TELKOM
NIKA
, Vol. 11, No. 6, June 20
13, pp. 3445
~
3
450
e-ISSN: 2087
-278X
3445
Re
cei
v
ed
Jan
uary 25, 201
3
;
Revi
sed Ap
ril 21, 2013; Accepted Ap
ril 29, 2013
Cotton Pests and Diseases Detection based o
n
Image
Processing
Qinghai He
1
, B
e
n
x
ue
M
a
*
1
, Duan
y
a
ng
Qu
1
, Qiang Zhang
1
,Xinmin Hou
2
,Jing Zhao
3
1
Colle
ge of Me
chan
ical a
nd El
ectrical En
gin
e
erin
g, Shih
ezi Univers
i
t
y
,
Sh
i
hezi 8
3
2
003,C
h
in
a
2
T
he
T
eaching Office of the U
n
iver
sit
y
of the
Broadc
ast and
T
e
levision
in Xinjiang, Shi
hez
i 832000,China
3
School of el
ec
trical an
d electr
onic e
ngi
ne
erin
g, Shand
on
g U
n
iversit
y
of T
e
chno
log
y
, Z
i
b
o
255
04
9, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: mbx_s
h
z@
1
63.com
A
b
st
r
a
ct
Extract the d
a
m
a
g
e
d
i
m
age
form the c
o
tton
i
m
ag
e
i
n
orde
r to
meas
ure
the
da
ma
ge r
a
t
i
o of t
h
e
co
tto
n
l
e
a
f
wh
ich
ca
u
s
e
d
b
y
th
e
d
i
se
a
s
e
s
or p
e
s
ts. Se
ve
ra
l
al
g
o
r
i
t
hm
s like
im
ag
e
e
n
h
a
n
c
eme
n
t, im
age
filterin
g w
h
ich suit for cotton leaf
process
i
n
g
w
e
re explore
d
in this pap
er. T
h
ree differe
nt color mod
e
ls fo
r
extracting th
e
da
ma
ged
i
m
a
g
e
from cotton
l
eaf i
m
a
ges
w
e
re i
m
pl
e
m
e
n
te
d, na
me
ly RG
B color
mod
e
l,
HSI
color
mo
del, a
nd YCb
C
r col
o
r mo
de
l. T
he ratio of da
ma
ge (
γ
) w
a
s chosen as fe
atur
e to me
asure
th
e
degr
ee of d
a
m
a
ge w
h
ic
h cause
d
by dis
e
ases or p
e
st
s. T
h
is pap
er al
so show
s the
comparis
on
of the
results o
b
tain
e
d
by the i
m
ple
m
e
n
tin
g
in
different
co
lor
mo
dels, the c
o
mparis
on of res
u
lts show
s go
o
d
accuracy i
n
bot
h color
mo
dels
and YCb
C
r co
lor spac
e is co
nsid
ered as th
e best color
mode
l for extracting
the da
ma
ge
d i
m
a
ge.
Ke
y
w
ords
: cotton pests an
d dise
ases, color
mod
e
l,
i
m
a
ge
process
i
ng, rati
o of da
mag
e
.
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
There a
r
e m
any ha
rmful
biologi
cal
ha
zard
s like di
seases, p
e
st
s
and oth
e
rs h
appe
ned
throug
hout th
e crop
gro
w
i
ng pe
riod
s
which
ca
used
nume
r
ou
s lo
sse
s
ea
ch
yea
r
, so th
e
correct
identificatio
n
and a
c
cu
rate
analysi
s
of
h
a
rm
wa
s t
he
first thing i
n
t
he p
r
o
c
e
ssin
g
of preventi
o
n
and control in
orde
r to achi
eve real
-tim
e, accurate p
r
e
d
iction a
nd control.
