TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 14, No. 3, June 20
15, pp. 508 ~ 5
1
5
DOI: 10.115
9
1
/telkomni
ka.
v
14i3.723
5
508
Re
cei
v
ed
De
cem
ber 2
4
, 2014; Re
vi
sed
May 11, 20
15
; Accepte
d
May 25, 20
15
A Skin and Clothes Matching Seeded by Color System
Selectio
n
Rahmadi Ku
rnia*,
Meza Silv
ana,
Ikhw
a
n
a
Elfitri
Dep
a
rtment of Electrical E
ngi
neer
ing, An
dal
as
Univ
ersit
y
,
Kampus Bar
u
Lima
u Manis,
Pada
ng, Indo
n
e
isa 2
5
1
63, T
e
lp/F
ax
: +
62(0
7
51) 72
58
4/+
62(
075
1)72
56
6
*Corresponding author, e-mail:
rahmadi_kurnia@ft.unand.ac.id
A
b
st
r
a
ct
This w
o
rk has desig
ned a
n
auto
m
ate
d
system to
reco
mme
nd cl
othe for peo
ple b
a
s
ed on the
i
r
skin. Skin colors from
var
i
ous rac
e
s w
e
r
e
rel
a
ted
w
i
th a var
i
ety
of clothi
ng
col
o
rs to o
b
tain
thei
r
har
mo
ni
z
a
ti
on
valu
e. T
h
is res
earch
invo
lve
30 res
pon
de
nts throug
h q
u
e
s
tionn
aire to
pr
ovid
e in
itial v
i
e
w
s
to a v
a
riety
of colors
to
match the
skin
col
o
r vari
at
io
n. In
put fro
m
th
e r
e
spo
n
d
ents is
then
an
aly
z
e
d
to
choos
e the rig
h
t color syste
m
to be linke
d w
i
th skin color. To deter
mi
ne th
e relati
onsh
i
p b
e
tw
een skin co
lo
r
and
cl
othes
col
o
r
fu
z
z
y
lo
gic r
u
les
w
e
re
use
d
. The
syste
m
w
a
s then
teste
d
a
g
a
i
n t
o
3
0
r
e
spo
n
d
ents. T
h
e
results show
ed
that variations
in skin co
lor an
d cl
othi
ng ar
e best on a co
mbin
ation
of Cr and Saturati
on.
Ke
y
w
ords
: ski
n color, cloth
e
s
color, har
mo
ni
z
a
ti
on
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The cl
othe co
lors
cho
s
e
n
a
nd se
d by pe
rso
n
c
an indi
cate ma
ny things. For i
n
st
ance: it
can
sup
p
o
r
t some
one’
s a
ppea
ran
c
e,
make th
e ski
n
become m
o
re b
r
illiant,
make
a pe
rson feel
more relaxed
and can be u
s
ed a
s
on
e a
s
pe
ct in dete
r
minin
g
one'
s personality
Curre
n
tly, the ha
rmoni
zat
i
on of
clothe
and
skin
colors i
s
d
e
termined
by
conje
c
ture
aspect
s, or propriety that the ac
curacy is still relatively low. In
th
at case, harmony in clothes
coul
d be matche
d acco
rdi
ng to someo
ne but it l
ooks tacky by others. T
herefo
r
e, it require
d
a
certai
n method in determi
ning the com
p
atibility with
a
system that can be used to accurately and
are universal. Since the dif
f
erence of human
skin
col
o
r
will have
a different level of harmony
to
any colo
r of his cloth
e
s, thi
s
tech
nolo
g
y wa
s ada
pt
ed
to the cha
r
act
e
risti
cs of hu
man skin colo
r.
Two i
m
po
rta
n
t point
s to
create th
e
suit
a
b
ility
o
f
one'
s cloth
i
ng were ski
n
color
cla
ssifi
cation and
h
u
man race determin
a
tion.
