TELKOM
NIKA
, Vol.14, No
.2, June 20
16
, pp. 523~5
3
0
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2947
523
Re
cei
v
ed
No
vem
ber 2, 20
15; Re
vised Janua
ry 2
7
, 20
16; Accepted
February 16,
2016
Data Analysis f
o
r Image Transmitted using Discrete
Wavelet
Transform and Vector Quantization
Compression
Musta
pha Khelifi
1
*, A. Moula
y
lakhdar
1
, Iman Ela
w
ady
2
1
Departme
n
t of Electrical En
gi
neer
ing, T
ahri M
ohamm
ed U
n
iversit
y
, Bech
ar, Algeri
a
/ lab:
T
I
T
2
Departme
n
t of Electrical En
gi
neer
ing, T
ahri Mohamm
ed U
n
iversit
y
,
Bechar, Alg
e
ri
a / lab: CAOSEE, BP 417 Rou
t
e Kenads
a, Béchar 0
8
0
00, Algeri
a
, +213 4
9
23 89 9
3
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: Khelifi
_
m@
yaho
o.fr
A
b
st
r
a
ct
In this p
aper w
e
are
go
ing
to
study the
effect
of chan
nel
no
ise i
n
i
m
a
ge c
o
mpress
ed w
i
th vecto
r
qua
nti
z
a
t
i
on a
nd discr
ete w
a
velet transfor
m
. T
he obj
ecti
ve
of this study i
s
to analy
z
e
a
nd u
ndersta
nd
the
w
a
y that the n
o
ise
attack tra
n
smitted
data
by do
in
g l
o
t
of
tests like
div
i
d
i
ng th
e i
n
d
i
ces
in
different
lev
e
ls
accord
ing to
di
screte w
a
velet
transform
an
d divid
i
n
g
eac
h l
e
vel i
n
fra
m
es
of bits. T
he coll
ected i
n
for
m
ati
o
n
w
ill hel
ps us to
propos
e sol
u
ti
ons to make th
e receiv
ed i
m
a
ge
more res
i
sti
b
le to the ch
an
nel n
o
ise
also t
o
ben
efit from th
e goo
d repr
ese
n
tation o
b
tai
n
e
d
by us
in
g vect
or qua
nti
z
a
t
i
o
n
and discr
ete w
a
vel
e
t transform.
Ke
y
w
ords
: DWT, V
Q, LBG
,
BSC channel, BER
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In re
cent ye
a
r
s l
o
t of rese
arch h
a
s be
e
n
do
ne to
en
han
ce th
e qu
ality of re
ceiv
ed ima
ge
after tran
smi
ssi
on in
noi
sy chan
nel, this p
r
obl
em can be solve
d
usi
ng
diffe
rent
te
chniq
u
e
s
(so
u
rce
codin
g
, chann
el co
ding , joint so
urce ch
ann
el
codi
ng or by optimize in
re
storatio
n filters)
as
shown in artic
l
es
[1-3].
A good rep
r
ese
n
tation a
nd comp
ression of imag
e
can m
a
ke the differe
nce
on the
quality of re
ceived imag
e, also to u
nde
rstand
ho
w the ch
annel
noi
se atta
ck th
e
transmitted d
a
ta
can be p
r
ovid
e an efficient techni
que
s wi
th less
con
s
u
m
ption in re
source (p
ro
ce
ssi
ng unit, time,
data si
ze, …) and ban
dwi
d
th usag
e.
In image com
p
re
ssi
on, we
kno
w
that, the co
mbi
natio
n betwe
en vector qu
antization an
d
discrete
wav
e
let transfo
rm provide a good repr
ese
n
tation of recon
s
tru
c
ted i
m
age. Since
the
discrete
wavelet tran
sfo
r
m analy
z
e i
m
age
and
give
us lot of
dat
a ab
out 2
D
signal
whi
c
h
make
the differen
c
e in the qua
lity of recon
s
tru
c
ted ima
ge ho
wever
In case of transmi
ssion t
he
comp
re
ssion
usin
g VQ
and
DWT i
s
so
sen
s
itive to the ch
annel n
o
ise
as
sho
w
n in
the
artic
l
es
[4-5].
