Information Technology Reference
In-Depth Information
become a proven technique. And infrared image de-noising became an important
subject[3]. In 1997, Johnstone with his team gave out the wavelet threshold
estimator about the related wavelet de-noising method. In 2000, Chang, put
forward self- adaptive wavelet threshold de-noising method in spatial domain
which combines translation invariant method with selfadaptive threshold[4]. For
the past few years, there were some new infrared image de-noising methods based
on other theories, such as Fuzzy Mathematics[5], Neural Network[6], Curvelet
Transform and Nonsubsampled Contourlet Transform and so on[7].
This paper mainly studies on the infrared image of the fault of the power
transmission lines which are interfered by the various noises, where the software
of Matlab is taken as the workbench.
2 Infrared Image De-noising
2.1 Neighborhood Linear Filtering
If the pixel value of processed image is
f
(
u,v
), the grey value could be
g
(
u,v
)
after transformation, the space neighborhood mean de-noising algorithm could
be given as follows
g
(
u,x
)=
1
N
f
(
u
−
m,v
−
n
)
(1)
(
m,n
)
∈S
Where
N
denotes the number of pixel value in the neighborhood, while S de-
notes the neighborhood. The size and shape of the neighborhood are decided by
infrared image, usually rectangle is used, because the image data is matrix like
A
×
B
, so the neighborhood areaS can be 3
×
35
×
57
×
7 and so on. For example,
when
S
is 3
×
3, Eq.1 can be described as
1
1
g
(
u,x
)=
1
9
f
(
u
+
i,v
+
j
)
(2)
i
=
−
1
j
=
−
1
The mean template is as 1
/
9
∗
[111;111;111].
Neighborhood mean method is a simple and effective image smoothing method
in spatial domain, which takes place of current pixels with the average of adja-
cent pixels, making the gray value in spatial domain even and playing the role
of smoothing the gray value. The weighted mean method is: if the pixel value of
processed image is also f(u,v), the gray value could be g(u,v) after transforma-
tion, then the weighted mean method could be defined as
g
(
u,v
)=
h
(
u,v
)
×f
(
u,v
)
(3)
Where h(u,v) denotes the matrix of weighted template. The common weighted
template matrix has the kinds of form as follows
Search WWH ::
Custom Search