Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
1
0
,
No.
1
,
Febr
uar
y
2020
, p
p. 78
6~80
0
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
1
.
pp786
-
800
786
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
A
n
ew
p
artial
i
m
age
e
nc
ryption
m
ethod fo
r
d
ocum
ent
i
ma
ges
u
sing
v
arian
ce
b
ased
q
uad
t
ree
d
ecomp
osition
C
.
R
.
Re
vann
a
1
, C
.
Kesh
av
amu
r
thy
2
1
Resea
rch
Schol
ar
a
t
Ja
in
Univ
er
sit
y
,
B
anga
lor
e,
India
1
Facul
t
y
of ECE, Government En
gine
er
ing
Col
le
g
e,
R
amana
gar
a,
Karna
ta
k
a, I
ndi
a
2
Profess
or,
Depa
rtment
of
ECE, S
RS
IT,
Bangalo
re,
K
arn
ataka
,
In
dia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
18
, 201
8
Re
vised
A
ug
2
1
,
20
19
Accepte
d
Se
p
27
, 20
19
The
proposed
m
et
hod
par
tial
l
y
and
comple
tel
y
enc
r
y
p
ts
the
gra
y
sca
l
e
Docum
ent
images.
The
comple
te
image
enc
r
ypti
on
is
al
so
p
erf
orm
ed
to
compare
the
p
er
form
anc
e
with
t
he
exi
sting
enc
r
y
pt
ion
m
et
hods.
The
par
ti
a
l
enc
r
y
pt
ion
is
ca
rrie
d
out
b
y
s
egmenti
ng
the
i
m
age
using
the
Quad
-
tre
e
dec
om
positi
on
m
et
hod
base
d
o
n
the
v
ar
i
ance
o
f
the
image
b
lock.
Th
e
imag
e
bloc
ks
with
un
i
form
pixe
l
le
ve
ls
are
conside
r
e
d
insigni
fi
ca
n
t
bloc
ks
and
othe
rs
the
sign
ifi
c
ant
bloc
ks
.
The
pixels
in
the
significant
bloc
ks
ar
e
per
m
ute
d
b
y
usi
ng
1D
Skew
te
nt
cha
ot
ic
m
ap.
Th
e
par
t
ia
l
l
y
en
cr
ypte
d
image
b
loc
ks
are
fur
th
er
per
m
ute
d
usi
ng
2D
Henon
m
ap
to
inc
re
ase
the
sec
ur
i
t
y
le
ve
l
and
fed
as
input
to
compl
et
e
enc
r
y
pt
ion.
The
comple
te
e
ncr
y
p
ti
on
i
s
ca
rri
ed
out
b
y
diffusing
the
par
tially
en
cr
y
p
te
d
image.
Tw
o
le
v
el
s
of
diffusion
are
p
erf
orm
ed.
Th
e
first
le
v
el
sim
pl
y
m
odif
ie
s
the
pixe
ls
i
n
the
par
ti
a
lly
encr
y
pt
ed
image
wi
th
the
B
ern
oulli’
s
cha
otic
m
ap.
The
sec
ond
le
ve
l
esta
b
li
shes
the
interde
pend
ency
bet
w
ee
n
ro
ws
and
col
um
ns
of
the
first
le
ve
l
diffuse
d
i
m
age
.
The
exp
eri
m
ent
is
cond
uct
ed
for
bo
th
par
tial
an
d
comp
le
te
imag
e
enc
r
y
pt
ion
on
the
Docum
ent
images.
The
propos
ed
sche
m
e
y
i
el
ds
b
et
t
er
r
e
sults
for
both
par
tial
and
complete
en
cr
y
pt
ion
on
Speed
,
stat
isti
ca
l
and
d
y
nami
ca
l
atta
ck
s.
The
resul
ts
e
nsure
better
sec
urity
when
compare
d
to
ex
i
sting
en
cr
y
pti
on
sche
m
es.
Ke
yw
or
d
s
:
MSE
NP
CR
PSN
R
SSI
M
UA
C
I
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
C
.
R
.
Re
vanna,
Re
search
Sc
hola
r
at
Jain
U
nive
rsity
, Ban
galo
re, India
Faculty
of EC
E,
Governm
ent Enginee
rin
g
C
ollege,
Ram
anag
ara
, Kar
nataka
, In
di
a
.
Em
a
il
:
rev
an
na
cr20
08@g
m
ai
l
.co
m
1.
INTROD
U
CTION
The
sto
ra
ge
a
nd
e
xc
ha
ng
e
of
im
ages
with
hi
gh
de
finiti
on
,
re
dundan
ci
es
and
c
orre
la
ti
on
of
a
n
i
m
po
rtant
doc
um
ent
at
a
hig
her
rate
of
tra
ns
m
issi
on
it
self
ta
kes
a
long
er
ti
m
e
and
the
encr
ypti
on
of
su
c
h
i
m
ages
ta
kes
add
it
io
nal
com
pu
ta
ti
on
al
tim
e
too
.
T
hi
s
requires
a
balance
bet
ween
sec
uri
ty
and
synch
ronizat
io
n
f
or
real
ti
m
e
app
li
cat
ion
s
.
In
s
uch
sit
uati
on
s
,
w
her
e
high
de
finiti
on,
low
m
e
m
or
y
a
nd
lo
w
powe
r
are
the
lim
it
at
ion
s
o
f
the
res
ources,
pa
rtia
l
encr
ypti
on
of
the
data
is
adv
anta
ge
ous
than
encr
y
pting
a
n
entire
im
age.
Partia
l
encr
ypt
ion
helps
i
n
re
du
ci
ng
t
he
co
m
pu
ta
ti
on
s
an
d
ba
ndwi
dth
.
The
do
c
um
ent
i
m
ages
consi
st
of
c
orr
el
at
ed
an
d
unc
orrelat
ed
par
ts
.
Enc
ryptio
n
of
only
the
c
orre
la
te
d
pa
rt
suffi
ces
tha
n
e
ncr
y
pting
the
com
plete
i
m
age.
The
co
r
e
idea
of
pa
rtia
l
encr
ypti
on
is
to
first
identi
fy
the
sign
ific
a
nt
pix
el
s
or
re
gion
of
pix
el
s
a
nd
the
n
enc
rypt
that re
gion.
Pa
rtia
l
encr
y
ption
al
so i
nclud
e
s
e
ncr
y
ption
of d
at
a wi
th
dif
fer
e
n
t
se
cur
it
y
le
vels
to
s
uit
e
nd
us
e
r
c
us
to
m
er
requirem
ent.
A
sig
nifica
nt
re
gion
from
the
c
om
plete
i
m
age
m
a
y
be
s
el
ect
ed
ei
ther
sta
ti
sti
cal
ly
or
dynam
i
cal
ly
fo
r
pa
rtia
l
encr
y
ption
t
o
fu
lfil
l
sec
ur
i
ty
and
com
pu
t
at
ion
al
ti
m
e
fo
r
real
tim
e
app
li
cat
ion
s.
S
o
the
p
ar
ti
al
encr
ypti
on
increases
the
eff
ic
acy
of
en
crypti
on
by
re
du
ci
ng
c
om
pu
ta
ti
on
siz
e. To en
sure
a go
od sec
ur
it
y l
evel a m
ini
m
al
d
at
a o
f
12.5%
has
t
o be e
ncr
y
pted [1
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A n
ew
parti
al i
mage
e
ncry
ption met
hod
f
or
do
c
ument i
mag
es u
si
ng vari
an
ce b
as
e
d qua
d
...
(
C R Rev
anna
)
787
In
[
2], the
a
uthors p
rop
os
e
d
a
bit
-
le
vel sc
ram
bling al
gorith
m
s
to scr
am
ble b
it
posit
ions.