To solve this pro
b
lem
we
are
goin
g
to
analyze dat
a du
ring t
r
an
smissio
n
to
discove
r
what is the m
o
st impo
rtant parts that can
make
the differen
c
e in the
quality or re
ceived image
by
doing l
o
t of te
sts i
n
tra
n
smi
tted indices a
c
cordi
ng to
th
e de
com
p
o
s
e
d
levels p
r
ovided by
discre
t
e
wavelet transform.
The pap
er is organi
ze
d as follow: In the first part
we will intro
d
u
ce the com
p
re
ssi
on
usin
g
vecto
r
quanti
z
ation and
di
screte wavelet
tra
n
sform. The
se
con
d
pa
rt th
e effect of th
e
cha
nnel n
o
ise in image
co
mpre
ssed
with vector q
u
a
n
tization a
nd
discrete
wav
e
let tran
sform. In
third pa
rt the effect of the cha
nnel n
o
ise in ea
ch lev
e
l gene
rate
d by discrete
wavelet tran
sfo
r
m
with different BER. In the fourth part we will divi
de each level to ei
ght subl
evels finally we try to
prop
ose som
e
solutio
n
s
a
c
cordi
ng to the orbit
ed re
sults to that
can e
nha
nce
the quality of
received ima
ge.
2. Compres
s
i
on Using DWT and Vec
t
or Quan
tiza
tion
A quanti
z
er
simply redu
ces th
e nu
m
ber
of
bits
need
ed to
store the
tran
sform
e
d
coeffici
ents
by redu
cin
g
the pre
c
i
s
io
n of thos
e
values [6].
The ba
si
c p
r
inci
ple of v
e
ctor
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 523 – 53
0
524
quanti
z
ation
based imag
e
comp
re
ssi
on
techniq
u
e
s
is to match e
a
ch in
put vector with a co
de-
vector i
n
the
cod
ebo
ok
so that the di
storti
on
bet
ween the
input
vector
and
the cho
s
en
code-
vector is mi
ni
mum [7] by evaluating the
Euclid
e
an di
stance b
e
twe
e
n
the input vector a
nd ea
ch
cod
e
word
in t
he
cod
ebo
ok
[8] . On
ce th
e
clo
s
e
s
t
cod
e
w
ord i
s
fo
und
, the ind
e
x of
that co
de
word
is
sent
thro
ug
h a
ch
ann
el .
W
he
n the
en
cod
e
r re
ceive
s
the
ind
e
x of
the
cod
e
word, it re
pla
c
e
s
the
index
with the
asso
ciate
d
[
9
]. Quanti
z
ati
on i
s
a
n
ir
rev
e
rsi
b
le
pro
c
e
ss.
That i
s
, in
gen
eral, th
ere is
no
way to fin
d
the
origi
nal
value fro
m
th
e qu
antized v
a
lue [1
0]. Th
e differe
nce b
e
twee
n the
in
put
and outp
u
t si
gnal
s of the quantizer be
co
mes the
qu
an
tizing erro
r, or quanti
z
ing n
o
ise [11].
The Discrete
Wavelet Tra
n
sform
p
r
ovi
des
suffi
cient
informatio
n
both for anal
ysis a
n
d
synthe
sis of the origi
nal si
gnal, with a si
gnifi
ca
nt redu
ction in the computation ti
me [12].
Discrete
Wa
velet Tran
sf
orm
(DWT
) is u
s
e
d
in
a variety
of sign
al p
r
oce
s
sing
appli
c
ation
s
,
su
ch a
s
v
i
deo
comp
re
ssi
on, Intern
et comm
uni
cation
s
com
p
re
ssi
on, ob
ject
recognitio
n
, and nume
r
ical
analysis. It can effici
entl
y
repre
s
e
n
t some sig
nal
s, espe
cially o
nes
that have localize
d
ch
ang
e [13]
The ste
p
s
co
mpre
ssion u
s
ing DWT and
VQ as sho
w
n
in Figure 1 a
r
e:
1)
De
comp
ose the image u
s
i
ng the DWT tran
sform to 3
levels.
