ter
video
r
e
s
izing
is
the
c
onve
r
ti
ng
of
a
ll
r
e
s
ult
e
d
video
f
r
a
mes
f
r
o
m
R
GB
s
pa
c
e
int
o
N
T
S
C
(
or
Y
I
Q)
s
pa
c
e
.
T
he
Y
c
omponent
r
e
pr
e
s
e
nt
s
the
il
lum
ination
inf
o
r
mation;
I
a
nd
Q
r
e
pr
e
s
e
nt
the
c
hr
omi
na
nc
e
inf
or
mation.
T
he
N
T
S
C
c
olor
s
ys
tem
is
int
e
nde
d
to
income
a
dva
ntage
of
human
r
e
s
pons
e
c
ha
r
a
c
ter
is
ti
c
s
to
the
c
olor
.
T
his
s
tep
is
done
by
a
p
plyi
ng
(
5)
on
a
ll
the
f
r
a
me
o
f
the
r
e
s
ize
d
video.
[
]
=
[
0
.
299
0
.
587
0
.
114
0
.
596
−
0
.
274
−
0
.
322
0
.
211
−
0
.
523
0
.
312
]
[
]
(
5)
−
T
he
thi
r
d
s
tep
in
the
a
ppr
oa
c
h
is
a
pplyi
ng
the
s
tee
r
a
ble
pyr
a
mi
d
de
c
ompos
it
ion
on
e
a
c
h
laye
r
(
Y,
I
,
a
nd
Q)
f
r
om
the
video
f
r
a
mes
indi
vidually.
T
he
de
c
ompos
it
ion
is
us
e
d
to
f
a
c
tor
ize
the
video
f
r
a
mes
int
o
s
c
a
lable
im
a
ge
s
f
or
dif
f
e
r
e
nt
leve
ls
o
f
de
c
ompos
it
io
n.
T
he
s
tee
r
a
ble
pyr
a
mi
d
is
a
tr
a
ns
f
o
r
m
that
a
na
lys
e
s
a
n
im
a
ge
ba
s
e
d
on
s
pa
ti
a
l
s
c
a
le,
or
ienta
ti
on
,
a
nd
pos
it
ion.
I
n
thi
s
p
r
oc
e
s
s
2
-
D
dis
c
r
e
te
F
our
ier
tr
a
ns
f
or
m
(
DF
T
)
is
a
ppli
e
d
ove
r
a
ll
the
r
e
s
ize
d
video
f
r
a
mes
in
YI
Q
s
pa
c
e
.
S
ubs
e
que
ntl
y,
the
s
pa
ti
a
l
f
il
t
e
r
s
of
dif
f
e
r
e
nt
s
ize
a
nd
or
ienta
ti
on
a
r
e
a
ppli
e
d.
T
he
s
tee
r
a
ble
pyr
a
mi
d
pr
oduc
e
s
a
li
ne
a
r
mul
ti
-
s
c
a
le
a
nd
mul
ti
-
or
ient
a
ti
on
de
c
ompos
e
d
im
a
ge
.
E
a
c
h
leve
l
f
r
om
the
ou
tput
is
a
r
r
a
y
of
c
ompl
e
x
number
s
,
whic
h
c
ons
is
ts
of
a
mpl
it
ude
a
nd
their
pha
s
e
f
o
r
e
a
c
h
e
lem
e
nt
in
the
a
r
r
a
y.
−
T
he
r
e
s
ult
e
d
ba
nds
f
r
om
the
pr
e
vious
s
tep
e
nter
e
d
t
o
tempor
a
l
f
il
te
r
.
I
n
or
de
r
to
pa
s
s
only
the
int
e
r
e
s
t
b
a
nds
of
f
r
e
que
nc
ies
to
a
mpl
if
y
them
in
the
ne
xt
s
tep.
−
I
n
thi
s
s
tep,
the
r
e
s
ult
e
d
ba
nd
c
oe
f
f
icie
nts
f
r
om
the
t
e
mpor
a
l
f
il
ter
a
r
e
mul
t
ipl
ied
by
the
magnif
ica
ti
on
f
a
c
tor
to
a
mpl
if
y
the
i
nter
e
s
ted
tempo
r
a
l
f
r
e
que
nc
ies
.
−
S
ubs
e
que
ntl
y,
the
magnif
y
ba
nds
c
ombi
ne
wit
h
the
or
ig
inal
ba
nds
that
r
e
s
ult
e
d
f
r
om
the
s
pa
ti
a
l
de
c
ompos
it
ion
in
s
tep
3.
−
I
nve
r
s
e
of
s
tee
r
a
ble
c
ompl
e
x
pyr
a
mi
d
is
a
ppli
e
d
on
t
he
r
e
s
ult
s
of
the
pr
e
vious
s
tep
to
r
e
c
ons
tr
uc
t
the
v
ideo
f
r
a
mes
a
f
ter
a
mpl
if
ica
ti
on
in
YI
Q
s
pa
c
e
.
−
S
ubs
e
que
ntl
y,
c
onve
r
t
f
r
a
mes
f
r
om
Y
I
Q
s
pa
c
e
i
nto
R
GB
s
pa
c
e
to
obtain
the
or
ig
inal
c
olor
of
v
ideo.
T
his
s
tep
is
done
by
a
pplyi
ng
(
6
)
on
a
ll
the
f
r
a
me
o
f
the
v
ideo
.
[
]
=
[
1
0
.
956
0
.
619
1
−
0
.
272
−
0
.
647
1
−
1
.
106
1
.
703
]
[
]
(
6)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
De
v
e
loped
appr
oac
h
for
phas
e
-
bas
e
d
E
uler
ian
v
ideo
magnifi
c
ati
on
(
Haide
r
I
s
mae
l
Shahadi
)
2395
−
F
inally,
the
output
video
f
r
om
the
p
r
e
vious
s
tep
is
r
e
s
izing
-
up
by
200
%
of
he
ight
a
nd
width
of
the
f
r
a
mes
to
ge
t
the
s
a
me
o
r
igi
na
l
s
ize
of
the
s
our
c
e
video
by
us
ing
L
a
nc
z
os
-
3
r
e
s
a
mpl
ing
method.
5.
RE
S
UL
T
S
AN
D
DI
S
CU
S
S
I
ON
T
his
s
e
c
ti
on
pr
e
s
e
nts
e
xpe
r
im
e
ntal
r
e
s
ult
s
to
c
ompar
e
the
modi
f
ied
P
B
-
E
VM
a
ppr
oa
c
h
with
the
c
onve
nti
ona
l
P
B
-
E
VM
a
ppr
oa
c
h.
F
igur
e
4
s
ho
ws
t
he
us
e
d
videos
in
our
tes
ts
,
whic
h
include
:
B
a
by1
video
with
dim
e
ns
ion
960×
544×
3×
301
a
nd
a
f
r
a
me
r
a
t
e
of
30
f
r
a
me
pe
r
s
e
c
ond(
f
ps
)
,
E
ye
video
with
di
mens
ion
1152×
896×
3×
140
a
nd
a
f
r
a
me
r
a
te
o
f
30
f
ps
,
B
a
by2
video
with
a
dim
e
ns
ion
640×
352×
3×
900
a
nd
a
f
r
a
me
r
a
te
of
30
f
ps
,
B
a
by3
video
with
a
d
im
e
ns
ion
1280×
7
20×
3
×
22
2
a
nd
a
f
r
a
me
r
a
te
of
30f
ps
.
