Information Technology Reference
In-Depth Information
010
111
010
111
121
111
121
242
111
1
5
1
10
1
16
When the smaller the center element of h ( u,v ) takes up, the better the smooth-
ing resultis, but the fuzzier infrared image will be. Totally, we need to choose
appropriate template practically. Fig. 1 shows the de-noised results by different
mean filtering windows and different weighted templates under Gaussian noise
(noise mean: 0, noise variance: 0.15).
1
2
3
SSS
J
Fig. 1. The results with Gaussian noise de-noised by the mean filters
It is obvious that the neighborhood mean filtering can remove Gaussian noise
well. While the bigger the filtering window is, the better the de-noising effect
will be, but the fuzzier the image will be. Although the weighted mean filter-
ing algorithm could keep the details in the image. The drawback is that while
the template is going to be large, the image will be relatively poor, though the
de-noising results become better. So it takes us great trouble to diagnose the
fault.
2.2 Median Filtering
Median filtering is a nonlinear filtering technology, which could inhibit the ran-
dom noise without making the details of the image fuzzy, and it is very effective
to filtrate the impulse interference and image scanning noise. Since image signal
is generally related in two dimensions, 2D sliding window is used for the median
filtering, the 2D median filter could be described as
y ij = Me S {
x mm }
(4)
Search WWH ::




Custom Search