2)
Partition obta
i
ned ba
nd
s in
to 4x4 blocks
then apply Vector
Qua
n
tization on ea
ch block an
d
result is the compresse
d
image.
Figure 1. Image com
p
ressi
on step
s u
s
in
g DWT and V
Q
.
This sy
stem
got high co
mpre
ssion ra
tio wi
thout loss of mu
ch
informatio
n b
e
ca
use
applying
DWT tran
sform
we mi
nimize
the dom
ain
of co
deb
oo
k vecto
r
. Thi
s
will
helpful
in
achi
eving hig
h
comp
re
ssio
n ratio witho
u
t
loss of information [14].
3. Transmiss
i
on Chann
e
l
In this
sectio
n, it is explai
ned the
re
sul
t
s of
re
se
arch and
at the
same
time is
given the
comp
re
hen
si
ve discussio
n
.
Result
s ca
n
be pre
s
e
n
te
d in figure
s
, grap
hs, table
s
and oth
e
rs
that
make the
rea
der un
de
rsta
nd ea
sily [2,
5]. The discu
ssi
on can be
made in seve
ral su
b-ch
apt
ers.
The
bin
a
ry symmetric ch
annel (BSC)
is
def
in
ed
by the ch
an
nel dia
g
ra
m
sho
w
n i
n
Figure 2, and
its chan
nel
matrix is give
n by Equation
(1):
(1)
The ch
ann
el has two inpu
ts (
x
1 =
0,
x
2 = 1) and t
w
o output
s (
y
1 = 0,
y
2
= 1). The
cha
nnel i
s
sy
mmetric
be
ca
use th
e prob
ability of rece
iving a 1 if a 0 is sent is t
he same a
s
t
he
prob
ability of receiving a
0
if a 1 is
sent.
This
comm
on
transitio
n p
r
o
bability is de
n
o
ted by p [15]
.
The e
r
ror ev
ents
are
al
so
inde
pend
ent
of the
data
bits [16]. T
h
i
s
i
s
the
sim
p
lest mo
del
of a
cha
nnel
with
errors, yet it captu
r
e
s
m
o
st of
the
co
mplexity of the ge
neral p
r
oble
m
[17].
The
cap
a
city of this ch
ann
el given by Equatio
n (2):
p
1
p
p
p
1
P(y/x)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Data Anal
ysi
s for Im
age Transm
i
tted usi
ng Di
scre
te
Wa
velet Tran
sform
…
(Mu
s
tapha Khelifi)
525
(2)
With the bina
ry entropy fun
c
tion given by
Equation (3
):
(3)
Figure 2. Binary symmetri
c
ch
ann
el
4. The Effe
ct of Error in Image Compr
essed
w
i
th DWT
+
QV
In this part we are goi
ng to see ho
w th
e noise atta
cks the ima
ge after tran
smission we
will
take an example
of Lena
im
age size
512x5
1
2
,
the block size is
4x4. The co
deb
oo
k is
gene
rated by
LBG algo
rith
m, with different BER (10
-3
, 10
-2
) as sho
w
n in Figu
re
3.
BER=
10-
3
, PSN
R=
2
8
.
91
dB
BER=
1
0
-2
, P
S
N
R
=
2
4
.
33
dB
Figure 3. Re
ceived image
usin
g DWT a
nd VQ com
p
ression
with di
fferent BER
For tra
n
smi
s
sion
with the
same p
a
ra
meters but
with image
compresse
d
b
y
VQ the
simulatio
n
, re
sults a
r
e represe
n
ted a
s
shown in Figu
re 4.
BER=
10-
3
, PSN
R=
2
8
.
30
dB
BER=
10
-2
, PSNR= 25.59
dB
Figure 4. Re
ceived image
comp
re
ssed
by VQ with different BER
use
channel
bit
in
1
H(p)
C
p)
(
p)
(
p
p
p
H
1
log
1
log
2
2
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 523 – 53
0
526
From
th
e sim
u
lation re
sult
s
the quality of
re
ceived
i
m
age usi
ng VQ
is better than
the
received one us
ing DWT
and VQ. However the
PSNR of rec
o
ns
truc
t
ed image
us
ing DWT and
QV (PSNR=3
3
.35dB) i
s
be
tter than the
recon
s
tr
u
c
ted
one
usi
ng V
Q
(PSNR=30
,67dB). Its lo
ok
like
we
can n
o
t benefit fro
m
this en
han
ceme
nt du
ri
n
g
the tran
smi
ssi
on a
nd thi
s
is
wh
at we
are
going to inve
stigate it in the next parts.