I
t i
s also s
ai
d
that
the
pro
pose
d
al
gorith
m
pr
ovides
flexibili
ty
to
sel
ect
any
i
m
ag
e
as
the
inpu
t
so
urce
i
m
age,
any
deco
m
po
sit
io
n
te
chn
i
qu
e
f
or
ob
ta
ini
ng
the
bi
t
plane,
a
ny
de
com
po
sed
bit
plane
a
s
the
s
ecur
it
y
key
bit
plan
e
and
any
sc
ram
bling
te
ch
niqu
e
fo
r
the
bit
-
l
evel
pe
rm
utatio
n.
I
n
[3
]
,
a
c
hao
ti
c
syst
em
base
d
pa
rtia
l
im
age
encr
y
ption
is
propose
d.
T
he
pro
pose
d
schem
e
includes
bit
pla
ne
deco
m
po
sit
io
n
of
s
ource
i
m
age.
Af
te
r decom
posing, the sign
if
ic
ant b
it
p
la
ne
are select
ed
for
en
crypti
on. E
ncr
y
ption
is
achieve
d
by g
e
ne
rati
ng
the pseu
dora
ndom
n
um
ber
se
qu
e
nce
us
i
ng c
hao
ti
c syst
em
.
In
[4
]
,
a
c
ha
otic
base
d
pa
rtia
l
gr
ey
sal
e
im
ag
e
encr
y
ption
is
propose
d.
It
is
ob
se
r
ved
t
hat
the
auth
or
s
pro
po
se
d
a
bi
t
plane
dec
om
po
sing
m
et
ho
d
f
or
enc
ry
ption.
T
he
va
rio
us
bit
p
la
ne
s
(signific
a
nt
an
d
insig
nificant)
a
re
ide
ntifie
d
ba
sed
on
a
uto
c
orrelat
ion
thre
s
ho
l
d
of
diff
e
re
nt
bina
ry
pla
ne
s.
The
key
se
qu
e
nce
ob
ta
ine
d
by
c
hao
ti
c
m
ap
is
us
e
d
to
enc
ry
pt
the
co
rr
el
at
ed
bit
pla
nes.
In
[
5],
bo
t
h
se
le
ct
ive
an
d
c
om
ple
te
i
m
age
encr
y
ption
us
i
ng
the
s
equ
e
nce
of
ch
aotic
m
ap.
At
first
the
c
ha
otic
m
ap
is
us
ed
to
ge
ner
at
e
a
key
t
o
com
plete
ly
encr
ypt
the
plai
n
im
age.
Sec
ond,
f
or
the
s
a
m
e
inp
ut
im
age
sel
ect
ive
portio
n
is
e
nc
rypted
.
Finall
y
the
co
m
ple
te
and
se
le
ct
ive
encr
ypt
ed
res
ults
are
com
bin
ed
by
XO
R
ope
rati
on
t
o
achie
ve
bette
r
secur
it
y.
A
ne
w
te
ch
nique
c
al
le
d
grap
h
c
olo
ri
ng
pro
blem
(G
CP
)
f
or
par
t
ia
l
encr
ypti
on
of
m
edical
i
m
age
i
s
pro
po
se
d
in
[
6]
.
The
GCP
te
chn
i
qu
e
is
use
d
to
sel
ect
the
op
ti
m
al
po
sit
ion
s
of
t
he
pixe
ls
fr
om
the
i
nput
m
edical
i
m
ag
e.
In
[
7],
resea
r
cher
s
pr
ese
nte
d
a
par
ti
al
i
m
a
ge
enc
ryptio
n
te
chn
i
qu
e
base
d
shuf
fling
the
pix
el
s
within a b
loc
k.
Pixels sh
uffli
ng
a
re
ac
hieve
d
base
d
on
the
se
qu
e
nce
of
c
ha
ot
ic
m
ap.
By
sel
ect
ing
v
arie
d
bl
ock
siz
e, d
at
a e
nc
r
ypti
on
quant
um
can
be
va
rie
d.
In
[8
]
,
a
par
ti
al
gr
ay
scal
e
enci
ph
e
rin
g
ba
sed
on
t
he
cha
otic
m
ap
is
propose
d.
T
he
gray
sca
le
i
m
age
is
deco
m
po
se
d
i
nto
ei
gh
t
bin
a
ry
pla
nes.
Scr
a
m
bling
is
do
ne
f
or
m
os
t
sign
ific
a
nt
bit
planes
.
T
he
C
hao
ti
c
seq
uen
ce
ge
ne
rated
by
the
S
kew
te
nt
m
ap
is
us
e
d
t
o
sc
ram
b
le
the
pla
nes.
The
scra
m
bled
bit
plan
es
are
encr
y
pted
to
obta
in
ci
pher
i
m
age.
In
[9
]
,
auth
or
s
pro
po
s
ed
a
trans
form
do
m
ai
n
approac
h
f
or
par
ti
al
i
m
age
encr
y
ption.
F
our
sub
-
ba
nds
a
re
ar
rive
d
at
by
app
ly
in
g
D
WT
for
t
he
in
pu
t
plain
im
age.
T
he
lo
w
fr
e
qu
e
ncy
su
b
-
band
is
e
ncr
y
pted
by
t
he
se
quence
ge
ner
at
e
d
by
t
he
L
ogist
ic
chao
ti
c
m
ap.
At
la
st
enc
ryp
te
d
lo
w
fr
e
qu
e
ncy
ba
nd
an
d
non
-
e
ncry
pted
high
f
re
qu
e
ncy
ba
nd
a
r
e
com
bin
ed
an
d
inv
e
rse
disc
r
et
e
wav
el
et
transfor
m
is
app
li
e
d
to
obta
in
the
ci
phe
r
im
age.
Discr
et
e
Cosine
T
ra
ns
f
or
m
(D
CT
)
base
d
par
ti
al
encr
y
ption
of
colo
r
i
m
age is p
r
opose
d
i
n
[
10]
.
The
DCT
is
ap
plied
f
or
the
in
pu
t
s
ource
im
a
ge
f
or
all
three p
la
nes
.
Th
en
t
he
enc
ryptio
n
i
s
perf
or
m
ed
by
the
sequ
e
nc
e
gen
e
rated
by
the
Lo
gisti
c
chao
ti
c
m
ap.
The
encr
ypte
d
pla
nes
ar
e
com
bine
d
an
d
inv
e
rse
DCT
is
app
li
ed
to
obta
in
ci
pher
im
age.
I
n
[
11]
,
the
authors
propose
d
an
enci
pherin
g
te
chn
i
que
based
on
tw
o
le
vel
s
of
pe
rm
utati
on
and
substi
tuti
on.
For
ever
y
input
source
im
age,
a
dynam
i
c
key
is
gen
er
at
ed
wh
ic
h
res
ults
in
the
ba
sis
of
e
nc
ryptio
n
proce
sses.
T
he
n,
a
nonlinea
r
S
-
bo
x
m
et
ho
d
is
us
ed
for
im
age
su
bs
ti
tuti
on
w
hich
is
fo
ll
owe
d
by
a
m
at
rix
m
ulti
p
li
cat
ion
wh
ic
h
is
per
f
or
m
ed
for
im
age
diff
us
io
n.
T
hese
t
wo
processes
finall
y
resu
lt
s
in
a
ci
pher
im
age.
In
[12],
pro
posed
a
te
chn
iq
ue
w
hich
c
om
bin
es
bo
th
par
ti
al
encr
y
ption
a
nd
i
m
age
com
pr
essio
n.
The
c
om
pr
essi
on
is
ac
hieve
d
by
t
he
qu
a
d
tree
an
d
SP
HI
T
im
age
com
pr
essi
on
te
ch
ni
qu
es
.