Skin
co
lor ide
n
tificati
on an
d the d
e
termin
ation
of a
person
'
s
race
ca
n be don
e
by seve
ral method
s su
ch
a
s
regi
on-b
a
se
d
m
e
thod
s
[1], ski
n co
lor
clu
s
ters u
s
in
g
neu
ral
netwo
rk [2, 3], an
d
neural
fu
zzy [
4
]. Subseque
ntly, the suita
b
ility of cloth
e
s
colo
rs
can
be
dete
r
mine
d
usin
g le
ast
square fitting
method
[7], makin
g
th
e d
e
sig
n
of i
n
tell
igent
system
s to
create a
n
effici
ent an
d effe
ctive clot
h
e
s [8
] and
produ
ction a
u
tomatio
n
in
a
co
rpo
r
ate
fashio
n garm
ent usin
g fuzzy logic [9].
The purpose of this
research
was to
determi
ne the level
of
compatibility of clothes
colo
rs with
o
ne's skin
col
o
r. We
mad
e
an auto
m
ate
d
sy
stem to
determi
ne th
e high
est l
e
vels of
colo
r suitabilit
y between
ski
n
and
clothin
g
. The ne
w a
ppro
a
ch offered in this
re
search
wa
s: the
con
n
e
c
tivity b
e
twee
n the fuzzy mem
b
e
r
ship func
tio
n
s of
skin colo
r and
cl
othing. The
sy
stem will
recomme
nd
an app
rop
r
iat
e
colo
r of clo
t
hes when
th
e type of ski
n colo
r wa
s
detecte
d by inpu
t
came
ra. In this re
sea
r
ch, we u
s
ed fu
zzy logic metho
d
to match the variation of
skin
colo
r in
put
with its
app
ro
priate
cloth
e
s. We a
nalyze
d a lot
of dat
a input f
r
om
30 respon
den
ts that gave
their
valuation by que
stione
r sh
eet
to
d
e
term
ine
the
suitab
ility
colo
r spa
c
e model. Th
en,
we creat
ed
a fuzzy me
mbershi
p
function of this colo
r m
odel
and de
sign i
t
s correl
ation
with skin
color
membe
r
s
h
ip f
unct
i
o
n
by
f
u
zzy
rule
sy
st
e
m
s.
Fi
nally,
we
obtain
ed t
h
ree
of the
m
o
st h
a
rm
onio
u
s
clothe
s colors for each skin
color.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Skin and Cl
othes Mat
c
hi
ng Seede
d b
y
Col
o
r Syste
m
Selection (Rahm
adi Kurnia)
509
2. Rese
arch
Metho
d
This
study
con
s
i
s
ted of
several sta
ges: Th
e first stag
e wa
s hum
an
skin colo
r
cla
ssif
i
cat
i
on.
I
n
t
h
is
st
a
ge,
30
s
k
in
col
o
r
sa
mple
s of
h
u
man
we
re
t
r
ied o
u
t
t
o
cla
ssif
y
t
h
e
m
int
o
three
ra
ce
s b
a
se
d on thei
r colo
r mod
e
l
s
. We fixed t
he Ycb
C
r col
o
r mo
del. From, these
ra
ce
c
l
as
s
i
fic
a
tion,
the fuz
z
y
members
h
ip func
tion
s wa
s de
t
e
rmine
d
t
o
make a r
u
le
s of
f
u
zzy
.
The suitability of skin col
o
r variation was
de
sig
ned
with cloth
e
s
as a second
stage. A
que
stionn
aire
wa
s di
strib
u
ted t
hat
conta
i
ned a l
o
t of i
m
age
s of vari
ous
hum
an ra
ce
s with
a
wi
de
variety of cl
o
t
hes
col
o
rs t
o
30
re
sp
on
dents.
Ea
ch
respon
dent
wa
s given
3
0
variatio
ns
of a
comm
on colo
r of clothing
worn by the peopl
e fo
r e
a
ch type of ski
n col
o
r. Resp
ond
ent g
a
ve
rating
s of 1 to very misma
t
ched an
d 10
for very
harmoniou
s. The
three highe
st score re
sult
s of
total resp
ond
ents was
ran
ged a
s
a dom
inant col
o
r of each skin col
o
r.