5. The Effe
ct of Error in the Lev
e
ls of
Image Comp
resse
d
w
i
th
DWT and VQ
Figure 5. Flowchart to stu
d
y the noise
ef
f
e
ct
s on t
h
e
indice
s in lev
e
ls
Figure 6. PSNR in fun
c
tio
n
of BER
10
-3
10
-2
10
-1
14
16
18
20
22
24
26
28
30
32
BER
PS
N
R
[
d
B]
P
S
N
R
i
n
f
unc
t
i
on
of
B
E
R
(
B
oa
t
i
m
ag
e)
I
n
d
i
ce
s
B
F
I
n
d
i
ce
s
l
e
v
e
l
3
I
n
d
i
ce
s
l
e
v
e
l
2
I
n
d
i
ce
s
l
e
v
e
l
1
10
-3
10
-2
10
-1
14
16
18
20
22
24
26
28
30
32
34
BER
PS
N
R
[
d
B]
P
S
N
R
i
n
f
u
nc
t
i
on of
B
E
R
(
L
e
na i
m
a
ge)
In
d
i
c
e
s
B
F
I
n
d
i
c
e
s
le
v
e
l
3
I
n
d
i
c
e
s
le
v
e
l
2
I
n
d
i
c
e
s
le
v
e
l
1
10
-3
10
-2
10
-1
16
18
20
22
24
26
28
30
32
BER
PS
N
R
[d
B
]
P
S
N
R
in
f
u
n
c
t
i
o
n
o
f
B
E
R
(
G
o
l
d
h
ill
im
a
g
e
)
I
ndi
c
e
s
B
F
I
n
d
i
c
e
s
le
v
e
l 3
I
n
d
i
c
e
s
le
v
e
l 2
I
n
d
i
c
e
s
le
v
e
l 1
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Data Anal
ysi
s for Im
age Transm
i
tted usi
ng Di
scre
te
Wa
velet Tran
sform
…
(Mu
s
tapha Khelifi)
527
We
are
goi
n
g
to divid
e
t
he in
dices of
vector qu
an
tization
acco
rding to
the l
e
vels
of
discrete
wav
e
let tra
n
sfo
r
m (BF
ind
e
x, index
level
3
,
index l
e
vel
2, index
level
1) ea
ch
time
we
are goi
ng to test the effect of the noise in a matr
ix
of bits and its indices, the
n
cal
c
ulate
s
the
PSNR of the degrade
d ima
ge, as sho
w
n
in Figure 5.
For the sim
u
lation re
sults we use two
images
(Le
na, Boat,Goldhill), ea
ch index is
rep
r
e
s
ente
d
by 8 bit
s
with differe
nt B
E
R (0.001,
0.01, 0.02,
0.0
5
, 0.1), a
s
shown in
Figu
re 6.
The
simulatio
n
re
sult
s sho
w
that the m
o
st sen
s
itive levels a
r
e: le
vel 3 and B
F
level co
mpa
r
ed
with other lev
e
ls
6. The Effe
ct of the Chan
nel Noise in the Frame
s
of bits
In this part we are g
o
ing t
o
divide the
bits of
indice
s in 8 fram
es of bits. Each
frame is
con
s
tru
c
ted
b
y
bits of indice
s with the
same
we
i
ght. We test the
effect of noise in one fra
m
e
each time the
n
we calculat
e the PSNR
of receiv
ed i
m
age
with different BER (0
.001, 0.01, 0.02,
0.05, 0.1), as
sho
w
n in Fig
u
re 7.
Figure 7. PSNR in fun
c
tio
n
of BER
7. The Appli
cation o
f
Un
equal Error
Protec
tion in Lev
e
ls
To
te
st
the rightne
ss
of our study an
d
colle
cted i
n
formatio
n
we u
s
e
une
qu
al erro
r
prote
c
tion a
s
appli
c
ation to
our data. Th
e main step
s
are a
s
sh
own
in Figure 8.