On
ly
13
-
27
%
of
the
qua
d
tre
e
com
pr
esse
d
data
a
nd
le
ss
t
han
2%
of
the
SPHIT
c
om
pr
e
ssed
data
is
pa
rtia
ll
y
encr
y
p
te
d
f
or
s
ecur
it
y.
I
n
[
13
]
an
ef
fici
ent
sel
ect
ive
i
m
age
encr
y
ption
te
c
hniq
ue
com
bin
ing
sa
w
to
oth
fi
ll
ing
,
sel
ect
ed
pix
el
s
,
non
-
li
near
chao
ti
c
m
ap
and
S
VD(si
ngular
va
lu
e
de
com
po
sit
ion
)
is
pr
op
os
ed
.
Im
age
scram
bling
is
done
us
i
ng
the
saw
too
th
fill
ing,
w
her
ea
s
si
gn
i
fica
nt
pix
el
s
are
sel
ect
ed
base
d
on
the
pi
xels
of
interest
.
Fi
nally
,
dif
fu
si
on
is
perform
ed
on
the
m
or
e
wei
ghte
d
pix
el
s
us
i
ng
cha
otic
m
a
p
a
nd
sin
gu
la
r
value
deco
m
po
sit
io
n
as
the
key.
A
su
bst
it
ution
bo
x
[S
-
box]
a
nd
l
inear
fr
act
io
nal
trans
f
or
m
te
chn
iq
ue
is
pro
pos
ed
in
[14]
f
or
par
ti
a
l
i
m
age
enc
ryption.
T
he
pro
po
s
ed
te
ch
nique
us
es
a
li
fting
wa
velet
tra
ns
f
or
m
in
fr
e
quenc
y
do
m
ai
n,
which
p
r
ovides a
se
nsi
ti
ve
inf
or
m
at
i
on that ca
n
be
encr
y
pted by the se
qu
e
nce
ge
ner
at
e
d
f
r
om
c
hao
ti
c
m
ap.
T
he
du
al
proces
s
of
c
onf
us
io
n
a
nd
diffusio
n
a
re
ca
r
ried
out
via
pe
rm
utati
on
,
di
ffusio
n
a
nd
s
ub
s
ti
tuti
on
process
.
In
[
15
]
,
a
par
ti
al
i
m
age
encr
ypt
ion
us
in
g
DC
T
and
li
ght
w
ei
gh
t
stream
t
echn
i
qu
e
is
propose
d.
By
con
sid
erin
g
the
basic fun
dam
ental
at
ta
cks
li
ke
sta
ti
sti
ca
l
at
ta
ck,
re
place
m
ent
at
ta
ck
and
d
if
fer
e
ntial
at
ta
ck,
DCT
C
oeffici
ents
based
tra
nsfo
rm
ation
te
c
hniq
ue
is
pro
posed.
I
n
[16],
a
par
ti
al
im
age
encr
y
ption
base
d
on
bit
plane
dec
om
po
sit
ion
is
pro
posed
.
T
he
in
put
sour
ce
i
m
age
is
s
egm
ented
int
o
ei
ght
bit
pl
anes.
The
n
the
s
ig
nif
ic
ant
bina
ry
bit
plane
is
c
on
si
der
e
d
for
e
ncr
y
ption.
Using
t
he
te
nt
ch
aotic
m
ap,
a
key
se
quence
is o
btained
for
encr
y
ption.
In
[
17
]
,
a
no
n
-
ada
ptive
par
ti
al
encr
y
ption
of
gray
scal
e
im
age
is
pe
rform
ed
us
i
ng
Ch
aotic
m
ap
is
pro
po
se
d.
T
he
input
source
i
m
a
ge
is
su
b
-
di
vid
e
d
into
ei
ght
bin
ary
pla
nes
.
By
the
Tent
m
ap
m
et
ho
d
a
pp
ly
in
g
ps
e
udor
a
ndom
seq
uen
ces
,
f
our
si
gn
i
ficant
bit
planes
a
re
encr
y
pted.
The
par
ti
al
i
m
age
encr
y
ption
is
obta
ined
by
scram
bling
us
in
g
non
sin
usoidal
wavel
et
s
is
pro
pose
d
in
[
18]
.
Kekre’
s
W
al
s
h
Se
que
nce
proce
dure
is
us
e
d
to
scram
ble
the
im
age
in
the
tra
nsfo
rm
do
m
ai
n
(
wav
e
le
t).
The
tra
ns
form
do
m
ai
n
helps
i
n
pr
e
ve
ntin
g
the
at
ta
cks
by stat
ist
ic
al
m
ea
ns
.
I
n
[19],
a
pa
rtia
l
encr
y
ptio
n
te
ch
ni
qu
e
is pro
po
se
d.
E
ncry
ption
i
n
both Sp
at
ia
l
and
t
rans
f
or
m
do
m
ai
n
has
be
en
de
pl
oyed
in
this
case.
T
he
bit
plane
de
com
po
sit
ion
te
chn
i
qu
e
is
us
e
d
f
or
encr
y
pting
the
i
m
age
in
spa
ti
al
do
m
ai
n.
At
la
st,
the
rati
o
of
enc
ryptio
n
ti
m
e
and
e
nc
od
i
ng
tim
e
is
cal
culat
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
1
,
Febr
uar
y
2020 :
78
6
-
800
788
to
il
lustrate
the
sp
eed
perfor
m
ance.
In
[20
]
,
a
DCT
bas
ed
pa
rtia
l
color
im
age
encry
ption
is
pres
ented.
The
DCT
is
us
e
d
to
sel
ect
the
si
gn
ific
a
nt
re
gions
i
n
three
di
ff
e
re
nt
pla
nes
fro
m
the
col
or
i
m
age.
The
n
the
sel
e
ct
ed
reg
i
ons
are
enc
rypted
or
diffuse
d
us
in
g
the
Arn
o
ld
c
hao
ti
c
m
ap
w
hich
res
ults
in
ci
ph
e
r
i
m
age.
Fr
om
the
li
te
ratur
e
r
evie
w,
th
ere
is
a
nee
d
t
o
enc
rypt
th
e
Do
c
um
ent
i
m
a
ges
par
ti
al
ly
to
synch
ronize
with
the
real
tim
e
app
li
cat
ion
s
a
nd
inc
r
ease
the
secu
rity
le
vel
fo
r
hand
hel
d
de
vices
an
d
ot
her
le
ss
com
pu
ta
ti
on
al
ly
capa
ble
ga
dg
et
s.
Wh
e
n
a
n
i
m
age
is
viewe
d
as
a
w
ho
le
w
it
h
bl
ock
s
,
t
he
blo
c
ks
with
un
ifor
m
pix
el
inte
ns
it
y
le
vels
(H
ist
ogr
a
m
)
exh
i
bit
le
ss
m
eaning
fu
l
i
nfor
m
at
ion
in
t
he
im
age
bu
t
bl
ock
s
with
un
e
qu
al
pix
el
intensit
y
le
vels
exh
i
bit
m
or
e
intel
l
igent
inf
or
m
at
ion
and
a
re
ref
e
rre
d
to
as
sig
nific
a
nt
reg
i
ons.
T
her
e
is
a n
ee
d
to
ide
ntify bloc
ks wit
h
un
e
qual
inten
sit
y l
evels. Th
e
pro
po
se
d
m
et
ho
d use
s a
dif
fere
nt app
ro
ac
h
t
o
fi
nd
the
sig
nifica
nt
re
gions
a
nd
e
ncr
y
pt
only
th
os
e
reg
i
ons.
F
or
i
de
ntifyi
ng
the
sig
nifica
nt
re
gions,
t
he
i
m
age
is
segm
ented
into
blo
c
ks
by
ta
ki
ng
va
riance
a
s
a
par
am
et
er
f
or
Q
ua
d
-
tre
e
Se
gm
entat
ion
te
chn
i
qu
e
.
