Image Processing Reference
In-Depth Information
Figure 6.2
Median filter and fine detail.
3 pixels [see
Fig. 6.2(a)], there are four pixels of value 1 and five of value 0. That is, after sorting,
the list looks like the following with 0 at the center:
A corner pixel is deleted because for a simple window of say 3
×
{0, 0, 0, 0, 0, 1, 1, 1, 1}.
!
Therefore, a value of 0 is placed at the center of the window.
This effect may be removed by giving the center pixel a weighting of 3. Hence,
it is placed in the list three times and results in the following list:
{0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1}
!
with the output now 1. The filter thus gives more importance to the center pixel.
However this supresses the noise reduction properties. The center weighting may
be calculated for larger windows as shown in Fig. 6.2(b). For an odd-sized square
window of dimension (2
n
+1)
(2
n
+ 1) placed at the corner pixel, the number of
background pixel values in the list corresponds to
×
3
n
2
+
2
n
.
(6.6)