10
-3
10
-2
10
-1
16
18
20
22
24
26
28
30
32
34
BER
PSN
R
[
d
B]
P
S
N
R
i
n
f
u
n
c
t
i
on
o
f
B
E
R
(L
ena i
m
age)
fr
a
m
e
1
fr
a
m
e
2
fr
a
m
e
3
fr
a
m
e
4
fr
a
m
e
5
fr
a
m
e
6
fr
a
m
e
7
10
-3
10
-2
10
-1
16
18
20
22
24
26
28
30
32
BE
R
PSN
R
[
d
B]
P
S
N
R
i
n
f
u
nc
t
i
o
n
of
B
E
R
(
B
oa
t
i
m
ag
e)
fr
a
m
e
1
fr
a
m
e
2
fr
a
m
e
3
fr
a
m
e
4
fr
a
m
e
5
fr
a
m
e
6
fr
a
m
e
7
10
-3
10
-2
10
-1
16
18
20
22
24
26
28
30
32
BE
R
PS
N
R
[
d
B]
P
S
N
R
in
f
u
n
c
t
i
o
n
o
f
B
E
R
(
G
o
l
d
h
ill ima
g
e
)
f
r
am
e 1
f
r
am
e 2
f
r
am
e 3
f
r
am
e 4
f
r
am
e 5
f
r
am
e 6
f
r
am
e 7
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 523 – 53
0
528
Figure 8. une
qual erro
r pro
t
ection Flo
w
chart (UEP)
For the
simulation
we use th
ree im
ages
(Lena, B
oat, Go
ldhill )
si
ze
512x512 with
different BER and co
deb
oo
k gen
erate
d
b
y
LBG t
he results are
sho
w
n in Figures 9
,
10 and 11.
PSNR= 33.3
533dB
PSNR= 31.3
346dB
PSNR= 31.8
231dB
Figure 9. Re
cived image
s with BER=10
-3
PSNR= 32.3
241dB
PSNR= 30.4
783dB
PSNR= 31.0
232dB
Figure 10. Re
cived imag
es
with BER=10
-2
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Data Anal
ysi
s for Im
age Transm
i
tted usi
ng Di
scre
te
Wa
velet Tran
sform
…
(Mu
s
tapha Khelifi)
529
PSNR= 32.9
681dB
PSNR= 26.3
794dB
PSNR= 28.1
216dB
Figure 11. Re
cived imag
es
with BER=0.0
5
We
simulate
the tran
smi
s
sion chann
el
of image u
s
i
ng DWT a
n
d
VQ for com
p
re
ssi
on,
whi
c
h p
r
od
uce indices.
We
decompo
se t
he indi
ce
s a
c
cording to th
e
levels ea
ch i
ndices i
s
cod
ed
in 8 bits
. All matrices
of bits
of
those in
dice
s are co
d
ed usi
ng RS
cod
e
.
The
simulatio
n
re
sult
s sho
w
the
rightn
e
s
s of ou
r d
a
ta analy
z
e. T
he u
s
e
of UEP as
a
solutio
n
opti
m
ize in the u
s
e of re
dund
a
n
t data from the less to the
most impo
rta
n
t data.
Of course, the quality of receive
d
imag
e is
goin
g
to be low
with hi
gh BER level
s
but it’s
still better co
mpared with t
he com
p
ressi
on usi
ng VQ and UEP a
s
shown in Tabl
e 1.
Table 1. The
quality of received image
usin
g VQ co
mpre
ssion a
n
d
UEP.