The third
sta
ge wa
s the relation
ship d
e
sig
n
betwee
n
human
skin
colo
rs
with the col
o
r of
clothin
g
. The sele
cted
colo
r model (RGB
, HIS or YCB
Cr) whi
c
h is suitabl
e for the clothe
s, the
n
cre
a
ted ea
ch
of its fuzzy membe
r
ship functio
n
and t
he app
rop
r
iat
e
rule for thi
s
system. Fina
lly,
the relation
sh
ip of fu
zzy
rul
e
b
e
twe
en fu
zzy
memb
ership fun
c
tion
o
f
ski
n
colo
r
a
nd
clothe
s
we
re
desi
gned to
get the best
suitabilit
y index. This index is needed to
select an appropriate
clothes
colo
r with th
e
variou
s in
se
rt skin
col
o
r. T
he su
itability index was
differentiate
d bet
wee
n
men
an
d
wome
n cloth
e
s.
2.1. Color Models and S
k
in Color
Colo
r m
odel
s provide a
standa
rd
way t
o
spe
c
ify
a particula
r col
o
r,
by definin
g
a 3D
coo
r
din
a
te system, and a sub
s
pa
ce t
hat contai
ns
all con
s
tru
c
ti
ble colo
rs wi
thin a particular
model. Any color that
can
be sp
ecifie
d
usin
g a mo
d
e
l
will co
rrespo
nd to a si
ngle
point within t
h
e
s
u
bs
pa
ce
its
d
e
f
in
es
.
2.1.1. RGB
The mo
st co
mmonly use
d
and po
pul
ar color
spa
c
e is
RGB.
Ho
wever, thi
s
sp
ace
pre
s
ent
s so
me limitation
s
: (i) th
e pre
s
en
ce
of
a n
egative pa
rt in the spe
c
tra
,
which doe
s not
allow th
e re
prese
n
tation of
certai
n colo
rs by a su
pe
rpo
s
ition of the t
h
ree
sp
ect
r
a,
(ii) the
difficul
t
y
to determi
ne
colo
r features li
ke the p
r
esen
ce
o
r
the ab
sen
c
e
of a given color, an
d (iii
) the
inability of the Euclidean di
stance
to correctly capture
color di
fferences in the RGB space.
A color i
n
this spa
c
e i
s
rep
r
ese
n
ted by a
tr
iplet of values typically b
e
twee
n ze
ro
and on
e
and i
s
u
s
u
a
lly scaled
by 2
5
5
for
an 8
-
bit
rep
r
e
s
ent
atio
n. Each
color ca
n be
broken do
wn
into
its
relative inten
s
ity in the th
ree pri
m
ari
e
s corre
s
pon
di
ng to the sp
ectral
respon
se of one of
the
three
types
of co
ne
s
pre
s
ent i
n
th
e h
u
man
eye:
red, g
r
ee
n a
n
d
blu
e
. Th
e
spa
c
e
is e
a
sily
rep
r
e
s
ente
d
as a three di
mensi
onal
cu
be wh
ere e
a
ch axis rep
r
e
s
ents the st
ren
g
th.
2.1.2. YCbCr
YCbCr i
s
a f
a
mily of col
o
r spa
c
e
s
u
s
e
d
as a
pa
rt of i
m
age
pipelin
e in video
an
d digital
photog
rap
h
y B and CR a
r
e the blue-differen
c
e an
d
red-diffe
ren
c
e
chro
ma com
pone
nts. Y (with
prime
)
i
s
di
stingui
shed
fro
m
Y whi
c
h i
s
luminan
ce,
meanin
g
that
light inten
s
ity is non
-lin
e
a
rly
encode
d u
s
in
g gamm
a
correctio
n
[12]. The Y in YCb
C
r d
enote
s
th
e lumina
nce
comp
one
nt, and
Cb and
Cr re
pre
s
ent the chromi
nan
ce factors .In
YCbCr, the Y is the brightne
ss (l
uma
)
, Cb
is
blue mi
nu
s lu
ma (B
- Y) an
d Cr i
s
red mi
nus luma
(R - Y) [5]. If R,
G an
d B a
r
e
given with
8
bit
digital p
r
e
c
isi
on, then Y
C
b
C
r f
r
om di
gital 8-bit ca
n b
e
obtain
ed from RGB coo
r
dinate
u
s
ing
a
transfo
rmatio
n matrix as i
llustrate
d in Equati
on (1).
