Graphics Reference
In-Depth Information
will be set as 0. Otherwise it will be set as 1. After the binarization process being
completed, read the binaries obtained by thresholding from the top-left corner one by
one clockwise. Then transform the string of binary into a decimal number, which is
looked on as a response of the center point.
To meet the needs of different sizes of the texture feature information, Ojala et al.
[13] have made a further improvement on the basic LBP operator. The original 3×3
neighborhoods can be extended to any neighborhoods, and circular neighborhoods are
used instead of the previous square neighborhoods. As a result, any radiuses and
numbers of neighborhood pixels could be obtained.
Fig. 2 shows three cases of circular
LBP
operator model, in which the radius
R
is 1, 2,
and 3 respectively, and the number of neighborhoods
P
is 8, 16, and 24 accordingly. If
the neighborhood point is just in the center of the grid, the gray value of the square in
which the pixel located can be regarded as its own gray value directly. If the neigh-
borhood point is located in a cross section of two squares, its gray value can be calcu-
lated by bilinear interpolation.
Fig. 2.
R=1, 2, 3, P=8, 16, 24 circular LBP operator model
So since the general process of the
LBP
operator is as follows. Set an arbitrary pixel
within a local area of an image is
f
(
x
c
y
,
)
. And look on this pixel as the center pixel,
c
g
, and the gray values of the neighbors in the local unit are
whose gray value is
respectively. The texture features T of the local region can be ex-
pressed as equation (1):
g
g
…
g
0
1
p
T
≈
t
(
g
−
g
,
g
−
g
,...,
g
−
g
)
(1)
0
c
1
c
p
c
Regard the gray value of the center pixel as the threshold, and the other pixels in the
neighborhood unit are processed by binarization following the method defined as
formula (2):
1
x
>
0
T
≈
t
(
s
(
g
−
g
,
g
−
g
,...,
g
−
g
)),
s
(
x
)
=
0
c
1
c
p
c
(2)
0
x
≤
0
Search WWH ::
Custom Search