The leaf i
s
th
e vital orga
ns of crop p
hot
osynthe
s
i
s
, a
nd al
so the
pl
ace
wh
ere th
e pe
sts
and di
sea
s
e
s
occur fre
que
ntly. Yutaka Sasa
kin jud
g
ed the ca
use
of crop
s thro
ugh the an
alysis
of the refle
c
ti
on spe
c
tral
curve of the
n
o
rmal
pa
rt an
d the di
se
ase
s
pot
part [1,
2]. Chen
Ji
aj
uan
detecte
d the
cotton l
eave
s
hole
a
nd
da
maged
e
dge
s ba
sed
o
n
th
e comp
uter vision,
whi
c
h
can
automatically determi
ne the
extent of
cotton pe
sts from
the surfa
c
e [
3
].
In contra
st to the approa
ches
me
ntion above, this re
sea
r
ch attem
p
ts to focuse
d on the
cotton im
age
whi
c
h d
a
ma
ged by the
d
i
sea
s
e
s
o
r
p
e
sts, the
Ma
chin
e Visio
n
techni
que
s a
nd
Image Proce
ssi
ng m
e
thod
s
were expl
ored in
this p
a
p
e
r in
o
r
de
r to
the autom
atically dete
c
tion
of
the cotton im
age.
2. Image Preproces
sing
The
effect o
f
image
seg
m
entation
di
rectly
affects
the perfo
rmance of
th
e
target
recognition [4]. Before the image processing, At
first the damaged
image shoul
d be cut off, th
is
prep
ro
ce
ssin
g ca
n redu
ce
the influen
ce
made
by
the
backg
ro
und.
A polygon th
at simila
r to t
he
edge
s of the leaves
wa
s u
s
ed to cut the dam
ag
ed i
m
age. The
seco
nd prep
ro
ce
ssi
ng is im
age
enha
ncement
whi
c
h ma
de
the image
more
suitabl
e for ba
ckground
sep
a
rated and fe
atu
r
e
extract, mean
while it ca
n el
iminate the in
fl
uence whi
c
h
made by the outsid
e
factors [5].
Image enh
an
ceme
nt app
ro
ach
e
s fall int
o
two bro
ad categori
e
s: sp
atial domain
method
s
and frequ
en
cy domain
me
thods [6]. T
h
e term
sp
atia
l domai
n refe
rs to
the ima
ge pla
ne it
se
lf,
and
app
roa
c
hes in thi
s
category
are
based
on
direct m
anipul
a
t
ion of pixel
s
in a
n
im
ag
e.
Freq
uen
cy d
o
main p
r
o
c
e
s
sing te
chni
qu
es are ba
sed
on modifying the Fouri
e
r transfo
rm of
an
image. T
here
is
no
gen
eral
theory
of ima
ge e
nha
ncem
ent. As fo
r th
e imag
e p
r
o
c
essing
ma
chi
ne
perceptio
n, the be
st ima
ge p
r
o
c
e
ssi
n
g
meth
o
d
would b
e
the
yielding th
e
best
machi
ne
recognitio
n
re
sults.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 6, June 201
3 : 3445 – 34
50
3446
As indi
cate
d
previously,
the term
spa
t
ia
l domai
n
refers to th
e
aggregate
of
pixels
comp
osi
ng
a
n
imag
e. Sp
atial dom
ain
method
s a
r
e
pro
c
e
dures
that ope
rate
dire
ctly on th
ese
pixels. Spatia
l domain p
r
o
c
esse
s w
ill be
denote
d
by the expre
ssi
on:
y
,
x
f
T
y
,
x
g
(1)
Whe
r
e f(x,y) is the input im
age, g(x,y) is
the pro
c
e
s
se
d image a
nd
T is an op
era
t
or on f,
defined
over som
e
n
e
igh
borh
ood
of
(x,y). T
he cotton imag
e af
ter en
han
ce
ment is sho
w
n i
n
Figure 1, it easy to se
pa
rate and extraction th
e foliage shap
e a
fter the pro
c
essing of im
age
enha
ncement
(gray-scal
e
t
r
an
sform
a
tion
), but a lo
t of
informatio
n
about the
col
o
r of di
sea
s
e
s
and pe
sts
we
re lost.