When rep
r
e
s
entin
g the signal
s in digi
tal
form, the results are
scal
ed and ro
un
ded, and o
ffsets are typically added.
F
o
r example, the
scaling an
d offset applied
to the Y compone
nt per
spe
c
ification
result
s in the value of 16 for
black an
d the
value of 235
for white
wh
en usi
ng a
n
8
-
bit rep
r
e
s
e
n
tation. The
standa
rd h
a
s 8
bit
digitize
d versi
ons of Cb an
d Cr
scaled to a different range of 16
– 2
40 [4].
B
G
R
Cr
Cb
Y
214
.
18
112
996
.
24
786
.
93
203
.
74
553
,
128
112
757
.
37
481
.
64
128
128
16
(1)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 3, June 20
15 : 508 – 51
5
510
2.1.3. HIS
The HSI col
o
r model is al
so
base
d
on the
char
acte
ri
stics of the hu
m
an's visual sy
stem. I
denote
s
the li
ght intensity, H
den
otes th
e hue that ind
i
cate
s the me
asu
r
e of the color pu
rity, S
is
the satu
ration
(the deg
ree
of a color p
e
rmeated the white colo
r). If a colo
r is with
high satu
rati
on
value, it means the color i
s
with the lo
w whit
e col
o
r. The rel
a
tion
ship between
HSI and RGB can
be de
scribe
d as [15]:
2
1
2
1
2
1
cos
B
G
B
R
G
R
B
R
G
R
(
2
)
B
G
R
I
3
1
(
3
)
B
G
R
B
G
R
S
,
,
min
3
1
(4)
2.1.4. Skin Color
The
ai
m
of skin
color
pixel cla
ssifi
cation
is
to determin
e
if
a col
o
r pi
xel
is a ski
n color or
non-skin colo
r.
Good ski
n colo
r
pixel
cl
assificatio
n
should p
r
ovide
coverage of
all different skin
types (bla
cki
sh, yello
wish
, bro
w
ni
sh,
whitish, et
c.) and
cater f
o
r a
s
many
different lig
hting
con
d
ition
s
as
possibl
e.
Skin
colo
r i
s
one
of the
ch
ara
c
te
risti
c
s of h
u
ma
n identifi
c
ation a
nd h
u
m
an race
cla
ssifi
cation.
Skin
colo
r h
a
s a
hig
h
se
nsitivity to ch
ange
s in
light
. Therefo
r
e, t
he alte
ration
of
ski
n color i
m
a
ge from
RGB
spa
c
e to Y
C
b
C
r m
odel
wa
s very suita
b
le
for dete
c
ting
ski
n color
du
e
to the influen
ce of lumina
n
c
e can be eli
m
i
nated du
rin
g
the image p
r
ocessin
g
[10
,
11].
In gen
eral,
h
u
man
skin
ca
n be
group
ed
into
3
main
race
s:
ski
n bl
ackish, brownish
and
whitish [9]. But in this re
search, the
skin colo
r a
r
e
cla
ssifie
d
into cau
kcsoid,
mongol
oid a
nd
negroid. Th
e
Skin-col
or
distrib
u
tion a
nd Ga
ussi
a
n
distri
bution in
Cg
-Cr sp
ace are
sho
w
n
Figure 1.
Figure 1. Skin-colo
r
distri
b
u
tion and G
a
us
sian di
strib
u
tion in Cb
-Cr spa
c
e [14]
Comm
only, the h
a
rm
oni
za
tion of
ski
n a
nd
clothe
s
co
lor in
dicated t
he
comp
atibil
ity of a
p
e
r
s
o
n'
s
ap
pe
a
r
an
ce
.
2.2. Fuzzy
Logic
Fuzzy logic
has a conti
nuou
s value,
wh
ich
can
be expressed in degre
e
s of a
membe
r
ship.