PSNR BER
(10
-
3
) BER
(10
-
2
) BER
(5x10
-
2
)
Lena
30.1708 dB
29.0494 dB
24.9971dB
Boat 28.5913
dB
27.9619dB
24.3983dB
G
o
ldhill 29.1297
dB
28.2048dB
24.5350dB
Comp
ared
with other
re
search p
ape
rs co
n
c
e
r
ne
d i
n
de-noi
sing
image [18
-
1
9
], our
contri
bution t
o
analyze the
data du
ring t
r
an
smi
ssi
on
provide
s
a
go
od way to un
derstand
ho
w the
noise attack t
he data, al
so
it tell us wh
at
is t
he mo
st sensitive d
a
ta
whi
c
h ma
ke t
he differe
nce
in
the quality of
the re
ceive
d
i
m
age, thi
s
ca
n help
u
s
to
prop
ose
som
e
techniqu
es
that ca
n p
r
ot
ect
the importa
nt data by usin
g cha
nnel
co
ding o
r
by
se
nding a
dditio
nal data that
can hel
p u
s
in
resto
r
atio
n.
8. Conclusio
n
In this paper
we study the
effect of channel noi
se in
image com
p
resse
d
by DWT and
VQ this stu
d
y sh
ow u
s
th
e
keys that
can
help
u
s
to
e
nhan
ce
the
q
uality of recei
v
ed imag
e, al
so
provide
an
alysis of d
a
ta a
nd
kno
w
led
g
e
ab
out h
o
w
the noi
se
ch
annel
effect
on the
imag
e
the
benefits
of this stu
d
y is
co
nce
n
trated
a
bout ho
w to
sele
ct a d
a
ta
that can
ma
ke a
differe
n
c
e in
the qu
ality of re
ceived
im
age
our sug
gestio
n
to e
nhan
ce
the quality
of
im
age by
u
s
ing
bit
corre
c
tor fo
r sen
s
itive part
of data.
Lot of flowch
arts
ca
n be
p
r
opo
sin
g
to p
r
otec
t the im
p
o
rtant d
a
ta in
image. Th
e
prop
osed
one is n
o
t the best a
s
we
guess, also it is applied
j
u
st in levels
so a
s
future
work may be
we
c
a
n d
e
s
i
gn
so
me
flow
ch
arts
us
in
g U
EP
a
p
p
lied
in le
v
e
ls
and
the f
r
ame
of bits which
can
give
us
a good results.
Referen
ces
[1]
Mohse
n
N
a
sri,
Abd
e
lh
amid
Hela
li, H
a
lim
Sgha
ier, H
a
ss
en M
aaref.
Efficient JPEG
200
0 Imag
e
Compress
io
n
Scheme for
Multiho
p
Wireless
Net
w
o
r
ks.
T
E
LKOMNIKA T
e
lec
o
mmunic
a
tio
n
Co
mp
uting El
e
c
tronics an
d C
ontrol
. 20
11; 9(
2): 311-3
18.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 523 – 53
0
530
[2]
Ali M F
a
dhi
l,
Haid
er M A
l
Sa
bba
gh.
Perfor
mance Analy
s
is for Bit-Error-
Rate
of DS-CDMA Sensor
Net
w
ork S
y
stems
w
i
t
h
Sourc
e
Coding.
T
E
LKOMNIKA T
e
leco
mmu
n
ic
ati
on Co
mputi
n
g
Electronic
s
and C
ontrol
. 2
012; 10(
1): 165
-170.
[3]
L
y
di
a Sar
i
, An
tonius
Adit
ya.
Raptor
Co
de
for En
erg
y
-Efficient
W
i
rel
e
ss
Bod
y
Are
a
N
e
t
w
o
r
k
Dat
a
T
r
ansmission.
TELKOMNIKA Telec
o
mmu
n
icati
on
Co
mp
uting
Electro
n
i
cs an
d C
ontro
l
. 201
5; 1
3
(1):
277-
283.
[4]
Deb
nath JK, Rahim
NMS, W
a
i-Keun
g
F
ung.
A mo
difie
d
Vector Quanti
z
a
t
i
on base
d
i
m
a
g
e
compressi
on
techn
i
qu
e
usin
g w
a
vel
e
t tran
sform
.
IEEE International J
o
int
Conference on Neura
l
Net
w
orks. 20
0
8
: 171-1
76.
[5]
Binit Amin, Pat
e
l Amrut
bhai.
Vector Qua
n
tiz
a
tion
bas
ed
Lo
ss
y
Ima
g
e
Co
mpressio
n
usin
g W
a
vel
e
ts
–
A Revie
w
.