All
the
v
ideos
e
xc
e
pt
B
a
by3
c
a
n
be
f
ound
on
the
we
bs
it
of
the
M
I
T
[
25]
.
T
he
B
a
by3
-
video
ha
s
be
e
n
r
e
c
or
de
d
by
the
P
r
e
ter
m
de
pa
r
tm
e
nt
-
c
e
ntr
a
l
c
hil
d
hos
pit
a
l
-
I
r
a
q
a
nd
the
ba
by
ha
s
c
a
s
ua
lt
y
vir
a
l
he
pa
ti
ti
s
.
I
n
o
r
de
r
to
c
ompar
e
the
pr
opos
e
d
a
ppr
oa
c
h
ove
r
the
c
onve
nti
ona
l
one
,
we
mea
s
ur
e
the
qua
li
ty
o
f
the
r
e
s
ult
e
d
videos
f
r
om
e
a
c
h
a
ppr
oa
c
h.
T
he
qua
li
ty
mea
s
ur
e
ment
is
a
c
hieve
d
by
mea
s
ur
ing
P
e
a
k
s
ignal
-
to
-
nois
e
r
a
ti
o
(
P
S
NR
)
a
c
c
or
din
g
to
(
7
)
f
or
e
a
c
h
f
r
a
me
;
=
10
log
255
2
(
7)
=
1
×
∑
∑
(
,
−
,
)
2
whe
r
e
,
M
S
E
is
the
mea
n
s
qua
r
e
e
r
r
or
,
I
a
nd
I
a
a
r
e
t
he
or
igi
na
l
a
nd
the
a
mpl
i
f
ied
f
r
a
mes
r
e
s
pe
c
ti
ve
ly,
M
a
nd
N
a
r
e
the
f
r
a
me
dim
e
ns
ions
.
I
n
a
ddit
ion
t
o
P
S
NR
,
m
a
xim
um
a
bs
olut
e
e
r
r
or
(
M
AX
E
R
R
)
is
a
ls
o
c
a
lcula
ted
in
the
tes
ts
that
r
e
pr
e
s
e
nts
the
a
bs
olut
e
maximum
s
qua
r
e
d
de
viati
on
of
the
input
to
the
output
video
f
r
a
mes
.
F
ur
th
e
r
mor
e
,
L
2R
AT
is
a
ls
o
us
e
d
to
mea
s
ur
e
the
r
a
ti
o
of
the
s
qu
a
r
e
d
nor
m
of
the
output
t
o
the
input
video
f
r
a
mes
.
W
e
ha
ve
im
pleme
nted
our
tes
ts
by
a
pplyi
ng
oc
tave
ba
ndwi
dth
of
c
ompl
e
x
s
tee
r
a
ble
py
r
a
mi
ds
that
ha
ve
wid
e
s
pa
ti
a
l
s
uppor
t.
S
e
ve
r
a
l
gr
oups
of
tes
ts
ha
ve
be
e
n
a
c
hieve
d
a
s
il
lus
tr
a
ted
in
F
igur
e
4
.
(
a
)
(
b)
(
c
)
(
d)
F
igur
e
4.
T
he
s
our
c
e
videos
that
us
e
d
in
the
e
xpe
r
i
menta
l
r
e
s
ult
s
;
(
a
)
B
a
by1
,
(
b
)
E
ye
,
(
c
)
B
a
by2
,
(
d
)
B
a
by3
a.
I
n
the
f
ir
s
t
gr
oup
of
tes
ts
,
B
a
by1
video
as
s
hown
in
F
igur
e
4
(
a
)
ha
s
be
e
n
us
e
d
a
s
a
s
our
c
e
video.
T
he
g
oa
l
of
thi
s
gr
oup
of
tes
ts
is
to
de
tec
t
the
we
a
k
br
e
a
thi
ng
of
the
ba
by
by
magni
f
ying
the
s
ubtl
e
mot
ion
of
the
r
e
gion
a
r
ound
the
c
he
s
t
a
nd
a
bdom
inal
a
r
e
a
.
I
nf
ini
te
im
puls
e
r
e
s
pons
e
(
I
I
R
)
ha
s
be
e
n
us
e
d
a
s
temp
or
a
l
f
il
ter
with
thr
e
e
s
e
ts
of
dif
f
e
r
e
nt
tes
ts
a
s
f
oll
ows
:
−
S
e
t1:
in
thi
s
s
e
t
of
tes
ts
,
the
va
lue
of
s
igm
a
is
c
h
a
nge
d
a
s
{1,
3,
5,
7,
a
nd
10}.
W
hil
e
,
the
f
r
e
que
n
c
y
bounda
r
ies
of
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
a
r
e
0
.
1
Hz
f
or
the
lowe
s
t
f
r
e
que
nc
y
a
nd
0
.
4
Hz
f
or
the
high
e
s
t
f
r
e
que
nc
y.
T
ha
t
mea
n
the
br
e
a
thi
ng
ha
ppe
ns
6
-
24
ti
mes
pe
r
mi
nute
.
T
he
magnif
ica
ti
on
f
a
c
tor
is
f
ix
e
d
with
va
lue
e
qua
ls
to
20
.
T
a
ble
1
s
hows
the
e
xp
e
r
im
e
ntal
r
e
s
ult
s
that
il
lus
tr
a
te
c
ompar
is
on
be
tw
e
e
n
c
onve
nti
ona
l
a
nd
pr
opos
e
d
a
ppr
oa
c
he
s
.
As
s
how
n
in
the
table
,
the
incr
e
a
s
ing
of
s
igm
a
va
lue
lea
d
s
to
incr
e
a
s
e
the
qua
li
ty
o
f
magni
f
ied
video
,
howe
ve
r
,
the
e
xe
c
uti
on
ti
me
incr
e
a
s
e
s
too.
T
he
table
s
ho
ws
the
s
upe
r
ior
it
y
of
the
pr
opos
e
d
ove
r
the
c
onve
nti
ona
l
in
r
e
duc
ing
the
pr
oc
e
s
s
ing
ti
me
upto
60%
with
ma
int
a
ini
ng
the
video
qua
l
it
y.
−
S
e
t2:
I
n
thi
s
s
e
t
of
tes
ts
,
α
va
r
ies
with
va
lues
{30
,
40,
50
,
60
,
a
nd
100}
a
nd
f
ixed
va
lue
o
f
s
igm
a
(
5)
.
T
he
f
r
e
que
nc
y
bounda
r
ies
of
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
a
r
e
{0.
1
-
0.
4Hz
}.
T
a
ble
2
s
hows
the
c
ompar
is
on
be
twe
e
n
the
pr
op
os
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
.
T
he
r
e
s
ult
s
of
the
table
s
how
the
s
upe
r
ior
it
y
of
the
pr
opos
e
d
ove
r
the
c
onve
nti
ona
l
method
in
r
e
duc
ing
the
e
xe
c
uti
on
ti
me
by
mor
e
than
60
%
a
nd
ke
e
ping
the
v
ideo
qua
li
ty
wi
thi
n
a
c
onve
nti
ona
l
P
B
-
E
VM
li
mi
t.