The me
mbe
r
shi
p
fun
c
tion
s u
s
ed i
n
thi
s
research
was the t
r
ian
g
u
lar m
e
mbe
r
ship
function
curv
e whi
c
h wa
s
sho
w
n in the
following e
q
u
a
tion [13]:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Skin and Cl
othes Mat
c
hi
ng Seede
d b
y
Col
o
r Syste
m
Selection (Rahm
adi Kurnia)
511
c
x
c
x
b
b
c
x
c
b
x
a
a
b
a
x
a
x
c
b
a
x
f
0
0
,
,
,
(5)
This fuzzy m
e
mbership function
wa
s illustrated in Fi
gure 2 below:
Figure 2. Tria
ngula
r
fuzzy membe
r
ship functio
n
3. Results a
nd Analy
s
is
3.1. Skin Color Range
Skin colo
r ra
nge wa
s obta
i
ned by
an
alyzing
of
3
0
ski
n color sampl
e
s, a
s
sh
own
the tabl
e
below. This
res
u
lt was
s
o
mewhat diffe
rent from previous
studie
s
[4, 6].
Table 1. Ra
n
ge of ski
n col
o
r in the YCb
C
r
colo
r syste
m
T
y
pe of
skin
Y
Cb
Cr
kaukasoid
180-216
197-250
189-200
mongoloid
134-186
160-203
155-192
negroid
80-140
100-166
100-165
Furthe
rmo
r
e,
from th
ese
value
s
, we
created
the
i
r fu
zzy m
e
mbershi
p
fu
nction
a
s
illustrated in Figure 3 below:
Figure 3. Sk
in c
o
lor fuzz
y membership func
tion [6]
3.2. Questio
nnaire Resul
t
s
Five high
est
score of
h
a
rmo
n
ization
between skin col
o
r an
d
cl
othing
f
r
om
30
respon
dent
s descri
bed in
Figure 4 and
5 belo
w
.
Thre
e
hi
ge
st
sco
r
e
fo
rm each con
d
itio
n
de
scri
bed
by
Figu
re 4 and 5 we
re use
d
to
anali
z
e of
ha
rmoni
zatio
n
color
between
cloth
e
a
nd h
u
man
skin. F
r
om th
e g
r
ap
h ab
ove, there
are
differe
nces i
n
ha
rmo
n
y
clothin
g
col
o
r fo
r mal
e
a
nd femal
e
o
n
the
same
type of
ra
ce. T
h
is
can
be
u
nde
rstood
a
s
th
e
level of
pro
p
riety clothe
s b
e
twee
n fem
a
l
e
an
d m
a
le i
s
diffe
rent. T
he
domina
n
t clot
hing color fo
r female in bl
ack in
the ha
rmony of the variou
s ra
ce
s. For male, the
blue cl
othing
colo
r wa
s do
minant in the
harmo
ny of the variou
s ra
ce
s. The
s
e colors be
com
e
a
r
e
ferenc
e for
c
r
eating fuzzy member
s
h
ip func
tions
in eac
h gender
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 3, June 20
15 : 508 – 51
5
512
Figure 4. Five highe
st sco
r
e of harmoni
zation
b
e
twe
en clothi
ng a
nd female (a) cau
c
a
s
oid, (b)
mongol
oid, (c) neg
roid
Figure 5. Five highe
st score of harmo
nization bet
wee
n
clothin
g
an
d male (a
) ca
uca
s
oi
d, (b)
mongol
oid, (c) neg
roid
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Skin and Cl
othes Mat
c
hi
ng Seede
d b
y
Col
o
r Syste
m
Selection (Rahm
adi Kurnia)
513
3.3. Clothing
Color Model
Selection
We a
nalyzed
a ha
rmony l
e
vel of skin
color
and
clot
hing that o
b
tained f
r
om
re
spo
nde
nt
by calculatin
g the cloth
e
colo
rs
distri
b
u
tion in
ea
ch
colo
r mod
e
l (RGB,HSI, Ycb
C
r). Then,
we
sele
ct an a
p
p
r
op
riate colo
r model to rel
a
te with skin
colo
r. The
re
sults
we
re
sh
own a
s
T
able
2
and 3.