T
his
m
et
hod
deco
m
po
ses
a
n
im
age
into
s
ign
ific
a
nt
an
d
insig
nificant
bl
ocks.
It
is
e
nough
t
o
e
ncr
y
pt
only
the
si
gn
i
fican
t
blo
c
ks
to
pa
rtia
ll
y
encr
ypt
an
i
m
age.
The
bl
ock
siz
e
is
a
pr
eset
value,
de
pe
nd
i
ng
on
the
r
equ
i
rem
ent
based
on
the
le
vel
of
se
cur
it
y.
The
part
ia
l
encr
ypti
on
is
per
f
or
m
ed
us
in
g
a
cha
otic
syst
e
m
.
The
pr
op
os
ed
wor
k
is
al
so
exten
ded f
or c
om
plete
en
crypti
on usi
ng m
i
xed cha
otic sy
s
tem
.
2.
RESEA
R
CH MET
HO
D
2.1.
Quad
t
ree se
g
ment
at
i
on
The
Qu
a
d
tr
ee
m
e
tho
d
is
ap
plied
f
or
the
d
ivisi
on
of
an
i
m
age
into
bl
ock
s/
re
gions
by
ap
plyi
ng
recursio
n
[
21]
.
The
pa
rtit
ion
e
d
bl
ock
s
a
re
ar
range
d
in
the
f
or
m
of
hiera
rc
hal
tree
struct
ure.
T
he
r
oo
t
blo
ck
i
s
cal
le
d
as
par
e
nt
blo
c
k
a
nd
th
e
pa
rtit
ion
e
d
bl
ock
s
are
cal
le
d
as
child
blo
c
ks
.
The
par
e
nt
bloc
k
is
se
gme
nted
into
four
(
quad
)
e
qu
al
ly
siz
ed
sub
-
blo
c
ks
an
d
eac
h
s
ub
-
bloc
k
is
s
ubje
ct
ed
to
a
te
st.
A
blo
ck
is
c
heck
e
d
to
s
e
e
if
the
crit
erio
n
fo
r
hom
og
ene
it
y
is
m
et
,
if
i
t
m
eet
s,
then
no
furthe
r
di
vision
is
m
ade
and
t
he
node
i
s
le
ft
undiv
i
ded
an
d
is
cal
le
d
as
a
le
af
no
de.
If
t
he
crit
erion
is
not
m
et
,
then
div
id
e
the
bl
oc
k
int
o
f
our
sub
blo
c
ks
or
reg
i
on
s
a
nd
a
pp
ly
the
te
st
crit
eria
again.
The
ab
ove
pr
ocedu
re
is
perfo
rm
ed
un
ti
l
each
sub
-
bl
ock
ob
ey
s
the
crit
eria.
T
he
refor
e
each
node/
par
e
nt
is
e
it
her
hav
e
no
c
hildr
e
n
or
has
four
c
hildr
e
n.
Hen
ce
the
qua
d
tree
deco
m
po
si
ng
t
echn
i
qu
e
par
ti
t
ion
s
the
im
age
into
sub
bl
oc
ks
or
re
gions
that
are
m
or
e
ho
m
og
e
neous
than
the im
age itse
l
f.
T
he
le
ast
bloc
k
siz
e is
var
ia
ble on
e a
nd
de
pends o
n
the
r
e
qu
i
rem
ents that suit t
he
a
pp
li
c
at
ion
.
The dec
om
po
sing l
ast
s whe
n Qu
a
d
t
ree r
ea
c
hes
it
s m
ini
m
um
size [2
2].
The
qu
a
d
tre
e
deco
m
po
sit
io
n
is
show
n
in
th
e
Fig
ur
e
1
with
a
tree
diag
ra
m
.
The
root
no
de
in
dicat
es
the
w
ho
le
im
a
ge,
this
node
is
par
ti
ti
on
e
d
int
o
eq
ually
siz
ed
four
s
ub
-
blo
c
ks
if
it
fail
s
to
sat
isfy
the
crit
eria
of
ho
m
og
e
neity
. T
he
le
af
no
de
i
nd
ic
at
es a
b
l
oc
k
sat
isfie
s t
he ho
m
og
e
neity
crite
ria.
Figure
1. (
a
) S
a
m
ple I
m
age,
(
b) Qua
d
T
ree
Deco
m
po
sit
io
n St
r
uctur
e
2.2.
Ima
ge Deco
m
po
s
iti
on
Crite
ri
a
The
qua
d
tree
deco
m
po
sit
io
n
is
m
ade
with
var
ia
nce
of
t
he
i
m
age
a
s
the
crit
eria.
The
var
ia
nce
of
a
blo
c
k
is
cal
culat
ed
by
co
m
pu
ti
ng
the
m
ean
of
t
he
pi
xels
in
that
bl
oc
k.
The
va
riance
e
qu
at
io
n
can
be
giv
e
n
as
(
)
=
1
×
∑
[
−
µ
(
)
]
2
=
1
.
(1)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A n
ew
parti
al i
mage
e
ncry
ption met
hod
f
or
do
c
ument i
mag
es u
si
ng vari
an
ce b
as
e
d qua
d
...
(
C R Rev
anna
)
789
w
he
re
µ
(
)
is t
he
m
ean
of that
bloc
k.
µ
(
)
=
1
×
[
∑
]
.
=
1
(
2)
The
var
ia
nce
of
the
par
e
nt
blo
ck
cal
culat
ed
first
an
d
t
he
va
riance
of
t
he
par
ti
ti
on
e
d
c
hi
ldre
n
bl
ocks
are
dete
rm
ined
ind
i
viduall
y.
Now
c
hec
k
f
or
the
va
riance
of
the
i
nd
i
vidual
blo
c
k,
i
f
it
is
great
er
tha
n
it
s
par
e
nt
blo
c
k,
th
en
de
com
po
se
the
c
hild
bl
oc
k
f
ur
t
her.
Ot
herwise
the
child
blo
c
k
is
le
ft
as
a
le
af
bl
ock
/
node
with
no
furthe
r
di
visio
n.
T
his
resu
lt
s
in
dec
om
po
sin
g
the
i
m
age
wi
th
une
qu
al
siz
e
par
ti
ti
oned
bl
ock
s
.
Fig
ure
2
is
an
exam
ple
of
a pi
ct
ur
e d
oc
um
e
nt
w
hich
is
dec
om
po
sed
i
nto
s
m
al
le
r
blo
cks
. A
4
×
4
dec
om
posed
Le
na
im
age
is
represe
nted
i
n Fi
gure
2(
a
)
.
Fi
gure
2(b
)
re
pre
sents the
corre
sp
on
ding m
apped
blo
c
ks
.
Figur
e
2. (a
) Q
uad Tree
D
ec
om
po
sit
ion
w
it
h
Bl
ock Size
4
×
4,
(b) Decom
po
se
d
Bl
oc
ks
m
apped
for
Le
na
im
age
2.3.
Chaotic
Map
2.3.1.
Skew
ten
t
m
ap
This
c
hao
ti
c
m
ap
is
one
dim
ension
al
in
natu
re.
It
i
s
al
so
cal
le
d
as
asy
m
m
et
ric
te
nt
m
ap.
Ma
them
a
ti
call
y, it
is g
ive
n b
y
+
1
=
(
)
∶
=
{
,
∈
[
0
,
]
,
1
−
1
−
,
∈
[
,
1
]
.
(3)
w
he
re
is co
ntr
ol p
a
ram
et
er r
ang
e
s fr
om
[
0,
1]
an
d
x
n
is t
he
s
ta
te
o
f
syst
em
w
hose
val
ue
ra
ng
e
s fr
om
[
0,
1]
.
At
= 0
.
5,
(
)
bec
om
es a r
eg
u
la
r
tent m
ap.