F
igu
r
e
5
s
hows
s
a
mpl
e
s
o
f
the
a
mpl
i
f
i
e
d
f
r
a
mes
with
r
e
s
pe
c
t
to
the
s
our
c
e
f
r
a
mes
f
or
both
th
e
pr
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
,
whe
r
e
α
=
100
a
nd
s
igm
a
=
5.
F
r
om
F
igur
e
5
we
s
e
e
that
the
a
mpl
if
ied
f
r
a
mes
us
ing
the
pr
opos
e
d
method
ha
ve
qua
l
it
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
5
,
Oc
tober
2020:
2391
-
2400
2396
r
e
s
e
mbl
e
to
the
a
mpl
if
ied
f
r
a
mes
us
ing
the
c
onve
nti
ona
l
method
with
e
xe
c
uti
on
ti
me
les
s
than
ha
lf
of
that
r
e
quir
e
d
in
the
c
onve
nti
ona
l
a
ppr
oa
c
h.
F
ig
ur
e
6
s
how
c
ompar
is
on
in
P
S
NR
va
lues
be
twe
e
n
the
c
onve
nti
ona
l
a
nd
pr
opos
e
d
a
ppr
oa
c
h
f
or
B
a
by
1
video
f
r
a
mes
.
F
ur
ther
mor
e
,
the
s
ur
vival
o
f
the
vi
de
o
qua
li
ty
of
the
pr
opos
e
d
a
ppr
oa
c
h
r
e
mains
c
los
e
with
the
c
onve
nti
ona
l
method
.
T
a
ble
1
.
S
e
t1
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
B
a
by1
a
t
α
=
20
S
ig
ma
va
lu
e
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
1
218.27
32.15
39.65
130.30
0.9993
82.72
32.25
40.36
124.79
0.9989
3
219.67
32.51
36.45
128.82
0.9994
78.92
32.49
36.79
123.96
0.9992
5
225.23
32.79
34.22
125.26
0.9994
85.36
32.74
34.72
121.95
0.9992
7
223.49
32.99
32.65
123.55
0.9994
84.18
32.99
33.37
120.10
0.9992
10
230.94
33.21
31.04
120.94
0.9995
85.44
33.09
32.06
118.08
0.9993
T
a
ble
2
.
S
e
t2
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
B
a
by1
a
t
s
igm
a
=
5
M
a
gni
f
y
f
a
c
to
r
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
α =
30
225.78
31.42
46.89
137.78
0.9993
85.79
31.49
46.24
125.32
0.9989
α =
40
226.86
30.08
63.81
143.69
0.9989
86.20
30.18
62.72
128.48
0.9986
α =
50
228.22
28.91
83.52
144.84
0.9985
86.72
28.97
82.95
133.61
0.9983
α =
60
227.28
27.99
103.35
143.48
0.9980
86.36
27.99
103.58
138.11
0.9979
α =
100
229.44
24.07
282.43
155.17
0.9958
87.18
24.07
282.43
155.17
0.9958
(
a
)
(
b)
(
c
)
F
igur
e
5.
S
a
mpl
e
s
of
the
e
xpe
r
im
e
ntal
r
e
s
ult
s
f
or
B
a
by1
a
t
α
=
100
a
nd
s
igm
a
=
5
(
a
)
s
our
c
e
f
r
a
mes
,
(
b)
a
mpl
i
f
ied
f
r
a
mes
us
ing
c
onve
nti
ona
l
P
B
-
E
VM
,
(
c
)
a
mpl
if
ied
f
r
a
mes
us
ing
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
−
S
e
t3:
I
n
thi
s
s
e
t
o
f
tes
ts
,
the
bounda
r
y
f
r
e
que
nc
ies
of
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
ha
ve
be
e
n
c
ha
n
ge
d
with
va
lues
{0.
1
-
0.
4
,
0
.
15
-
0.
4,
0.
2
-
0.
38}
Hz
.
W
hi
le
us
e
f
ixed
va
lue
f
or
s
igm
a
5,
a
nd
the
magni
f
ica
ti
on
f
a
c
tor
α
is
20.
T
a
ble
3
s
hows
the
e
xpe
r
im
e
ntal
r
e
s
ult
s
f
or
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
h
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
De
v
e
loped
appr
oac
h
for
phas
e
-
bas
e
d
E
uler
ian
v
ideo
magnifi
c
ati
on
(
Haide
r
I
s
mae
l
Shahadi
)
2397
F
r
om
the
r
e
s
ult
s
of
the
table
we
c
a
n
s
e
e
that
the
bounda
r
y
f
r
e
que
nc
ies
{0.
2
-
0.
38
Hz
}
give
higher
qua
l
it
y
than
the
othe
r
f
r
e
que
nc
y
li
mi
ts
.
Als
o,
i
t
is
c
lea
r
that
the
p
r
opos
e
d
method
is
s
upe
r
ior
ove
r
the
c
onve
nti
o
na
l
method
in
r
e
duc
ing
the
im
p
leme
ntation
ti
me
b
y
mor
e
than
61%
with
maintaining
the
magni
f
ied
video
qua
li
ty.
T
a
ble
3
.
S
e
t3
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
B
a
by1
a
t
α
=
20
a
nd
s
igm
a
=
5
B
a
nd
-
pa
s
s
(
H
z
)
va
lu
e
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
0.09
-
0.4
223.15
35.65
17.71
111.37
0.9998
84.79
35.95
16.52
98.01
0.9998
0.15
-
0.4
220.06
38.09
10.09
95.72
0.9998
81.42
38.59
8.99
91.53
0.9998
0.2
-
0.38
204.64
40.38
5.95
72.37
0.9999
77.76
40.86
5.33
70.12
0.9999
b.
I
n
the
s
e
c
ond
gr
oup
of
tes
ts
,
E
ye
video
a
s
s
hown
in
F
igur
e
4
(
b
)
is
us
e
d
a
s
a
s
our
c
e
video.
T
he
goa
l
he
r
e
is
to
magnif
y
the
s
ubtl
e
mot
ion
ha
ppe
ns
in
the
e
y
e
mus
c
le
a
nd
it
s
c
a
pil
lar
ies
.
F
ini
te
im
puls
e
r
e
s
po
ns
e
(
F
I
R
)
ha
s
be
e
n
us
e
d
a
s
tempor
a
l
f
il
ter
in
thr
e
e
di
f
f
e
r
e
nt
s
e
ts
of
tes
ts
that
a
r
e
e
xplaine
d
be
low.
−
S
e
t1:
S
igm
a
is
c
ha
nge
d
with
va
lues
{1,
3
,
4
,
7,
1
0}.
W
hil
e
,
the
bounda
r
y
f
r
e
que
nc
ies
of
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
va
lues
a
r
e
{30
-
50
Hz
}
with
α
=
75.
T
a
ble
4
s
hows
the
e
xp
e
r
im
e
ntal
r
e
s
ult
s
f
or
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
methods
.