Gra
y
image
Enhancement image
Figure 1. The
Cotton Leaf Image after E
nhan
cem
ent
Histo
g
ra
ms
are th
e ba
si
s techniq
u
e
s
fo
r the n
u
m
ero
u
s
sp
atial dom
ain p
r
oce
s
sing.
Histo
g
ra
ms
manipul
ation
is ve
ry pop
ular i
n
ima
g
e
enh
an
cem
ent. It provi
de u
s
eful im
age
statistics
abo
ut image, m
e
anwhile it’s u
s
eful
in
othe
r image
proce
ssi
ng a
ppli
c
a
t
ions,
such a
s
image com
p
ression and segmentatio
n.
In
Figure
2,
the origi
nal i
m
age
cont
ra
st is hi
ghe
r, the
distin
ction b
e
t
ween th
e pro
s
pe
ct an
d the
backg
rou
nd i
m
age i
s
la
rge
r
, it’s ea
sy to segm
entation
,
and it p
r
ovid
e the th
re
sh
old for se
gm
entation.
Ima
ge hi
stog
ram
equali
z
atio
n
pro
c
e
s
sing
wa
s
use
d
in this
pape
r to increase the con
t
rast
of the i
m
age, streng
then
the ed
g
e
s of ba
ckground
image be
ca
u
s
e the ed
ge o
f
backgro
und
image is d
a
rker.
The original imag
e
Image after
histogram
equalization process
Histogram of the
original
image
Histogram of the
image after
histogram equalization
process
Figure 2. Image after Hi
sto
g
ram Equ
a
lization Pro
c
e
s
s
Spatial linear filte
r
ing
Space nonlinear
filter
ing
Freque
nc
y
doma
i
n filtering
Color filtering
Figure 3. Image after Filtered
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Cotton Pest
s
and Di
se
ases Detectio
n ba
sed o
n
Im
age Proce
s
sing (Qingh
ai He
)
3447
Whe
n
we g
o
t the ima
g
e
s, d
ue to t
he influe
nce
of light inte
nsity an
d th
e sen
s
or
temperature
and
other factors, th
e ima
g
e
was affe
cte
d
by n
o
ise p
o
llution
whi
c
h
made it
ha
rd
for
segm
entation
,
so the pro
c
ess of
filtering wa
s use
d
in this pape
r
to eliminate the influen
ce
of
noise. The Fi
ltering g
ene
ral incl
ude
sp
atial filter
ing
and fre
que
ncy domain filte
r
ing.
We cho
o
se
different filter for different image
s,but for t
he leaf cotton image
s which d
a
ma
ged by pest
s
or
dise
ases, th
e
experi
m
ent
shows
better
result
s after spatial linear filteri
ng, b
u
t it
also
lo
st a lot
of
details; Spa
c
e nonli
nea
r
filter is a
b
le
to ac
hi
eve
better result while
prese
r
ving the
ba
sic
cha
r
a
c
teri
stics of inform
ation, and
conv
enient.
Fre
q
u
ency do
main
filtering ca
n
achieve
better
visual effects, however, fo
r machi
ne proce
s
sing
can
’
t achieve a better re
sult, for colo
r ima
ges
colo
r filter al
so ca
n obtain
a better visua
l
effect
s a
nd
machi
ne p
r
o
c
essing
re
sults, poor d
e
tails
in
the spa
c
e
no
nlinea
r filteri
n
g. In
summ
a
r
y, sp
ace n
o
n
linea
r filteri
n
g o
r
oth
e
r m
e
thod
s
can
be
sele
cted to
e
x
tract the p
e
s
t and
disea
s
e
cha
r
a
c
teri
stics, the sp
ace
of linea
r filtering, spa
c
e
nonlin
ear filte
r
ing, col
o
r filtering
can b
e
sele
cted
to extract of the whole
le
af characte
ri
stics.