Table 2. Vari
ous femal
e
cl
othes
colo
r m
odel
Skin T
y
pe
Textile Color
R
G
B
H
S
I
Y
Cb
Cr
Kaukasoid
Red
206 18
51
350,532
80,01
92
84 151
208
Black
35
37 34
100,0000
3,79
35 47
143
128
Gree
n
46 130
67 134,4240
43
81
100
158
128
Mongoloid
Red
206 18
51 350,5320
80,01
92
84 151
208
Black
35
37 34
100,9080
3,79
35 47
143
128
Darkgre
y
95
102 110
211,8960
18,63
163 103
176 128
Negroid
Black
35
37 34
100,9080
3,79
35 47
143
128
White
239 241
238 100,9080
0,56
239
222 232
128
Light
gre
y
151 147
158 233,9280
3,23
151
143 198
128
Table 3. Vari
ous mal
e
clot
hes
colo
r mo
del
Skin T
y
pe
Textile Color
R
G
B
H
S
I
Y
Cb
Cr
Kaukasoid
Red
206 18
51 350,532
80,01
92
84 151
208
Blue
3
67 132
209,988
96,27
67
64
186
128
Oran
ge
241
88
48
11,412
61,781
126
127 149
198
Mongoloid
Red
206 18
51 350,532
80,01
92
64 186
128
Black
35
37 34
100,908
3,79
35 47
143
128
White
239 241
238 100,908
0,56
239
222 232
128
Negroid
Black
35
37 34
100,908
3,79
35 47
143
128
White
239 241
238 100,908
0,56
239
222 232
128
Light
gre
y
151 147
158 233,928
3,23
151
143 198
128
From the sixt
h figure
s
abo
ve, HSI color system had
a regul
ar val
ue obtaine
d betwe
en
the re
sult
s of
the
colo
r g
r
o
ups a
c
cordi
n
g to the
type
of skin
col
o
rs for b
o
th
wo
men a
nd
me
n is
the HSI
colo
r sp
ace, whil
e
in an
othe
r color
sp
ace
v
a
lue
s
a
r
e
not
reg
u
lar a
n
d
overla
p. In t
h
a
t
ca
se,
t
he col
o
r of
clot
he
s i
n
t
h
is re
sea
r
c
h
wer
e
co
nv
e
r
t
ed int
o
HI
S
.
3.4. Fuzzy
Rule
A rule of fuzzy wa
s fixed base
d
on range
of the
output fuzzy system me
m
bership
function
inde
x. In this research
we
ob
tai
ned th
e
caucasoid
skin color ha
d
the outp
u
t fu
zzy
system m
e
m
bership fu
ncti
on ind
e
x in th
e ran
ge of
0
up to 0.5. Th
e mong
oloid
ski
n color ha
d the
output fuzzy system m
e
m
bership fu
nct
i
on index
in
the ran
ge of
0.5 up to 0
.
7. Finnaly, the
negroid ski
n colo
r
the ca
u
c
a
s
oid skin color had
th
e
output fuzzy
system m
e
m
bership
fun
c
tion
index mo
re t
han 0.7. F
o
r example if
we h
ad
pixel
ski
n with
Cb = 1
67
and
the output f
u
zzy
system m
e
m
bership fu
ncti
on ind
e
x of clothe
s
colo
r
saturation
=
0.5, so it wou
l
d get the va
lue
of output = 0
752. Thi
s
val
ue indi
cated
that
the relati
onship bet
we
en skin
colo
r Cb =
167 a
nd
clothes at sat
u
ration
= 0.5
will be
suitabl
e for negroid
ski
n col
o
r. Th
e com
p
lete rule fuzzy
system
wa
s sh
own in
Figure 6.
Figure 6. Fuzzy rule outp
u
t system [6]
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 3, June 20
15 : 508 – 51
5
514
3.5. Index of
Harmoni
zati
on
The ha
rmoni
zation index of
skin color to
colo
r
clothe
s in the syste
m
was illu
stra
ted in
Figure 7 and
Figure 8.