As
s
hown
in
the
tabl
e
,
the
incr
e
a
s
ing
of
s
igm
a
lea
ds
to
incr
e
a
s
ing
the
qua
li
ty
of
magnif
ied
video
a
nd
e
xe
c
uti
on
ti
me.
Als
o,
we
c
a
n
c
lea
r
ly
s
e
e
the
s
upe
r
ior
it
y
of
the
pr
opos
e
d
meth
od
ove
r
the
c
onve
nti
ona
l
method
in
r
e
duc
ing
the
e
xe
c
uti
on
ti
me
in
a
r
ound
50%
with
maintaining
the
vid
e
o
qua
l
it
y.
F
igur
e
7
s
how
c
ompar
is
on
in
P
S
NR
va
lues
be
twe
e
n
the
c
onve
nti
ona
l
a
nd
pr
opos
e
d
a
ppr
oa
c
h
f
o
r
the
magnif
ied
e
ye
f
r
a
mes
.
F
r
om
th
is
f
ig
ur
e
it
is
c
lea
r
that
the
video
f
r
a
me
qua
li
ty
o
f
the
pr
opos
e
d
a
ppr
o
a
c
h
r
e
mains
c
los
e
with
the
c
onve
nti
ona
l
method
.
T
a
bl
e
4
.
S
e
t1
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
E
ye
a
t
α
=
75
S
ig
ma
va
lu
e
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
1
205.58
29.52
72.65
200.78
0.9899
103.60
29.55
72.54
199.42
0.9899
3
205.79
29.69
69.72
201.34
0.9915
106.92
29.77
68.99
198.19
0.9913
4
211.98
29.79
68.25
200.73
0.9917
102.80
29.88
67.12
195.66
0.9915
7
245.46
29.98
65.29
197.48
0.9919
120.64
30.09
63.69
194.33
0.9918
10
251.95
30.11
63.38
195.35
0.9920
121.74
30.24
61.89
192.75
0.9918
F
igur
e
6.
C
ompar
is
on
be
twe
e
n
the
p
r
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
f
or
B
a
by1
video
a
t
α
=
20
F
igur
e
7.
C
ompar
is
on
be
twe
e
n
the
p
r
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
f
or
E
ye
video
a
t
α
=
75
−
S
e
t2:
the
magnif
ica
ti
on
f
a
c
tor
α
is
va
r
ied
wit
h
va
lues
{60,
85,
90,
100}
a
nd
a
f
ixed
va
lue
of
the
s
igm
a
=
4.
T
he
f
r
e
que
nc
y
bounda
r
ies
of
th
e
b
a
nd
-
pa
s
s
tempor
a
l
f
il
ter
a
r
e
{30
-
50
Hz
}
f
o
r
a
ll
va
l
ue
s
of
α
.
T
a
ble
5
s
hows
the
e
xpe
r
im
e
ntal
r
e
s
ult
s
of
P
B
-
E
VM
method
f
or
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
5
,
Oc
tober
2020:
2391
-
2400
2398
T
hr
ough
the
r
e
s
ult
s
of
the
table
we
c
lea
r
ly
s
e
e
that
s
upe
r
io
r
it
y
of
the
pr
opos
e
d
method
o
ve
r
the
c
o
nve
nti
ona
l
method
in
r
e
duc
ing
the
pr
oc
e
s
s
ing
ti
me
by
a
r
ound
50%
,
while
the
video
qua
li
ty
ke
e
ps
withi
n
c
onve
nti
ona
l
P
B
-
E
VM
li
mi
ts
.
T
a
ble
5
.
S
e
t2
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
E
ye
a
t
S
igm
a
=
4
M
a
gni
f
y
f
a
c
to
r
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
α =
60
211.73
30.58
56.94
199.23
0.9929
108.89
30.69
55.79
199.89
0.9926
α =
85
210.41
29.35
75.56
209.30
0.9909
106.36
29.47
74.49
204.15
0.9907
α =
90
212.75
29.15
79.09
210.14
0.9907
105.65
29.25
78.05
205.50
0.9904
α =
100
211.15
28.79
85.93
213.21
0.9898
110.05
28.86
84.90
209.63
0.9896
−
S
e
t3:
the
bounda
r
y
f
r
e
que
nc
ies
f
or
the
ba
nd
-
pa
s
s
te
mpor
a
l
f
il
ter
is
us
e
with
va
r
ious
va
lues
{30
-
35,
40
-
50,
50
-
60}
Hz
.
W
hil
e
us
e
f
ixed
va
lue
f
o
r
S
igm
a
=
4
a
nd
α
=
75.
T
a
ble
6
s
hows
the
e
xpe
r
im
e
ntal
r
e
s
ult
s
of
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
by
us
ing
E
ye
video
a
s
input
.
Aga
in
,
the
table
s
hows
that
the
pr
opos
e
d
a
ppr
oa
c
h
s
upe
r
ior
in
ter
ms
o
f
t
im
e
r
e
duc
ti
on
ove
r
the
c
onve
nti
ona
l,
whe
r
e
the
r
e
quir
e
d
pr
oc
e
s
s
ing
ti
me
f
or
the
pr
opos
e
d
is
a
bout
ha
l
f
of
th
a
t
in
the
c
onve
nti
ona
l
one
without
e
f
f
e
c
t
the
qua
li
t
y
of
the
magnif
ied
video
.
T
a
ble
6
.
S
e
t3
r
e
s
ult
s
f
or
the
pr
opos
e
d
ba
s
e
d
P
B
-
E
VM
f
or
E
ye
a
t
S
igm
a
=
4
a
nd
α
=
75
B
a
nd
-
pa
s
s
(
H
z
)
va
lu
e
C
onve
nt
io
na
l
P
B
-
E
V
M
[
10]
P
r
opos
e
d i
mpr
ove
d P
B
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
E
xe
c
ut
io
n
T
im
e
P
S
N
R
M
S
E
M
A
X
E
E
R
L
2R
A
T
30
-
35
208.39
29.43
74.17
199.83
0.9924
106.67
29.49
73.17
192.82
0.9921
40
-
50
208.22
30.59
56.66
194.15
0.9937
108.84
30.67
94.60
188.07
0.9923
50
-
60
214.06
31.99
41.07
175.20
0.9946
107.36
32.04
76.25
167.60
0.9942
c.
I
n
the
thi
r
d
gr
oup
of
tes
ts
,
B
a
by2
video
a
s
s
hown
i
n
F
igur
e
4
(
c
)
is
us
e
:
T
he
goa
l
he
r
e
is
to
magnif
y
he
a
d
moveme
nt
r
e
s
ult
ing
f
r
om
the
c
hil
d's
br
e
a
thi
ng
.
I
I
R
ha
s
be
e
n
us
e
d
a
s
tempor
a
l
f
il
ter
,
a
nd
the
magni
f
ica
ti
on
f
a
c
tor
va
r
ies
with
va
lues
{20
,
40,
60
,
80,
100
a
n
d
150}
a
nd
f
ixed
va
lue
of
s
igm
a
(
5)
.
T
he
f
r
e
que
nc
y
bounda
r
ies
f
or
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
a
r
e
{
0.
3
-
0.
5
Hz
}.