Spatial linear filtering is si
mple and fea
s
ible an
d the
result
s are g
ood. The ima
ges after filte
r
ed
were sh
own in Figure 3.
3. Comparis
on of the
Col
o
r Model
3.1. RGB
Col
o
r Model
The RGB
col
o
r mod
e
l is a
n
y additive color mo
del b
a
se
d on the
RGB col
o
r m
odel.
A
particula
r
RG
B col
o
r
mod
e
l is defin
ed
by the th
re
e
chromati
cities of the
re
d,
gree
n, a
nd
bl
ue
additive p
r
im
arie
s, an
d can p
r
od
uce
any ch
rom
a
ticity that is t
he tria
ngle
d
e
fined by th
ose
primary colors.
The thre
e compon
ent im
age nam
ely R com
pon
en
t image, G comp
one
nt image, B
comp
one
nt image were
shown in Figure 4, we c
an d
r
aw a con
c
lu
sion that the three
comp
on
ent
have a little
e
ffect by the ill
umination
of t
he light.
T
he
White
point of
the ima
ge h
a
s
little effect
on
image p
r
o
c
e
s
sing, the influ
ence of the White poi
nt
can eliminate i
n
later imag
e
pro
c
e
ssin
g
. And
the G comp
o
nent imag
e was u
s
e
d
to th
e se
gment
ati
on of the di
sease or pe
st
image fo
rm t
h
e
origin
al imag
e.
R component im
age
G component im
age
B component image
Figure 4. Co
mpone
nt Image in RGB M
odel
3.2. HSI Color Model
The HSI
colo
r mod
e
l [7, 8, 9] is very importa
nt and
attractive co
lor mo
del for image
pro
c
e
ssi
ng
a
pplication
s
b
e
ca
use it re
pre
s
ent
s
col
o
rs simil
a
rly
how th
e h
u
m
an eye
se
nse
s
colo
rs. T
he HSI color m
o
d
e
l rep
r
e
s
ent
s every col
o
r
with thre
e co
mpone
nts: hu
e (H), sat
u
rat
i
on
(S), intensity (I). The equ
a
t
ions u
s
ed fo
r conv
e
r
ting i
m
age from
RGB colo
r spa
c
e to HSI col
o
r
s
p
ac
e
ar
e
:
G
B
360
G
B
H
(
2
)
1/2
B)]
G)(G
(R
2
G)
[(R
B)]
(R
G)
(1/2)[(R
arccos
θ
(3)
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 6, June 201
3 : 3445 – 34
50
3448
)]
,
,
[min(
3
1
B
G
R
B
G
R
S
(4)
3
B
G
R
I
(5
)
As sho
w
n th
e comp
one
nt image in HSI color mo
del in Figure
5, we can
dra
w
a
con
c
lu
sio
n
th
at the differe
nce
between
the dise
a
s
e
spot a
nd b
a
ckgroun
d is very larg
e i
n
I
comp
one
nt i
m
age. M
ean
while
I comp
onent
ca
n eff
e
ct
ively supp
ress the
influe
nce
which
ca
use
d
by the noise and strong lig
ht. So I comp
onent wa
s
u
s
ed to the seg
m
entation
of the dise
ase spot
image an
d the backg
rou
n
d
image.
H component im
age
S component image
I component ima
ge
Figure 5. Co
mpone
nt Image in HSI Mo
del
3.3. YCbCr
Color Model
The Y
C
b
C
r
color
model
[1
0, 11] h
a
s
be
en d
e
fined i
n
re
spo
n
se to
increa
sing
de
mand
s
for digital alg
o
rithm
s
in ha
ndling vide
o informatio
n, and ha
s sin
c
e
become a
wi
dely use
d
mo
del
in a digital video. It belon
gs to the fam
ily of television tran
smi
ssi
on col
o
r mo
d
e
ls. The
s
e
color
model
s sepa
rate RGB (Re
d
-G
ree
n
-Blu
e
)
into lumina
nce a
nd
chro
minan
ce info
rmation an
d a
r
e
useful i
n
com
p
re
ssi
on a
ppl
ication
s
h
o
we
ver the sp
e
c
ification
of col
o
rs is
som
e
what
unintuitiv
e
.