Figure 7. Harmonization in
dex of female [6
]
Figure 8. Harmonization in
dex of male
From th
e figu
res, th
e m
o
st
stabl
e
syste
m
is
th
e
com
b
ination
of Cb-S. The
be
st
re
sult of
t
he sy
st
e
m
w
a
s o
b
t
a
ine
d
if
we
set
t
he
s
k
i
n col
o
r i
n
Cb
membe
r
ship f
u
nctio
n
an
d clothing
colo
r i
n
saturation
m
e
mbe
r
ship fu
nction
on
thei
r relation
shi
p
. Othe
rwi
s
e,
p
oor
re
sult i
s
t
he
combi
nati
o
n
of Cr-I color matchin
g
sy
stem. This
re
sult oc
cu
rred
beca
u
se the
rang
e value
s
of Cr
wa
s
not
stable fo
r all
ski
n tone
s
while th
e val
ues
of In
t
e
n
s
it
y
clot
he
s
colo
r alm
o
st
t
he sam
e
f
o
r
all
condition.
3.6. Sy
stem
Ev
aluation
System evaluation wa
s
perfo
rmed b
y
insert
ing 2
4
other vari
ous
ski
n col
o
rs.
We
obtaine
d
a suitable clothi
ng colo
r clo
s
ed
to experi
m
ental re
sult.
The re
sult
s of
this evalu
a
tion
were sh
own Figure 9.
(a)
(b)
Figure 9. Evaluation re
sult
(a) fem
a
le [6], (b) male
The
evaluatio
n result
s
sh
o
w
ed
that the
system
was run
well. It ca
n be
seen
fro
m
mo
st
of the output value is in the
range d
e
si
re
d output.
4. Conclusio
n
This
wo
rk ha
s p
r
op
osed
an auto
m
atic co
rrel
a
ti
on
system
of skin col
o
r va
ria
t
ions to
recomme
nd
the approp
ri
ate colo
r of clothing.
T
he syste
m
a
l
so succe
s
sfully manage
d to
separate the
compatibility level cl
othes by
sex. Relat
i
ve
element
and subjective
assessment in
this research
we
re elimi
n
a
t
ed by the n
u
m
ber
of
re
sp
onde
nts who use
d
t
he
re
sults of research
and testin
g b
a
ck to the re
spond
ent.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Skin and Cl
othes Mat
c
hi
ng Seede
d b
y
Col
o
r Syste
m
Selection (Rahm
adi Kurnia)
515
Ackn
o
w
l
e
dg
ements
This resea
r
ch
was
con
d
u
c
ted und
er fina
ncial
sup
port
from DIPA No. 023.04.2.4150
61,
Faculty of Enginee
ring, An
dala
s
Unive
r
sity, Padang-Indon
esi
a
Referen
ces
[1]
Rudr
a PK P
o
ude
l, et a
l
.
R
egi
on B
a
se
d
Skin C
o
l
o
r D
e
tection
, Proc
e
edi
ng Int
e
rnati
ona
l Jo
urn
a
l
Confer
ence
on
Comp
uter Vis
i
on, Imag
in
g a
nd C
o
mp
uter
Graphics T
heo
r
y
and
App
lic
a
t
ions. Rom
e
,
Ital
y
. 20
12; 1: 301-
306.
[2]
Hani K Al M
o
hair, et al, Hu
man Skin C
o
l
o
r De
tectio
n: A Revie
w
on
Neur
al Net
w
o
r
k Perspective
.
Internatio
na
l Journ
a
l of Innov
ative
Co
mputin
g, Informati
on
and C
ontrol.
2
012; 8(1
2
): 811
5-81
31.
[3]
Mu Z
hang, et a
l
. Neura
l
Net
w
ork Based C
o
l
o
r Reco
gniti
on
for Bobbi
n Sort
ing Mac
h
in
e.
TELKOMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2013;
7(1
1
): 3728-
373
5.
[4]
Mohamm
ad S
a
ber Iraj
i. Skin
Color
Segm
ent
ation
in Y
C
BC
R Co
lor S
pace
w
i
t
h
Ad
aptiv
e
F
u
zz
y
N
eur
al
Net
w
ork (Anfis
).
Internation
a
l
Journ
a
l Image,
Graphics an
d Sign
al Process
i
ng.
20
12; 4: 35-41.