T
a
ble
7
s
hows
the
c
ompar
is
on
be
tw
e
e
n
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
a
ppr
oa
c
he
s
.
T
he
t
im
e
r
e
duc
ti
on
in
the
pr
opos
e
d
method
is
a
bout
5
4%
c
ompar
e
to
the
c
onve
nti
ona
l
one
with
maintaining
the
magnif
ied
video
qua
li
ty.
T
a
ble
7
.
C
ompar
is
ons
be
twe
e
n
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
P
B
-
E
VM
f
or
B
a
by2
video
M
a
gni
f
y f
a
c
to
r
α =
20
α =
40
α =
60
α =
80
α =
100
α =
150
C
onve
nt
io
na
l
PB
-
E
V
M
[
10]
E
xe
c
ut
io
n T
im
e
225.69
226.39
227.92
228.35
228.82
229.95
P
S
N
R
35.77
33.17
31.35
30.05
29.06
27.38
M
S
E
18.56
37.22
57.96
77.62
95.62
135.13
M
A
X
E
E
R
63.50
88.50
107.74
119.90
128.85
146.40
L
2R
A
T
0.9982
0.9978
0.9974
0.9970
0.9966
0.9957
P
r
opos
e
d
im
pr
ove
d
PB
-
E
V
M
E
xe
c
ut
io
n T
im
e
103.81
104.13
104.84
105.03
105.25
105.77
P
S
N
R
34.04
32.22
30.39
29.66
28.03
26.84
M
S
E
25.64
39.01
59.44
70.32
110.91
146.22
M
A
X
E
E
R
72.42
93.93
111.40
122.21
128.87
138.32
L
2R
A
T
0.9979
0.9975
0.9970
0.9966
0.9957
0.9948
d.
B
a
by3
video
a
s
s
ho
wn
in
F
igur
e
4
(
d)
ha
s
be
e
n
us
e
d
in
the
f
our
th
gr
oup
o
f
tes
ts
.
T
he
ba
by
in
thi
s
vi
de
o
ha
s
vir
a
l
he
pa
ti
ti
s
a
nd
ha
s
dif
f
iculty
br
e
a
thi
ng.
T
he
r
e
f
or
e
,
the
goa
l
o
f
magnif
ica
ti
on
is
to
moni
tor
his
br
e
a
thi
ng.
I
I
R
is
us
e
d
a
s
a
tempor
a
l
f
il
ter
,
α
wit
h
va
r
ious
va
lues
{30,
60,
80,
100,
a
nd
150}
.
W
hil
e
,
the
va
lues
of
bounda
r
y
f
r
e
que
nc
ies
f
or
the
ba
nd
-
pa
s
s
tempor
a
l
f
il
ter
a
r
e
{0
.
2
–
0.
4
Hz
}
a
nd
f
ixed
va
lue
o
f
s
igm
a
is
us
e
d,
S
igm
a
=
5
.
T
a
ble
8
s
hows
the
e
x
pe
r
im
e
ntal
r
e
s
ult
s
f
or
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
methods
.
T
he
ti
me
r
e
duc
ti
on
in
the
pr
opos
e
d
one
is
a
r
ound
70%
c
ompar
e
to
the
c
onve
nti
ona
l
one
with
maintaining
the
magnif
ied
video
qua
li
ty
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
De
v
e
loped
appr
oac
h
for
phas
e
-
bas
e
d
E
uler
ian
v
ideo
magnifi
c
ati
on
(
Haide
r
I
s
mae
l
Shahadi
)
2399
T
a
ble
8
.
C
ompar
is
ons
be
twe
e
n
the
pr
opos
e
d
a
nd
c
onve
nti
ona
l
P
B
-
E
VM
f
or
B
a
by3
vi
de
o
M
a
gni
f
y f
a
c
to
r
α =
30
α =
60
α =
80
α =
100
α =
150
C
onve
nt
io
na
l
PB
-
E
V
M
[
10]
E
xe
c
ut
io
n T
im
e
231.37
232.52
232
.73
247.89
245.54
P
S
N
R
32.88
29.29
28.06
27.19
25.73
M
S
E
37.86
85.78
112.69
136.44
188.87
M
A
X
E
E
R
135.28
164.80
175.65
182.16
189.21
L
2R
A
T
0.9994
0.9983
0.9975
0.9967
0.9946
P
r
opos
e
d
im
pr
ove
d
PB
-
E
V
M
E
xe
c
ut
io
n T
im
e
71.72
72.08
72.14
74.36
73.66
P
S
N
R
32.74
29.22
27.99
27.13
25.69
M
S
E
38.56
86.60
113.56
137.28
189.66
M
A
X
E
E
R
129.74
161.92
173.01
180.15
189
L
2R
A
T
0.9993
0.9983
0.9975
0.9967
0.9946
6.
COM
P
AR
I
S
ON
WI
T
H
T
HE
RE
L
AT
E
D
WORK
T
his
s
e
c
ti
on
c
ompar
e
s
be
twe
e
n
the
pr
opos
e
d
m
e
thod
a
nd
the
r
e
late
d
methods
in
the
li
ter
a
tur
e
.
T
a
ble
9
s
hows
the
e
xe
c
uti
on
ti
me
a
nd
video
qua
li
t
y
in
ter
ms
of
P
S
NR
f
or
di
f
f
e
r
e
nt
methods
.
I
t
is
e
a
s
y
to
s
e
e
the
s
igni
f
ica
nt
r
e
duc
ti
on
in
pr
oc
e
s
s
ing
ti
me
of
the
p
r
opos
e
d
method
ove
r
a
ll
othe
r
s
.
F
ur
ther
mo
r
e
,
it
is
maintaining
video
qua
li
ty
s
a
me
a
s
in
c
onve
nti
ona
l
P
B
-
E
VM
method
that
ha
ve
the
s
upe
r
ior
it
y
in
te
r
ms
of
video
qua
li
ty
c
ompar
e
to
the
other
method
in
li
te
r
a
tur
e
.
I
n
a
ll
the
tes
t
of
T
a
ble
9,
B
a
by1
ha
s
be
e
n
us
e
d
a
s
a
video
s
our
c
e
f
il
e
.
F
igur
e
8
s
hows
a
n
e
xa
mpl
e
of
the
magnif
ied
video
f
r
a
me
us
ing
dif
f
e
r
e
nt
magnif
ica
ti
on
tec
hniques
a
t
α
=
20.
T
he
f
igu
r
e
s
hows
the
dif
f
e
r
e
nc
e
in
f
r
a
me
qua
li
ty
vis
ua
ll
y.
T
a
ble
9
.
C
ompar
is
on
of
the
pr
opos
e
d
a
nd
other
r
e
l
e
va
nt
video
magnif
ica
ti
on
methods
I
nput
V
id
e
os
M
e
th
ods
LB
-
E
V
M
[
8]
E
nha
nc
e
E
V
M
[
15]
C
onve
nt
io
na
l
PB
-
E
V
M
[
10]
FPB
-
E
V
M
[
16]
P
r
opos
e
d
PB
-
E
V
M
E
xe
c
ut
io
n
T
im
e
P
S
N
R
E
xe
c
ut
io
n
T
im
e
P
S
N
R
E
xe
c
ut
io
n
T
im
e
P
S
N
R
E
xe
c
ut
io
n
T
im
e
P
S
N
R
E
xe
c
ut
io
n
T
im
e
P
S
N
R
B
a
by1
85.67
31.75
101.32
34.91
225.23
32.79
117.45
31.91
85.36
32.74
(
a
)
(
b)
(
c
)
(
d)
(
e
)
(f)
F
igur
e
8.