Furth
er
detai
ls can
be
h
a
d on this m
ap
in
[2
3]
.
2.3.2.
Be
rnou
ll
i ma
p
This
c
hao
ti
c
m
ap
is
al
so
one
dim
ension
a
l
in
nat
ur
e
.
G
ener
al
form
ula
for
Be
rno
ulli
m
ap
can
be
wr
it
te
n
as
foll
ows
+
1
=
(
)
1
ℎ
∈
[
0
,
1
]
.
(4
)
The
co
ntr
ol
pa
ram
et
er
of
Be
rno
ulli
m
ap
(i.e.,
b)
sho
uld
be
ta
ke
n
in
the
ra
ng
e
of
1
to
5
to
kee
p
chao
ti
c
be
hav
i
or [24,
25
]
.
F
or
=
2
,
E
qu
at
io
n
(4)
can
be writt
en a
s foll
ows
+
1
=
(
2
)
1
∶
=
{
2
,
∈
[
0
,
1
2
]
,
2
−
1
,
∈
[
1
2
,
1
]
.
(
5)
w
he
re
0
=
0.270
9
is t
a
ken as th
e init
ia
l value.
Fu
rt
her detai
ls
can
be had
on t
his m
ap
in [2
6]
.
2.3.3.
Henon m
ap
In
discrete
ti
m
e
dynam
ic
syst
e
m
s
Henon
m
ap
ex
hib
it
good
c
ha
otic
beh
a
vior.
It
ta
kes
the
poin
t
(
,
)
in t
he
s
pace
and m
aps
it
to a
n
e
w po
i
nt.
Ma
them
a
ti
call
y i
t can for
m
ulate
d
as
+
1
=
−
1
+
2
,
(6)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
1
,
Febr
uar
y
2020 :
78
6
-
800
790
+
1
=
.
(7)
The
init
ia
l
val
ue
0
∈
(
0,
1)
an
d
0
∈
(0
,
1)
ca
n
be
use
d
as
t
he
key
f
or
the
syst
em
(
0
,
0
)
.
T
he
He
non
m
ap
m
ai
nly
dep
en
ds
on
tw
o
par
a
m
et
ers
and
,
th
e
researc
h
res
ul
ts
sh
ow
that
t
he
value
for
=
1.4
a
nd
=
0.3
,
the H
e
non m
ap
ex
hib
it
s c
hao
ti
c n
at
ure
[27]
.
2.4.
Part
i
al
im
age
encryp
tio
n
The result
ant
de
com
po
sed
im
age
blo
c
ks
of
m
ini
m
u
m
size
(v
a
ria
ble)
a
re sc
ram
bled
us
in
g a Ske
w
te
nt
chao
ti
c m
ap.
T
he parti
al
en
c
r
ypti
on
is
p
e
rfo
rm
ed
detai
le
d belo
w.
1.
Gen
e
rate t
he
c
hao
ti
c se
quenc
es u
si
ng the
Skew tent m
ap
f
or eac
h bloc
k
si
ze.
2.
The
c
ha
otic se
qu
e
nce
ge
ner
at
ed fo
r
a
pa
rtic
ul
ar b
l
ock size
i
s u
se
d
t
o
c
onf
use
that
blo
c
k
as
foll
ow
s
.
a.
Convert
the
ge
ner
at
e
d
cha
otic
seq
uen
ces
f
or
the
blo
c
k
siz
e
into
na
tu
ral
nu
m
ber
s
by
m
ulti
plyi
ng
with
a
factor
of
10
1
5
(p
recisi
on
of
real
num
ber
s)
and
obta
in
a
uniq
ue
in
dex
va
lue
w
hose
ra
nge
li
es
withi
n
the b
l
ock size
(
Perfo
rm
Mod
ul
us
)
.
b.
Scram
ble/perm
ute
the
va
rio
us
pi
xels
within
eac
h
blo
c
k
base
d
o
n
the
ind
e
x
values
gen
e
rated
in
the pre
vious st
ep.
3.
The
ste
ps
1
a
nd
2
a
re r
e
peate
d for all
the
bl
ock size
s
gen
e
r
at
ed.
The
blo
c
k
dia
gr
am
of
t
he
pa
rtia
l
and
com
plete
enc
ryptio
n
is
s
how
n
i
n
Figure
3
(a).
The
res
ultant
par
ti
al
enc
ryption
is
ref
e
rr
e
d
as
the
fir
st
le
ve
l
confu
si
on.
T
he
se
qu
e
nce
of
act
ion
s
on
t
he
Do
c
um
ent
i
m
age
is
pictorial
ly
depi
ct
ed
in
Fig
ure
3
(
b)
.
T
he
par
ti
al
ly
encr
ypte
d
re
su
lt
s
f
or
the
Le
na
i
m
age
with
blo
c
k
siz
e
s
of
8
×
8,
16
×
16
a
nd
32
×
32 a
re sh
own
i
n
Fi
gure
4 (a), 4
(b) an
d 4 (c)
r
es
pecti
ve
ly
.
Figure
3 (a)
. B
lock
diag
ram
o
f propose
d
sc
hem
e
Figure
3 (b).
Fl
ow
diag
ram
o
f
prop
os
ed
doc
um
ent i
m
age seg
m
entat
ion
u
si
ng qua
d
tree
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A n
ew
parti
al i
mage
e
ncry
ption met
hod
f
or
do
c
ument i
mag
es u
si
ng vari
an
ce b
as
e
d qua
d
...
(
C R Rev
anna
)
791
Figure
4. Parti
al
ly
En
crypte
d Lena im
ages for
(a
)
Bl
oc
k
Si
ze eq
ual to
8
×
8, (
b)
Bl
ock Size
equal t
o
16
×
16
,
(c)
Bl
ock Size
equ
al
t
o 32
×
32,
(d)
Sec
ond l
ev
el
co
nf
us
ed
im
age
2.5.
Co
m
plete
im
age encr
yptio
n
The
propose
d
schem
e
can
be
furthe
r
exte
nded
f
or
c
om
pl
et
e
i
m
age
encry
ption
by
the
add
it
io
n
of
seco
nd
le
vel
conf
us
io
n
an
d
diffusio
n.
I
n
diffusio
n,
w
her
e
the
pix
el
v
al
ues
are
m
od
ifie
d
acco
rd
i
ng
to
the
se
qu
e
nce
gen
e
rated
by
t
he
c
hao
ti
c
m
ap
incl
ud
i
ng,
e
sta
blishin
g
i
nterd
e
pe
nd
e
ncy
betwee
n
t
he
pi
xels.
Diffusi
on is ca
rr
ie
d o
ut at t
he first a
nd seco
nd levels
.
2.5.1.
S
econ
d level c
on
f
usio
n
1.
Divid
e
t
he
pa
rtia
ll
y
encr
ypte
d
i
m
age
into
non
-
overlap
ping b
loc
ks
of
siz
e
equ
al
to
th
e
le
ast
le
vel
fo
r
wh
i
c
h
the p
a
rtia
ll
y encrypted
b
l
ock s
iz
e is co
ns
ide
r
ed.
2.
Gen
e
rate c
hao
t
ic
sequence
u
si
ng 2D
He
non m
ap
eq
ual to
num
ber
of
blo
c
ks.
3.
Perm
ute the b
l
ocks acc
ordin
g t
o
se
quence
g
e
ner
at
e
d by He
non m
ap.
Th
e
r
e
su
lt
ant m
at
rix
is ‘C’
.
Figure
4 (d)
de
picts t
he
sec
on
d
le
vel c
onf
us
e
d
im
age o
btai
ne
d
a
fter the
part
ia
l encryptio
n.
2.5.2.
Fir
st
le
vel
d
iff
usion
1.
Gen
e
rate t
he
c
hao
ti
c se
quenc
e u
si
ng the Be
r
noulli
’s
m
ap
of
size [
1,
M
×
N].
