Digital Signal Processing Reference
In-Depth Information
R
Color
Object
Signal
Plane
B
R'
G
Principal Axis
Transform
Color Space
B'
G'
Noise
Planes
Eigenspace
Fig. 2.16
Color space transform to eigenspace
the desired and unwanted pixels that is described as follows:
<
Π
,
SA
1
NA
1
SB
1
NB
1
SA
2
−
(2.39)
NB
1
·
·
where
SA
1
=
s
k
·
w
k
−
1
,
s
NA
1
=
s
k
·
w
k
−
1
,
n
1
NA
2
=
s
k
·
w
k
−
1
,
n
2
SB
1
=
s
k
−
1
·
w
k
−
2
,
s
NB
1
=
s
k
−
1
·
w
k
−
2
,
n
1
NB
2
=
s
k
−
1
·
w
k
−
2
,
n
2
.
In (
2.39
), the first term represents the current pixel and the second term repre-
sents the previous pixel. The index
k
represents the current searching location and
k
1 represents the previous location. The term
s
k
and
s
k
−
1
represent the current
and previous RGB color pixel. The term
w
k
−
1
,
s
,
w
k
−
1
,
n
1
,and
w
k
−
1
,
n
2
represent the
eigenvectors of the
−
th vectors in signal and two noise planes. The terms,
w
k
−
2
,
s
,
w
k
−
2
,
n
1
,and
w
k
−
2
,
n
2
represent the eigenvectors of the
(
k
−
1
)
th vectors in
signal and two noise planes. The term of
SA
1
and
SB
1
represent the RGB pixels
projected on the signal plane, whereas,
NA
1
,
NA
2
,
NB
1
,and
NB
2
are the RGB pixels
projected on the noise plane. We find out that is appropriate to take
(
k
−
2
)
α
=
2,
β
=
1,