M
a
gnif
ied
videos
us
ing
di
f
f
e
r
e
nt
magnif
i
c
a
ti
on
tec
hniques
;
(
a
)
s
our
c
e
f
r
a
mes
,
(
b)
a
mpl
i
f
ied
f
r
a
mes
ba
s
e
d
on
LB
-
E
VM
a
ppr
oa
c
h,
(
c
)
a
mpl
if
ied
f
r
a
mes
ba
s
e
d
on
E
2VM
a
pp
r
oa
c
h,
(
d)
Ampli
f
ied
f
r
a
mes
ba
s
e
d
on
c
onve
nti
ona
l
P
B
-
E
VM
a
ppr
oa
c
h,
(
e
)
a
mpl
if
ied
f
r
a
mes
ba
s
e
d
on
F
P
B
-
E
VM
a
ppr
oa
c
h,
a
nd
(
f
)
a
mpl
if
ied
f
r
a
mes
ba
s
e
d
on
pr
opos
e
d
P
B
-
E
VM
a
ppr
oa
c
h
7.
CONC
L
USI
ONS
T
his
pa
pe
r
pr
e
s
e
nts
a
n
e
f
f
e
c
ti
ve
modi
f
ica
ti
on
f
or
P
B
-
E
VM
method
in
or
de
r
to
r
e
duc
e
the
r
e
quir
e
d
pr
oc
e
s
s
ing
ti
me.
T
he
p
r
opos
e
d
tec
hnique
us
e
s
the
L
a
nc
z
o
s
a
lgor
it
hm
that
ba
s
e
d
on
ke
r
ne
l
f
il
ter
to
r
e
s
ize
-
down
da
ta
of
the
s
our
c
e
video
be
f
or
e
pr
oc
e
s
s
ing
a
n
d
r
e
s
ize
-
up
a
f
ter
pr
oc
e
s
s
ing.
T
his
a
lgor
it
hm
m
a
int
a
ins
the
qua
li
ty
o
f
the
a
mpl
if
ied
video
withi
n
P
B
-
E
VM
qua
li
ty
li
mi
ts
.
E
xpe
r
im
e
ntal
r
e
s
ult
s
s
how
that
the
r
e
quir
e
d
pr
oc
e
s
s
ing
ti
me
is
r
e
duc
e
d
by
mor
e
than
ha
lf
.
M
or
e
ove
r
,
the
c
ompar
is
on
with
the
r
e
late
d
meth
od
s
hows
the
s
upe
r
ior
it
y
o
f
the
p
r
opos
e
d
ove
r
the
other
s
in
ter
ms
of
r
e
duc
ti
on
both
p
r
oc
e
s
s
ing
ti
me
a
nd
magnif
i
e
d
nois
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
5
,
Oc
tober
2020:
2391
-
2400
2400
T
he
p
r
opos
e
d
tec
hnique
s
uppor
ts
lar
ge
magni
f
ic
a
ti
on
f
a
c
tor
wi
th
mi
nim
um
nois
e
a
mpl
if
ica
ti
on
r
e
s
e
mbl
e
to
P
B
-
E
VM
.
T
he
r
e
f
or
e
,
it
c
a
n
be
us
e
d
e
a
s
il
y
to
de
tec
t
invi
s
ibl
e
c
ha
nge
s
in
the
videos
to
be
c
lea
r
ly
pe
r
c
e
pti
ble
to
a
nick
e
ye
.
S
o
that
s
e
ve
r
a
l
a
ppli
c
a
ti
ons
c
a
n
e
m
ploy
our
tec
hnique
s
uc
h
a
s
in
he
a
lt
h
c
a
r
e
,
e
nginee
r
ing
a
nd
biol
ogica
l
mi
c
r
os
c
opy.
RE
F
E
RE
NC
E
S
[1
]
Po
h
M
.
Z
.
,
McD
u
ff
D
.
J
.
,
Pi
card
R
.
W.
,
"
Non
-
c
o
n
t
act
,
au
t
o
m
at
e
d
card
i
ac
p
u
l
s
e
meas
u
remen
t
s
u
s
i
n
g
v
i
d
eo
i
ma
g
i
n
g
an
d
b
l
i
n
d
s
o
u
rce
s
ep
ara
t
i
o
n
,"
O
p
t
i
cs
E
x
p
r
e
s
s
,
v
o
l
.
1
8
,
n
o
.
1
0
,
p
p
.
1
0
7
6
2
-
1
0
7
7
4
,
2
0
1
0
.
[2
]
Po
h
M
.
Z
.
,
McD
u
ff
D
.
J
.
,
Pi
card
R
.
W.
,
"
A
d
v
a
n
cemen
t
s
i
n
N
o
n
co
n
t
ac
t
,
Mu
l
t
i
p
arame
t
er
Ph
y
s
i
o
l
o
g
i
cal
Mea
s
u
reme
n
t
s
U
s
i
n
g
a
W
eb
cam
,"
IE
E
E
T
ra
n
s
ac
t
i
o
n
s
o
n
B
i
o
me
d
i
ca
l
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
58
,
n
o
.
1
,
p
p
.
7
-
11
,
2
0
1
1
.
[3
]
V
erk
r
u
y
s
s
e
W
.
,
Sv
aas
an
d
L
.
O
.
,
N
el
s
o
n
J
.
S.
,
"
Remo
t
e
p
l
e
t
h
y
s
m
o
g
ra
p
h
i
c
i
ma
g
i
n
g
u
s
i
n
g
amb
i
en
t
l
i
g
h
t
,"
O
p
t
ic
E
xp
r
es
s
,
v
o
l
.
16
,
n
o
.
26
,
p
p
.
2
1
4
3
4
-
2
1
4
3
45
,
2
0
0
8
.
[4
]
Bal
ak
r
i
s
h
n
a
n
G
.
,
D
u
ran
d
F
.
,
G
u
t
t
ag
J
.
,
"
D
et
ect
i
n
g
p
u
l
s
e
fro
m
h
ea
d
mo
t
i
o
n
s
i
n
v
i
d
e
o
,"
P
r
o
cee
d
i
n
g
s
o
f
t
h
e
I
E
E
E
Co
m
p
u
t
er
S
o
ci
e
t
y
Co
n
f
e
r
en
ce
o
n
Co
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
ec
o
g
n
i
t
i
o
n
,
Po
r
t
l
a
n
d
,
O
R,
U
S
A
:
I
E
E
E
;
2
0
1
3
.
[5
]
Al
.
N
a
j
i
A
.
,
Ch
a
h
l
J
.
,
"
Co
n
t
act
l
es
s
card
i
ac
ac
t
i
v
i
t
y
d
et
ect
i
o
n
b
as
e
d
o
n
h
ead
m
o
t
i
o
n
mag
n
i
f
i
ca
t
i
o
n
,"
I
n
t
er
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
Im
a
g
e
a
n
d
G
r
a
p
h
i
c
s
,
v
o
l
.