2.
The
ge
ner
at
e
d
seq
uen
ce
is
c
onve
rted
into
in
te
ger
by
m
ultip
ly
ing
with
a
f
act
or
of
10
15
a
nd
m
od
ulus
it
with
255. A
rr
a
nge t
his se
qu
e
nce
of inte
ger
s
in or
der to
obta
in th
e m
a
trix ‘A’
of
size M
×
N.
3.
The
m
at
rix
‘B’
w
hich
is
t
he
resu
lt
of
first
le
vel
of
dif
f
usi
on
is
a
rr
i
ved
by
X
ORin
g
the
pi
xels
of
t
he
par
ti
al
ly
encr
ypte
d
im
age
(f
r
om
Sect
ion
2.4)
with
the
co
r
respo
nd
i
ng
pixe
ls
(ele
m
ents)
of
the
m
at
rix
‘A
’
ob
ta
ine
d
i
n
t
he
previ
ou
s
step
.
2.5.3.
Second l
evel
d
iffus
i
on
To
est
a
blish m
or
e
inter
depen
den
c
y
betwee
n t
he neig
hbori
ng
pix
el
s
1.
Convert t
he
i
nput im
age m
at
r
ix I o
f
siz
e M
×
N
int
o
a
n
a
rr
ay
.
Wh
e
re
n= 1,
2,
3……….
M
×
N.
2.
Esta
blish
for
w
ard r
ow w
ise
inter
dep
e
ndenc
y as sh
own
Fig
ur
e
5 su
c
h
t
hat
′
∶
=
{
,
=
0
,
⊕
−
1
′
,
=
1
,
2
,
3
…
.
.
×
N
.
(8)
3.
Esta
blish
bac
kwar
d ro
w wise
interde
pe
nd
e
nc
y as sh
own
Fig
ur
e
5 su
c
h
t
hat
′′
∶
=
{
′
,
=
×
N
,
′
⊕
+
1
′′
,
0
<
<
×
N
.
(9)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
1
,
Febr
uar
y
2020 :
78
6
-
800
792
4.
Arrange
′′
into
a
m
at
rix
J
of
s
iz
e
M
×
N,
a
rr
a
nge
el
em
ents
of
J
i
nto
a
n
ar
ra
y
′′′
usi
ng
c
olum
n
wise
pro
gr
essi
ve
sca
n
m
et
ho
d.
5.
Esta
blish c
olum
n
wise p
ixel i
nter de
pende
nc
y as sh
own
Fig
ur
e
5 su
c
h
t
hat
′′′
∶
=
{
′′
,
=
0
,
′
′
⊕
−
1
′′′
,
=
1
,
2
,
3
…
.
.
×
N
.
(10)
6.
Arrange
′′′
into
a
m
a
trix ‘D’
o
f
size
M
×
N.
7.
The
sec
ond
le
vel
dif
fu
s
ed
i
m
age
is
the
re
su
lt
ant
m
at
rix
‘E’
(Ciph
e
r
I
m
age
as
show
n
in
Fi
gure
6(
b)
)
ob
ta
ine
d by
X
ORin
g
the
f
irst
level di
ffuse
d im
age ‘
B’
with
m
a
trix ‘D’
obt
ai
ned
i
n
ste
p 6.
Figure
5. Inter
dep
e
ndency
m
at
rix gen
e
rati
on
Figure
6. (a
)
I
nput
plain le
na
i
m
age,
(
b) Co
m
ple
te
ci
ph
e
r
im
age
2.6.
Co
m
plete
im
age decr
yptio
n
The
dec
ryptio
n
is
the
re
verse
process
of
e
nc
ryptio
n.
To
de
ci
ph
e
r
the
e
ncry
pted
im
age
th
e
f
ollow
i
ng
ste
ps
a
re follo
wed.
1.
The
inter
depen
den
cy
m
at
rix
‘D
’
w
hich
is
sha
red
(
by
sen
de
r)
is
XO
Re
d
w
it
h
the
ci
ph
er
i
m
age/
m
at
rix
‘E’
to obtai
n
t
he first l
evel
diffus
ed
im
age ‘
B’.
2.
Gen
e
rate
the
m
at
rix
‘
A’
us
i
ng
Be
r
noulli
’s
m
ap
with
the
sam
e
init
ia
l
c
onditi
ons
a
nd
con
t
ro
l
pa
ram
e
te
rs
us
e
d
in
en
c
rypt
ion
(
As
in
Step
1
a
nd
2 of Sec
ti
on
2.5.2)
.
3.
The
sec
ond l
ev
el
co
nf
us
ed
im
age is
obta
ined
b
y
perform
ing
XOR o
f
m
at
rix ‘C’
with
‘B’.
4.
Gen
e
rate
the
Chaotic
seq
ue
nce
usi
ng
2D
Heno
n
m
ap
with
the
sam
e
init
ia
l
con
dit
ion
s
a
nd
co
nt
ro
l
par
am
et
ers
to
per
m
ute
the
no
n
-
overlap
ping
blo
c
ks
ob
ta
ine
d
at
the
sec
ond
le
vel
c
onfus
ed
im
age
and
t
hus
ob
ta
in
the
pa
rtia
ll
y encrypted
i
m
age.
5.
Gen
e
rate
the
chao
ti
c
se
que
nce
us
in
g
Skew
te
nt
m
ap
with
the
sam
e
init
ia
l
condi
ti
on
s
a
nd
c
ont
ro
l
par
am
et
ers
as
in
sect
ion
2.4
an
d
per
m
ute
the
pix
el
s
within
the
bl
oc
ks
ge
ner
at
e
d
ou
t
of
Q
ua
d
Tree
Deco
m
po
sit
io
n. T
he
re
su
lt
ant
i
m
age o
btai
ned is the
Or
i
gin
al
Plai
n
im
age.
3.
RESU
LT
S
A
ND AN
ALYSIS
To
determ
ine
how
m
uch
eff
ic
ie
nt
t
he
pro
po
se
d
e
nc
r
ypti
on
m
e
thod
to
pro
vi
de
the
sec
ur
it
y,
the
pe
rfo
rm
ance
analy
sis
wa
s
dev
el
op
e
d
i
n
MATAL
AB
R
2014a
s
of
t
ware
us
i
ng
a
La
pt
op
ha
ving
4G
B
RAM
and
80GB
Hard d
isc
. Th
e
sim
ulati
on
ou
t
pu
ts
o
f
t
he
pro
pose
d
al
gorithm
r
eveals t
hat v
a
rio
us
gray
scale
im
ages
of
diff
e
ren
t
si
zes
(
51
2
×
512
and
256
×
256)
are
fed
as
the
input
plain
im
age.
T
he
i
niti
al
co
nd
it
io
ns
a
nd
the
co
ntr
ol
pa
ram
et
ers
us
ed
are
as
sho
wn
in
Table
1.
T
he
var
io
us
blo
c
k
siz
es
us
e
d
a
re
8
×
8
,
16
×
16
a
nd
32
×
32
.
Table
1.
Dif
fere
nt m
aps
with
con
t
ro
l
par
am
et
ers
an
d
i
niti
al
c
onditi
ons
us
e
d
as
K
ey
s
MAP
SUB KE
YS
CONTRO
L
PAR
A
ME
T
ERS
INITIA
L
CON
DI
TI
ON
S
Sk
ew T
en
t
Map
k
1
a =
0
.5
x
0
=
0
.1
Bern
o
u
lli
m
ap
K
2
b
=
2
x
0
=
0
.2705
Hen
o
n
M
ap
k
3
a =
1
.4
x
0
=
0
.6
3
1
5
4
7
7
b
=
0
.3
y
0
=
0
.18
9
0
6
3
4
3
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A n
ew
parti
al i
mage
e
ncry
ption met
hod
f
or
do
c
ument i
mag
es u
si
ng vari
an
ce b
as
e
d qua
d
...