The equ
ation
s
used for
co
nverting ima
g
e
from RGB
color mo
del to YCbCr color
model a
r
e:
B
G
R
214
.
18
786
.
93
00
.
112
00
.
112
203
.
74
797
.
37
966
.
24
553
.
128
481
.
65
128
128
16
Cr
Cb
Y
(6)
After the Colo
r model tran
sforme
d, the co
mpone
nts im
age were
sh
o
w
ed in the Fi
gure
6.
The Y comp
onent ima
ge
wa
s cho
s
en
as
segm
entat
ion imag
e, b
e
ca
use the d
i
sea
s
e
sp
ot a
nd
backg
rou
nd b
ound
ary is cl
ear in this
co
mpone
nt ima
ge and it’s h
a
s
little affected by light.
Y
compone
nt image
Cb omponent
image
Cr ompon
ent ima
ge
Figure 6. Co
mpone
nt Image in YCb
C
r
Model
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Cotton Pest
s
and Di
se
ases Detectio
n ba
sed o
n
Im
age Proce
s
sing (Qingh
ai He
)
3449
4. Featur
e Selection
From
the
abo
ve re
sea
r
ch t
he b
e
st
co
mp
onent im
age
were d
e
termi
ned fo
r th
e e
x
traction
of disea
s
e
s
o
r
pe
sts imag
e
form
origi
nal
image.
The
ratio of da
mag
e
(
γ
) wa
s
c
hos
e
n
as
fe
a
t
ure
in order to m
easure
the d
egre
e
of d
a
m
age
whi
c
h
ca
use
d
by di
se
ase
s
o
r
pest
s
. The featu
r
e
wa
s
also
u
s
ed
to
comp
are the
extraction
re
sult in
ea
ch
colo
r m
odel
we
don
e b
e
fore
at the
sa
me
time. The formula for mea
s
ure the ratio
of damage
was:
)
%
100
t
A
A
(7)
In this form
ula A me
an
s th
e a
r
ea
of
cotton le
af which
ca
used
by th
e di
sea
s
e
s
or pe
sts,
and At mean
s the total are
a
of cotton le
af.
5. Experimental Re
sults
In the expe
ri
mental resea
r
ch, th
e d
a
m
age
cotton
pl
ant whi
c
h
ca
use
d
by the
dise
ases
and p
e
st
s we
re u
s
ed to te
st the re
sult
s
in different
co
lor spa
c
e m
o
dels. Th
e co
mputer
we
used
in this pa
pe
r
wa
s
DELL
co
mputer,
whi
c
h the
OS i
s
Wind
ows XP;
the P
r
o
c
e
s
sor i
s
Intel
core i5
-
2410M,
2.3G
Hz; th
e me
mory i
s
4GB
of sy
stem
RAM. The
software
we
use
d
is MAT
L
AB
R20
11a. The
extraction
re
sults image a
r
e sho
w
n in Fi
gure 7 a
nd Fi
gure 8.
The original imag
e
Extraction in R
G
B
model
Extraction in HSI
model
Extraction in
Y
C
bCr
mode
Total area o
f
cotton
leaf
Figure 7. The
Results of Extraction Di
se
ase
s
Cotton Image in Th
re
e Colo
r Mod
e
l
The original imag
e
Extraction in R
G
B
model
Extraction in HSI
model
Extraction in
Y
C
bCr
odel
Total area o
f
cotton
leaf
Figure 8. The
Results of Extraction Da
maged
Cotto
n Image whi
c
h cau
s
e
d
by Pests in Th
re
e Colo
r
Model
The compa
r
i
s
on of resul
t
s sho
w
s go
od accu
ra
cy
in both col
o
r mod
e
ls,
nd the
comp
ari
s
o
n
o
f
the results i
n
the th
re
e
color mod
e
l
were
showed
in the
Tabl
e
1. The
dam
a
ge
ratio (
γ
d, ca
u
s
ed by the di
sea
s
e
s
) of e
a
ch
colo
r mo
dels a
r
e RG
B color mo
d
e
l (81.60%
). HSI
colo
r mo
del (43.15%), YCbCr
col
o
r
m
o
del (5
8.40%).