17
,
n
o
.
1
,
p
p
.
1
-
18
,
2
0
1
7
.
[6
]
H
e
X
.
,
G
o
u
b
ran
R
.
A
.
,
L
i
u
X
.
P.
,
"
W
ri
s
t
p
u
l
s
e
mea
s
u
re
men
t
a
n
d
a
n
al
y
s
i
s
u
s
i
n
g
E
u
l
er
i
an
v
i
d
eo
ma
g
n
i
fi
ca
t
i
o
n
,"
3
r
d
IE
E
E
E
M
B
S
I
n
t
e
r
n
a
t
i
o
n
a
l
Co
n
f
er
e
n
ce
o
n
B
i
o
m
e
d
i
ca
l
a
n
d
H
ea
l
t
h
In
f
o
r
m
a
t
i
cs
,
p
p
.
4
1
-
4
4
,
2
0
1
6
.
[7
]
L
i
u
C
.
,
T
o
rral
b
a
A
.
,
Freeman
W
.
T
.
,
D
u
ra
n
d
F
.
,
A
d
el
s
o
n
E
.
H.
,
"
Mo
t
i
o
n
mag
n
i
f
i
cat
i
o
n
,
"
A
CM
D
i
g
i
t
a
l
Li
b
r
a
r
y
,
vol.
24
,
n
o
.
3
,
p
p
.
5
1
9
-
5
26
,
2
0
0
5
.
[8
]
W
u
H
.
Y
.
,
Ru
b
i
n
s
t
ei
n
M
.
,
Sh
i
h
E
.
,
G
u
t
t
a
g
J
.
,
D
u
ran
d
F
.
,
Freeman
W
.
,
"
E
u
l
eri
an
v
i
d
e
o
mag
n
i
fi
ca
t
i
o
n
fo
r
re
v
eal
i
n
g
s
u
b
t
l
e
ch
an
g
es
i
n
t
h
e
w
o
rl
d
,"
A
CM
D
i
g
i
t
a
l
Li
b
r
a
r
y,
v
o
l
.
31
,
n
o
.
4
,
p
p
.
1
-
8
, 2
0
1
2
.
[9
]
Si
g
i
t
R
.
,
Ro
j
i
C
.
A
.
,
H
ars
o
n
o
T
.
,
K
u
s
w
a
d
i
S.
,
"
Imp
ro
v
e
d
ech
o
card
i
o
g
rap
h
y
s
eg
me
n
t
at
i
o
n
u
s
i
n
g
ac
t
i
v
e
s
h
ap
e
mo
d
e
l
an
d
o
p
t
i
ca
l
fl
o
w
,"
TE
LK
O
M
NIK
A
Te
l
eco
m
m
u
n
i
ca
t
i
o
n
Co
m
p
u
t
i
n
g
E
l
ec
t
r
o
n
i
cs
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
17
,
n
o
.
2
,
p
p
.
8
0
9
-
8
1
8
,
2
0
1
9
.
[1
0
]
W
ad
h
w
a
N
.
,
Ru
b
i
n
s
t
ei
n
M
.
,
D
u
ran
d
F
.
,
Freeman
W
.
T.
,
"
Ph
as
e
-
b
as
e
d
v
i
d
e
o
mo
t
i
o
n
p
ro
ce
s
s
i
n
g
,"
A
CM
D
i
g
i
t
a
l
Li
b
r
a
r
y
,
v
o
l
.
32
,
n
o
.
4
,
2
0
1
3
.
[1
1
]
Po
rt
i
l
l
a
J
.
,
Si
mo
n
cel
l
i
E
.
P.
,
"
A
Paramet
ri
c
T
e
x
t
u
re
Mo
d
el
Ba
s
ed
o
n
J
o
i
n
t
St
a
t
i
s
t
i
cs
o
f
Co
mp
l
ex
W
av
el
e
t
Co
effi
c
i
en
t
s
,"
In
t
J
Co
m
p
u
t
V
i
s
.
,
v
o
l
.
4
0
,
n
o
.
1
,
p
p
.
49
-
71
,
2
0
0
0
.
[1
2
]
Si
mo
n
c
el
l
i
E
.
P
.
,
Freeman
W
.
T
.
,
A
d
el
s
o
n
E
.
H
.
,
H
eeg
er
D
.
J.
,
"
Sh
i
ft
ab
l
e
Mu
l
t
i
s
cal
e
T
ra
n
s
f
o
rms
,"
I
E
E
E
Tr
a
n
s
a
c
t
i
o
n
s
o
n
In
f
o
r
m
a
t
i
o
n
Th
eo
r
y
,
v
o
l
.
38
,
n
o
.
2
,
p
p
.
5
8
7
-
6
0
7
,
1
9
9
2
.
[1
3
]
G
au
t
ama
T
.
,
V
an
H
u
l
l
e
M
.
M.
,
"
A
p
h
as
e
-
b
as
e
d
ap
p
r
o
a
ch
t
o
t
h
e
es
t
i
ma
t
i
o
n
o
f
t
h
e
o
p
t
i
cal
fl
o
w
fi
e
l
d
u
s
i
n
g
s
p
a
t
i
a
l
fi
l
t
eri
n
g
,"
IE
E
E
Tr
a
n
s
Ne
u
r
a
l
Net
w
o
r
k
s
,
v
o
l
.
1
3
,
n
o
.
5
,
p
p
.
1
1
2
7
-
11
36
,
2
0
0
2
.
[1
4
]
Sh
ah
a
d
i
H
.
I
.
,
Al
-
a
l
l
a
q
Z
.
J
.
,
A
l
b
at
t
at
H
.
J.
,
"
E
ffi
c
i
en
t
d
en
o
i
s
i
n
g
a
p
p
r
o
ach
b
a
s
ed
E
u
l
eri
a
n
v
i
d
eo
ma
g
n
i
fi
ca
t
i
o
n
fo
r
co
l
o
u
r
an
d
mo
t
i
o
n
v
ari
a
t
i
o
n
s
,"
In
t
er
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
E
l
ect
r
i
c
a
l
a
n
d
Co
m
p
u
t
er
E
n
g
i
n
ee
r
i
n
g
(IJE
C
E
)
,
v
o
l
.
10
,
n
o
.
5
,
p
p
.
4
7
0
1
-
47
11
,
2
0
2
0
.
[1
5
]
L
i
u
L
.
,
L
u
L
.
,
L
u
o
J
.
,
Z
h
an
g
J
.
,
C
h
en
X
.
,
"
E
n
h
a
n
ce
d
E
u
l
eri
an
v
i
d
eo
ma
g
n
i
fi
ca
t
i
o
n
.
I
n
:
Ima
g
e
a
n
d
S
i
g
n
al
Pr
o
ces
s
i
n
g
(CISP),
"
2
0
1
4
7
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
Co
n
g
r
e
s
s
o
n
.
D
a
l
i
a
n
,
Ch
i
n
a:
IE
E
E
,
2
0
1
4
.
[
1
6
]
W
a
d
h
w
a
N
.