(
C R Rev
anna
)
793
3.1.
M
ea
n
sq
uare
error
an
d
pe
ak sign
al
to n
oise ra
tio
The
m
ean
s
quare
e
rror
is
no
t
hing
but
the
di
ff
e
ren
ces
in
int
ensity
le
vels
of
pix
el
s
betwe
en
i
nput
a
nd
ou
t
pu
t
im
age
r
epr
ese
ntin
g
the
noise
le
vel.
F
or
a
n
i
deall
y
com
plete
ly
encr
ypte
d
ci
phe
r
im
age,
the
MS
E
valu
e
is m
or
e b
ut
f
or the
par
ti
al
ly
en
cry
pted
im
age,
it
s v
al
ue
is
m
od
erate. MS
E is cal
culat
ed
by u
si
ng the
equati
on
=
∑
∑
[
(
,
)
−
ℎ
(
,
)
]
2
=
1
=
1
×
.
(11)
The
par
ti
al
e
nc
ryptio
n
is
al
ways
no
t
sec
ur
e
beca
us
e
only
the
c
orrel
at
ed
pa
rt
of
the
im
age
is
encr
y
pted
an
d
the
rem
a
ining
pa
rt
is
le
ft
un
e
ncr
y
pted.
The
am
ou
nt
of
im
age
e
ncr
y
pted
dete
r
m
ines
the
co
nf
i
den
ti
al
it
y
le
vel.
A
bette
r
co
nf
i
de
ntial
it
y
le
vel
is
ob
ta
ine
d
wh
e
n
a
m
ini
m
u
m
of
12.
5%
of
data
encr
y
pted [1
]
.
T
he pr
opose
d
m
et
ho
d
yi
el
ds
60%
of
e
nc
ryp
ti
on
for
a
bloc
k si
ze o
f 8×8
.
% of
enc
rypti
on
=
.
(
12)
The
PS
NR
of
the
e
nc
rypte
d
docum
ent
im
age
sho
uld
be
le
s
s
tha
n
30dB
i
n
orde
r
to
pr
e
ven
t
the
interce
ptor
from
extracti
ng
the
plain
i
m
age
out
of
t
he
no
is
e
prese
nt
in
t
he
ci
phe
r
im
age.
Ou
r
m
et
ho
d
yi
el
ds
a PSNR
of 8.8
925dB.
P
SN
R i
s
giv
e
n b
y
=
10
10
(
255
2
)
.
(13)
3.2.
Histogr
am
Histo
gr
am
is
a
pictorial
way
of
de
picti
ng
pi
xel
distri
bu
ti
on
of
var
yi
ng
i
ntensity
le
vels.
It
is
pl
otted
as,
the
total
nu
m
ber
of
pi
xels
with
va
ryi
ng
i
ntensity
al
ong
the
y
-
axis
a
nd
the
dif
fer
e
nt
in
te
ns
it
y
le
vels
al
ong
the
x
-
axis
.
T
he
histo
gr
am
of
p
la
in
an
d
ci
phe
r
i
m
age
are
s
ho
wn
in
Fi
gure 7
.
For
a
c
om
pletely
encr
ypte
d
i
m
age
the
histo
gr
am
i
s
flat
,
bu
t
the
pa
rtia
ll
y
encr
ypte
d
i
m
age
has
sp
ikes
in
it
and
is
sa
m
e
as
that
of
the
plain
i
m
age
since,
only
th
e
pe
rm
utati
on
of
the
pi
xels
t
akes
place.
T
his
dep
ic
ts
t
ha
t
an
y
kind
of
sta
ti
sti
cal
attac
k
on
the co
m
plete
ly en
c
rypted
im
a
ge
is i
m
po
ssibl
e but f
or the
pa
rtia
ll
y encrypte
d
im
age it
m
ay
be p
os
sible.
F
igure
7. Histo
gr
am
o
f (a)
Pla
in im
age,
(b)
C
iph
e
r
im
age (
c
om
plete
en
crypti
on)
3.3.
Co
rrel
at
i
on
A
cl
early
visi
bl
e
i
m
age
with
pro
per
br
i
gh
t
ne
ss
has
it
s
co
r
rela
ti
on
c
oeffi
ci
ent
eq
ual
to
on
e
,
bu
t
for
the
ci
phere
d
i
m
age
it
has
a
sig
nificantl
y
reduce
d
value
(al
m
os
t
equ
al
to
ze
ro).
A
n
encr
y
ption
al
gorithm
gen
e
rates
a
ci
ph
e
re
d
im
age
with
rand
om
l
y
distrib
ute
d
pix
el
s
of
dif
fer
e
nt
inte
ns
it
ie
s
and
ha
s
it
s
c
orr
el
at
io
n
coeffic
ie
nt
bet
ween
a
djacent
pix
el
s
cl
os
e
to
zero.
A
set
of
4000
pairs
of
t
wo
a
dj
a
cent
pi
xels
in
al
l
dire
ct
ion
s
(hor
iz
on
ta
l,
ve
rtic
al
and
diag
on
al
)
we
re
ra
ndom
ly
sel
ect
e
d
to
determ
ine
th
e
co
rr
el
at
ion
c
oe
ff
ic
ie
nt
from
the p
la
in
a
nd ciph
e
re
d
im
age.
The
c
orrelat
ion coe
ff
ic
ie
nt is
giv
e
n by
=
(
,
)
√
(
)
√
(
)
.
(
14)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
1
,
Febr
uar
y
2020 :
78
6
-
800
794
w
he
re
(
,
)
is t
he
C
ov
a
riance
b
et
w
een
x
a
nd y. it
is give
n by
(
,
)
=
1
×
∑
[
(
−
µ
(
)
)
(
−
µ
(
)
)
]
=
1
.
(15)
w
he
re
x
a
nd y
are tw
o
a
djace
nt p
i
xels
value
s in
t
he
im
age,
V
(
x
)
is t
he
v
a
r
ia
nce of varia
bl
e x
a
nd is
giv
e
n by
(
)
=
1
×
∑
[
−
µ
(
)
]
2
.
=
1
(16)
µ
(
)
is t
he
m
ean of varia
ble x.
µ
(
)
=
1
×
∑
=
1
.
(17)
The
c
orrelat
io
n
co
ef
fici
ent
for
the
c
om
plete
ly
ci
ph
ere
d
i
m
age
was
f
ound
cl
os
e
t
o
zero
i
n
al
l
the
directi
ons
and
a
value
between
0
and
1
for
the
par
ti
al
ly
encr
yp
te
d
im
age.
This
represe
nts
that
the
al
gorithm
i
s
resist
ant
agai
ns
t
sta
ti
sti
cal
a
tt
acks.
The
c
orrelat
ion
coe
ff
ic
ie
nts
for
Plai
n
and
ci
phe
r
i
m
a
ge
i
n
three
dif
fer
e
nt
directi
ons a
re show
n
in
Fig
ure
8
.
Figure
8
.
Co
rr
e
la
ti
on
:
(a
-
c)
C
orrelat
ion f
or Pl
ai
n
im
age in
th
ree
diff
e
ren
t
d
i
recti
on
s
(H
or
iz
on
ta
l,
Ve
rtic
al
an
d
Diag
on
al
),
(d
-
f)
C
orrelat
ion f
or Ci
ph
e
r
im
age b
loc
k wit
h si
ze 8×
8
in
th
ree
d
if
fer
e
nt
direc
ti
on
s
(Ho
rizo
ntal,
Ver
ti
cal
and
D
ia
gonal), (
g
-
i)
Correl
at
ion
f
or Ci
ph
e
r
im
age b
l
oc
k wit
h
siz
e
16×16 i
n
in
th
ree
diff
e
ren
t
directi
ons
(Hori
zon
ta
l,
Ver
ti
c
al
an
d Dia
gona
l) and
(j
-
l)
Co
r
relat
ion
for
C
om
ple
te
En
cry
pted
Ci
pher
im
a
ge
in
three
dif
fer
e
nt
directi
ons
(Hori
zon
ta
l,
Ver
ti
c
al
an
d Dia
gona
l)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A n
ew
parti
al i
mage
e
ncry
ption met
hod
f
or
do
c
ument i
mag
es u
si
ng vari
an
ce b
as
e
d qua
d
...