[5]
Chu
an
Lin,
et al. F
a
ce
Det
e
ction
Alg
o
rith
m Ba
se
d o
n
Multiori
entati
o
n
G
abor F
i
lters
an
d F
eatur
e
F
u
sion,
T
E
LK
OMNIKA Indon
esia
n Journ
a
l o
f
Electrical Eng
i
ne
erin
g.
201
3; 11(10):
59
86-
599
4
[6]
Meza Silv
an
a, Rahm
adi K
u
r
n
ia. Sistem P
e
ndeteks
i
a
n
Ke
serasi
an W
a
rn
a Kul
i
t dan B
u
sana S
e
car
a
Otomatis unt
u
k
Jen
i
s K
e
la
min P
e
remp
u
an B
e
rb
asis I
m
age
Process
i
ng.
J
u
rn
al N
a
sio
nal
T
e
kn
ik
Elektro
. 201
4; 1(3): 18-2
4
.
[7] B
Z
hang.
Res
earch
on
Nu
merical
Ana
l
ysis
for Col
o
r Mat
c
hin
g
in T
e
xtil
e Dye
i
ng
Bas
ed o
n
L
east
Squar
e F
i
ttin
g
.
IEEE Intern
ati
ona
l C
onfer
en
ce o
n
Inte
lli
ge
nce
and
Sec
u
r
i
t
y
Inf
o
rmatics
(ISI). Beijin
g,
Chin
a. 20
11; 1: 289-2
9
2
[8]
Yong YT
, et
al. A Ne
w
an
d Efficient Intelli
ge
nt Coll
ab
oratio
n Schem
e for F
a
shion
Desig
n
.
IEEE
T
r
ansactio
n
s o
n
System, MA
N and Cy
ber
n
e
t
ics-Part A: System a
nd Hu
man
. 201
1; 41(3)
: 463-47
5
[9]
LC W
a
ng, et
al.
Fo
rma
l
i
z
a
t
io
n
of F
a
shi
o
n Se
nsory
Da
ta Base
d
on
F
u
zz
y
S
e
t T
h
eory
. F
ourt
h
Internatio
na
l C
onfere
n
ce o
n
Natura
l Camp
u
t
ation. W
a
shin
gton DC, USA.
2008; 7: 80-
84
.
[10]
Phun
g SL, et a
l
. Skin Segm
en
tation Usi
ng C
o
lor Pi
xel Cl
as
sificatio
n
: Anal
ysis
and C
o
mp
ariso
n
.
IEEE
T
r
ansactio
n
s o
n
Pattern Ana
l
ysis and Mac
h
i
ne Intell
ig
ence
.
2005; 2
7
(1): 1
48-1
54.
[11]
Murinto, et a
l
., Deteksi Je
ni
s W
a
rna Ku
lit
W
a
jah U
n
tuk
Klasifik
asi R
a
s Man
u
sia
Meng
gun
aka
n
T
r
ansformasi W
a
rna. Und
e
r grad
uate T
hesi
s
. Indonesi
a
: Ahmad D
ahl
an
Univers
i
t
y
. 2
0
0
8
.
[12]
IAG Boaventur
a, et al. F
u
zz
y
Classific
a
tio
n
o
f
Human Skin
Color i
n
Co
lor Images. 20
06.
[13]
Xi
aomi
ng Z
h
o
u
, et al. T
i
ssue F
l
o
w
Detecti
on Us
ing F
u
z
z
y
L
o
g
i
c Meth
od i
n
Co
lor F
l
o
w
Imag
ing
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2014; 1
2
(9): 6
840-
684
5.
[14]
Kamarul Ha
w
a
ri, et.al. An
In
n
o
vative F
a
c
e
D
e
tection
bas
ed
on Sk
in C
o
l
o
r
Segme
n
tatio
n
.
Internatio
na
l
Journ
a
l of Co
mputer App
lic
ations
. 201
1; 34(
2): 6-10.
[15]
Rahm
adi Kur
n
i
a
, et al. Generati
on of Efficie
n
t dan User-fri
end
l
y
Qu
eries
for Help
er Rob
o
ts to Detec
t
T
a
rget Object.
Advanc
ed R
o
b
o
tic Journ
a
l
. 20
06; 20(5): 4
99-
517.
Evaluation Warning : The document was created with Spire.PDF for Python.