,
R
u
b
i
n
s
t
e
i
n
M
.
,
D
u
r
a
n
d
F
.
,
F
r
e
e
m
a
n
W
.
T.
,
"
R
i
e
s
z
p
y
r
a
m
i
d
s
f
o
r
f
a
s
t
p
h
a
s
e
-
b
a
s
e
d
v
i
d
e
o
m
a
g
n
i
f
i
c
a
t
i
o
n
,
"
2
0
1
4
I
E
E
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
.
S
a
n
t
a
C
l
a
r
a
,
C
A
,
U
S
A
:
I
E
E
E
;
2
0
1
4
.
[1
7
]
Al
-
N
aj
i
A
.
,
L
ee
S
.
H
.
,
Ch
ah
l
J
.
,
"
A
n
effi
ci
e
n
t
mo
t
i
o
n
mag
n
i
f
i
cat
i
o
n
s
y
s
t
em
fo
r
rea
l
-
t
i
me
ap
p
l
i
ca
t
i
o
n
s
,"
M
a
ch
V
i
s
A
p
p
l
S
p
r
i
n
g
er
,
v
o
l
.
29
,
n
o
.
4
,
p
p
.
5
8
5
-
6
0
0
,
2
0
1
8
.
[1
8
]
Sh
ah
a
d
i
H
.
I
.
,
A
l
b
at
t
a
t
H
.
J
.
,
A
l
-
A
l
l
a
q
Z
.
J
.
,
T
h
ah
ab
A
.
T.
,
"
E
u
l
eri
a
n
v
i
d
eo
mag
n
i
fi
ca
t
i
o
n
:
A
rev
i
ew
,"
In
d
o
n
e
s
i
a
n
J
o
u
r
n
a
l
E
l
ect
r
i
c
a
l
E
n
g
i
n
eer
i
n
g
Co
m
p
u
t
i
n
g
a
n
d
S
ci
e
n
ce
.
,
v
o
l
.
18
,
n
o
.
2
,
p
p
.
7
9
9
-
8
1
1
,
2
0
2
0
.
[1
9
]
Al
-
a
l
l
a
q
Z
.
J
.
,
Sh
ah
ad
i
H
.
I
.
,
A
l
b
at
t
at
H
.
J.
,
"
Po
w
erf
u
l
a
n
d
L
o
w
T
i
me
Ph
a
s
e
-
Bas
e
d
V
i
d
e
o
Mag
n
i
fi
ca
t
i
o
n
E
n
h
an
ci
n
g
T
ech
n
i
q
u
e
,
"
2
0
1
9
4
t
h
S
ci
e
n
t
i
f
i
c
In
t
er
n
a
t
i
o
n
a
l
Co
n
f
e
r
en
c
e
Na
j
a
f
(S
ICN)
,
2
0
1
9
.
[2
0
]
Freeman
W
.
,
A
d
el
s
o
n
E
.
H
.
,
H
ee
g
er
D
.
,
"
Mo
t
i
o
n
w
i
t
h
o
u
t
mo
v
eme
n
t
,
"
A
CM
S
IG
G
R
A
P
H
Co
m
p
u
t
e
r
G
r
a
p
h
i
cs
,
v
o
l
.
25
,
n
o
.
4
,
p
p
.
27
-
30
,
1
9
9
1
.
[2
1
]
Mo
raes
T
.
,
A
m
o
ri
m
P
.
,
Si
l
v
a
J
.
,
Ped
ri
n
i
H
.
,
"
3
d
l
a
n
c
zo
s
i
t
erp
o
l
a
t
i
o
n
f
o
r
med
i
cal
v
o
l
u
mes
,
"
1
5
t
h
In
t
e
r
n
a
t
i
o
n
a
l
S
ym
p
o
s
i
u
m
o
n
Co
m
p
u
t
e
r
M
et
h
o
d
s
i
n
B
i
o
m
ech
a
n
i
cs
a
n
d
B
i
o
m
e
d
i
c
a
l
E
n
g
i
n
ee
r
i
n
g
,
2
0
1
8
.
[
2
2
]
Y
o
o
D
.
S
.
,
C
h
a
n
g
J
.
,
P
a
r
k
C
.
H
.
,
K
a
n
g
M
.
G.
,
"
V
i
d
e
o
r
e
s
a
m
p
l
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
s
i
m
u
l
t
a
n
e
o
u
s
d
e
i
n
t
e
r
l
a
c
i
n
g
a
n
d
i
m
a
g
e
u
p
s
c
a
l
i
n
g
w
i
t
h
r
e
d
u
c
e
d
j
a
g
g
e
d
e
d
g
e
a
r
t
i
f
a
c
t
s
,"
E
U
R
A
S
I
P
J
o
u
r
n
a
l
o
n
A
d
v
a
n
c
e
s
i
n
S
i
g
n
a
l
P
r
o
c
e
s
s
i
n
g
,
v
o
l
.
8
8
,
pp.
1
-
24
,
2
0
1
3
.
[2
3
]
Pars
an
i
a
P
.
S
.
,
V
i
rp
ari
a
P
.
V.
,
"
A
co
mp
ara
t
i
v
e
an
a
l
y
s
i
s
o
f
i
mag
e
i
n
t
erp
o
l
a
t
i
o
n
a
l
g
o
ri
t
h
m
s
,"
In
t
er
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
ce
d
R
es
e
a
r
c
h
i
n
Co
m
p
u
t
er
a
n
d
Co
m
m
u
n
i
c
a
t
i
o
n
E
n
g
i
n
ee
r
i
n
g
,
v
o
l
.
5
,
n
o
.
1
,
p
p
.
2
9
-
3
4
,
2
0
1
6
.
[2
4
]
Mad
h
u
k
ar
B
.
N
.
,
N
aren
d
ra
R.
,
"
L
an
czo
s
res
amp
l
i
n
g
fo
r
t
h
e
d
i
g
i
t
a
l
p
ro
ce
s
s
i
n
g
o
f
remo
t
e
l
y
s
e
n
s
e
d
i
mag
es
,
”
Pro
ceed
i
n
g
s
o
f
In
t
er
n
a
t
i
o
n
a
l
C
o
n
f
er
e
n
ce
o
n
V
LS
I,
C
o
m
m
u
n
i
ca
t
i
o
n
,
A
d
v
a
n
ce
d
D
ev
i
ces
,
S
i
g
n
a
l
s
&
S
ys
t
em
s
a
n
d
Net
wo
r
ki
n
g
(V
C
A
S
A
N
-
2
0
1
3
)
,
2
0
1
3
.
[2
5
]
Mas
s
ac
h
u
s
et
t
s
I
n
s
t
i
t
u
t
e
o
f
T
ech
n
o
l
o
g
y
,
"
V
i
d
i
o
Ma
g
n
i
fi
ca
t
i
o
n
,
"
2
0
1
0
.
[
O
n
l
i
n
e].
A
v
ai
l
ab
l
e
:
h
t
t
p
:
/
/
p
e
o
p
l
e.
cs
a
i
l
.
mi
t
.
e
d
u
/
mru
b
/
v
i
d
ma
g
/
Evaluation Warning : The document was created with Spire.PDF for Python.