(
C R Rev
anna
)
795
3.4.
Key
sp
ace
ana
lysis
It’s
a
m
easur
e
of
th
e
total
diff
ere
nt
keys
that
are
us
e
d
in
the
enc
ry
ption
a
nd
de
cryp
ti
on
process
.
The
di
ff
e
ren
t
s
ecret
key
K
is
a
com
bin
at
ion
of
s
ub
keys
k
1,
k
2
an
d
k
3
w
hic
h
are
ta
ke
n
f
rom
the
var
io
us
m
aps
with
their
init
ia
l
con
diti
ons
a
nd
the
syst
em
par
am
et
ers
us
ed
in
the
s
yst
em
as
sh
ow
n
in
the
Table
1.
T
he
sub
keys
k
1,
k
2
a
nd
k
3
are
ext
r
act
ed
f
r
om
Sk
ew
Te
nt
Ma
p
,
Be
rno
ulli
’s
Ma
p
a
nd
He
non
Ma
p
res
pe
ct
ively
.
Con
si
der
the
preci
sion
ta
ke
n
for
th
e
keys
is
of
the
order
of
10
-
15
.
T
hen
t
he
t
otal
key
s
pa
ce
us
e
d
is
(
10
14
)
z
,
wh
e
re
z
is
t
he
nu
m
ber
of
init
i
al
co
nd
it
io
ns
a
nd
syst
em
par
a
m
et
ers
us
e
d
i
n
al
l
the
s
ub
key
s
(z=
8).
T
his
r
esults
in
a
ve
ry
la
r
ge
key
s
pace
an
d
is
s
uffici
ently
la
rg
e
to
resi
st
the
at
ta
cks.
Fo
r
an
enc
rypt
ion
sc
hem
e
based
on
chaos,
the
key
sp
ace
shou
l
d
be
great
er
tha
n
the
2
200
≈10
30
[28]
to
withsta
nd
or
re
sist
ant
with
the
brute
-
force
at
ta
ck.
T
he pr
opose
d m
et
ho
d y
ie
lds a k
ey
siz
e of
10
112
wh
ic
h
is
ver
y m
uch great
er tha
n 1
0
30
.
3.5.
Key
sensiti
vit
y test
An
enc
ryptio
n
al
gorithm
is
sai
d
to
be
go
od
if
it
produces
a
c
om
plet
ely
diff
e
re
nt
plain
im
age
f
or
a
ti
ny
cha
ng
e
in
the
key
K,
with
ou
t
m
aking
any
c
ha
nges
in
t
he
ci
ph
er
ed
im
age.
A
plain
im
age
of
siz
e
512
×
512
is
sub
j
ect
ed
to
an
e
nc
ryptio
n
key
K
=
{k
1
,
k
2
,
k
3
}
wh
e
re
k
1
,
k
2
a
nd
k
3
are
the
s
ub
keys.
T
he
c
orrect
key
val
ue
K=
K
1
=
{
0.5
,
0.1,
2,
0.2
705,
1.4
,
0.631
5477,
0.3
,
0.1
8906
343}
for
wh
ic
h
th
e
decr
y
pted
im
a
ge
is
a
s
sh
ow
n
in
Fig
ure
9a
.
It
is
ob
serv
e
d
t
hat
the
decip
her
e
d
i
m
age
is
exactl
y
sa
m
e
as
that
of
the
plain
i
m
age.
Fo
r
a
sm
al
l
chan
ge
in
the
key
K=
K
2
=
{0.5,
0.1,
2,
0.2
705,
1.9
,
0.6
3154
77,
0.6,
0.1
8906
343}
,
K
=
K
3
=
{0.5,
0.1,
5,
0.1
,
1.
4,
0.631
5477,
0.3,
0.1
8906
343}
a
nd
K=
K
4
=
{0
.3,
0.2,
2,
0.270
5,
1.4,
0.6
3154
77,
0.3
,
0.189
06343}
f
or
w
hich
t
he
de
crypted
im
ages
are
e
ntirel
y
diff
e
re
nt
from
that
of
the
plain
im
age
is
sh
own
i
n
Fi
gure
9b,
9c
and
9d
res
pecti
vely
.
A
ti
ny
c
hange
i
n
the
va
lue
of
key
produces
a
c
om
plete
ly
diff
ere
nt
pla
i
n
i
m
age f
or th
e s
a
m
e inp
ut
ci
ph
er im
age.
He
nc
e the alg
ori
thm
is k
ey
se
ns
it
iv
e.
Figure
9
.
Dec
r
ypti
on
:
(a)
Ci
pher
I
m
age D
ec
rypted
w
it
h C
orrect
Key K
1
,
(b
-
d) D
ec
rypte
d usi
ng wron
g Keys
[(b) wit
h K
2,
(c
)
wit
h K
3,
an
d (
d) w
it
h K
4
]
3.6.
NPCR
A
ND
UACI
An
interce
pt
or
can
m
ake
a
ti
ny
m
od
ific
at
ion
in
the
pla
in
i
m
age
and
ob
se
rv
e
the
changes
in
the
outp
ut
ci
phered
im
age.
By
do
i
ng
s
o,
t
he
sig
nificant
relat
ion
s
hip
b
et
w
een
the
in
put
a
nd
outp
ut
im
ages
ca
n
be
obse
rv
e
d.
I
f
it
is
ob
se
rv
e
d
that
sign
ific
a
nt
changes
ta
ke
n
place
in
t
he
ci
ph
e
red
im
age
for
a
ti
ny
ch
ang
e
i
n
the
plain
im
a
ge,
it
m
eans
that
the
interc
epto
r
can
e
xtr
act
the
key
us
ed
f
or
t
he
e
ncr
y
ption
a
nd
crack
the
al
gorithm
.
To
c
heck
f
or
t
he
ef
fici
ency
of
this
al
gorith
m
two
pa
ram
eter
s
are
cal
c
ulate
d.
T
hey
are
NP
CR
and UACI
. T
he
ideal v
al
ue fo
r
N
PCR
a
nd UACI a
re
100 an
d 33.33
res
pecti
vely
[
29]
.
NP
CR
=
∑
(
,
)
,
×
×
100
.
(18)
w
he
re M a
nd
N
a
re th
e
w
i
dth
a
nd h
ei
gh
t
of
the im
age.
(
,
)
Ca
n be
def
i
ned as
(
,
)
∶
=
{
1
,
1
(
,
)
≠
2
(
,
)
,
0
,
1
(
,
)
=
2
(
,
)
.
w
he
re
1
(
,
)
Gr
ey
value
of
ci
ph
e
r
im
age an
d
2
(
,
)
Grey
v
al
ue of new ci
pher
im
age
UA
C
I
=
1
×
×
∑
(
1
(
,
)
−
2
(
,
)
)
255
,
×
100.
(19)
The
res
ults
of
t
he
ab
ov
e
te
sts
for
both
par
ti
al
and
c
om
plete
are
show
n
in
t
he
Table
2
a
nd
Table
3.
T
he
r
esults
rev
eal
t
hat the
al
gorithm
is b
et
te
r
than
the e
xi
st
ing
m
et
ho
ds
.
Evaluation Warning : The document was created with Spire.PDF for Python.