The da
mag
e
ratio (
γ
p,
ca
use
d
by pe
st
s) of
each color
m
odel
s are RG
B colo
r mod
e
l
(5.29%).
HS
I color
mod
e
l (4.90%), Y
C
b
C
r
colo
r mo
d
e
l
(5.93%). And
every color
model ha
s th
eir ow
n adva
n
tage
s and d
i
sadva
n
tage
s, such a
s
RGB
model is ove
r
segme
n
ted, the are
a
of
the image di
se
ase
s
is too la
rge.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 6, June 201
3 : 3445 – 34
50
3450
Table 1. The
Comp
ari
s
o
n
of the Results in the Three
Colo
r Mod
e
ls
Damage r
a
tio
RGB model
HSI model
Y
C
b
C
r mod
e
l
γ
d 81.60%
43.15%
58.40%
γ
p 5.29%
4.90%
5.93%
6. Conclusio
n
s
In this pape
r, a method to measure the damag
e
ratio
of the cotton leaf whi
c
h ca
use
d
by
the disea
s
e
s
or p
e
st
s ba
sed on
ma
chi
ne visio
n
an
d
image
pro
c
e
ssi
ng
wa
s p
r
ese
n
t. The
re
sult
sho
w
s the im
age e
nha
nce
m
ent an
d im
age filteri
ng
we u
s
e
d
in t
h
is p
ape
r
ca
n get a
dam
aged
image with
strong featu
r
e
s
,
It sho
w
s
goo
d accu
ra
cy in
extractin
g
th
e dam
age
d i
m
age fo
rm th
e cotto
n imag
e in three
different col
o
r model (RGB
、、
HSI
Y
Cb
Cr), and the
co
mpari
s
o
n
sh
o
w
s that Ycb
c
r color m
odel i
s
the best color model for extractin
g
in this paper. Th
e ratio of damag
e (
γ
) we
cho
o
se is
reliabl
e.
The re
sult
s i
s
just fro
m
indoo
r expe
ri
ment, whe
n
it come into
real u
s
e, du
e to the
rand
om noi
se
interfere
n
ce
and cotto
n leaf shado
ws e
x
ist, the accu
racy to mea
s
ure the de
gre
e
of dama
ge m
a
y slo
w
d
o
wn, so th
e fu
rther
re
sea
r
ch
need fo
cu
se
d
on the
alg
o
ri
thm to imp
r
o
v
e
the accuracy
and sta
b
ility to
measure the degree of d
a
mage.
Akno
w
l
e
dge
ment
This study was by
su
ppo
rted
by
th
e Nation
al Und
e
rg
rad
uate T
r
ainin
g
Programs fo
r
Innovation an
d Entreprene
urship of Chi
na (No. 1110
7591
6).
Referen
ces
[1]
Sasaki Y, Oka
m
oto T
.
Auto
matic dia
g
n
o
si
s of pla
n
t dis
ease
2 reco
gn
ition
bet
w
e
en
hea
lth
y
an
d
dise
ased l
eaf.
Journ
a
l of the
Society of Agri
cultura
l
Machi
n
ery.
61(2): 119
-126.
[2]
Serna Jr
DG, Pinto F
A
C, Que
i
roz DM, Vi
ana
PA.
F
a
ll Arm
y
w
o
rm d
a
mag
e
d
maiz
e p
l
a
n
t i
dentific
atio
n
usin
g dig
i
tal im
ages.