Int
e
rnati
ona
l Jo
ur
nal
of Inn
o
vati
ve Res
ear
c
h
i
n
Scie
nce, E
n
gin
eeri
ng
an
d
T
e
chno
logy
.
201
4; 3(3): 105
17-1
052
3.
[6]
H Malep
a
ti. Di
gital Me
dia Pro
c
essin
g
DSP Al
gorit
hms Usin
g C. 1st Edition
.
Ne
w
n
es. 20
1
0
: 586.
[7]
Ajol Kum
a
r Ra
y, T
i
nku Achar
ya. Informati
on
T
e
c
hnol
og
y P
r
incip
l
es a
nd A
pplic
atio
ns. Prentic Ha
ll o
f
India.
20
04
.
[8]
KV Kal
e
, SC
Mehrotra.
Com
puter V
i
sio
n
a
nd
Inform
atio
n
T
e
chnolo
g
y
: Advanc
es and
App
licati
ons.
201
0; 50.
[9]
R
y
sz
ard S C
horas. Imag
e
Processi
ng
and
Commu
ni
cations
Cha
lle
nges. Spr
i
ng
e
r
Scienc
e &
Busin
e
ss Medi
a. 2013: 2
36-2
38.
[10]
Yun Q Shi, Hui
f
ang Sun. Imag
e and Vi
de
o C
o
mp
ressi
on for
Multimedi
a En
gin
eeri
ng. Sec
ond Ed
itio
n.
CRC Press. 20
08.
[11]
Z
hou W
a
n
g
, Alan C Bov
i
k. Mean s
quar
ed e
rror love it or
l
eave it.
IEEE signal pr
ocessing maga
z
i
ne.
200
9: 99-1
17.
[12]
Vale
ntina Zh
ar
kova. Artificial
Intellig
enc
e i
n
Reco
gniti
on
and Cl
assific
a
tion of Astro
p
h
y
sic
a
l an
d
Medic
a
l Image
s (Studies in C
o
mputati
o
n
a
l Intelli
ge
nce). Sprin
ger. 20
07: 343.
[13]
Micha
e
l W
eek
s. Digital Si
gn
al Process
i
ng
Using MAT
L
AB & W
a
velets. Seconf Editio
n. Jones &
Bartlett Learn
i
n
g
. 2010: 2
71.
[14]
T
e
jas S Patel
,
Ravi
ndra
M
odi, K
e
yur J
Patel. Imag
e
Compress
io
n
Using
DW
T
and V
e
ctor
Quantizati
on.
Internatio
na
l Journ
a
l of Inn
o
vative R
e
se
a
r
ch in Co
mp
uter and C
o
mmu
n
icati
on
Engi
neer
in
g
. 2013; 1(3): 6
53.
[15]
H
w
ei P Hs
u. Anal
og a
nd Di
git
a
l Comm
unic
a
tions
(Sch
aum'
s
Outlines). Se
cond E
d
itio
n. McGra
w
-
Hil
l
Educati
on. 20
0
2
.
[16]
Milan
So
nka,
Vaclav
Hlav
a
c,
Rog
e
r Bo
yl
e.
Image Pr
oces
sing A
n
a
l
ysis
and M
a
ch
ine
Visio
n
. T
h
ird
Editio
n. T
H
OMSON,
CL Engi
neer
ing. 2
007.
[17]
John Mi
ano. C
o
mpress
ed Image F
i
l
e
F
o
rma
ts. Addison W
e
sle
y
.
199
9.
[18]
Yan F
e
ng,
H
ua
Lu,
Xili
an
g
Z
eng.
Image
Rest
or
atio
n
Based
o
n
H
y
brid
Ant
Col
o
n
y
Al
gorithm.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol
. 20
15; 13(4): 1
298
-130
4.
[19]
Jian R
en, Hua
Lu, Xili
ang Z
e
n
g
. Image Den
o
i
s
ing
Bas
ed o
n
K-means Si
ng
ular Va
lue D
e
c
o
mpos
ition.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol.
20
15; 13(4): 1
312
-131
8.
Evaluation Warning : The document was created with Spire.PDF for Python.