Biosystem
s Engineering
. 2003; 8
5
: 44
9-45
4.
[3]
Che
n
Jia
j
u
an,
Ji Sho
u
w
e
n,
Li Ju
an, Z
h
a
o
X
ued
u. Auto
matic Meas
ure
m
ent of Da
ng
er De
gree
of
Cotton Ins
e
ct Pests Usin
g C
o
mputer V
i
si
o
n
.
T
r
ansactio
n
s
of the C
h
i
n
ese Soc
i
ety of
Agricu
ltura
l
Engi
neer
in
g.
2001; 02: 1
57-1
60.
[4]
Jun Sun, Yan
W
ang, Xi
ao
ho
ng W
u
, Xia
o
d
ong Z
h
a
ng, H
ong
ya
n Gao. A Ne
w
Ima
ge
Segme
n
tatio
n
Algorit
hm and I
t
s Applicati
on i
n
Lettuce Obj
e
ct Segmentati
o
n.
TELKOMNIKA.
2012; 10(
3
)
: 557-56
3.
[5]
Hans
han
Li, Z
h
i
y
o
ng L
e
i. Re
search o
n
Infra
r
ed Sp
ecia
l F
a
cula Vi
e
w
Me
a
s
ureme
n
t Meth
od Bas
ed o
n
Image Proc
es
sing T
e
chn
o
lo
g
y
.
T
E
LKOMN
I
KA Indon
esia
n Jour
na
l of E
l
ectrical
Eng
i
n
eeri
n
g
. 20
12;
10(6): 14
22-
14
29.
[6]
Gonzal
ez R
C
, W
ood
RE. D
i
gital
Imag
e P
r
ocessi
ng. T
h
ird Ed
ition.
Bei
J
ing: P
ubis
h
i
n
g H
ouse
o
f
Electron
ics Ind
u
st
y
.
20
10.
[7]
Doutre
C,
Nasi
opo
ulos
P, K
o
n
s
tantinos
N
Pla
t
ani
oti. A
F
a
st
Demos
a
icki
ng
Method
Dir
ectl
y
Prod
uci
n
g
YCbCr 4:2:0 O
u
tpu.
IEEE Transactions on Consumer Electr
onics.
20
07; 53
(2).
[8]
Sanj
a
y
Kr, Si
n
gh
DS, Ch
au
h
an, Ma
ya
nk V
a
tsa, Ric
ha S
i
ngh. A
Ro
bust
Skin
Co
lor B
a
sed
F
a
c
e
Detectio
n Alg
o
r
ithm.
T
a
mkan
g Journ
a
l of Sc
ienc
e an
d Engi
neer
ing.
2
003;
6(4): 227-
23
4.
[9]
Nabi
ye
v V, G
üna
y
A. Pro
b
l
e
ms
of C
y
ber
netics an
d
Inf
o
rmatics.
In V
.
Nabiyev (E
d.).
To
w
a
r
d
s
a
biom
etric purp
o
se ima
ge filter
acco
rdi
ng to skin detecti
on. 2
008; 10-
12.
[10]
Sun H
u
i-
xi
an,
Z
hang Y
u
-h
ua,
Luo F
e
i-lu. C
o
lo
r
edg
e d
e
te
ction b
a
se
d on
HSI color s
p
a
c
e.
Optical
T
e
chni
que.
2
0
09; 35(2).
[11]
Kekre HB, S
u
d
eep
D, T
hepad
e,
Parkar A. P
e
rformanc
e An
al
ysis
of Kekre
'
s Media
n
F
a
st Search.
In
H. B. Kekre (
E
d.).
Kekre'
s Centro
id F
a
st Search an
d Exh
austiv
e
Se
arch Use
d
for Colori
ng
a
Gra
y
sca
le Ima
ge. 201
0.
Evaluation Warning : The document was created with Spire.PDF for Python.