![background image](http://html1.mh-extra.com/html/snell-and-wilcox/prefix-cpp100/prefix-cpp100_operators-manual_1309061090.webp)
CPP100
SECTION 9
CPP1OPS
25/10/06
www.snellwilcox.com
Version 1 Issue 6
9.8
Spatial Filtering
Spatial filtering typically involves using an aperture which comprises adjacent pixels from the same field period.
Typically linear filters have been used for noise reduction formed from a weighted average of adjacent pixels.
Usually non-adaptive in nature they reduce noise by averaging contributions across several pixels. In plain
areas, the degree of noise reduction is proportional to the sum of the square of the weighted contributions. In
general a larger aperture will allow a higher level of noise reduction.
Typical levels of noise reduction for equally weighted contributions are shown in the table below:
Aperture Size
Noise power dB
3
-4.7
5
-6.9
7
-8.4
A major disadvantage of this approach is that high frequency picture detail is also filtered and this leads to a
softening of the picture on edges and on textures.
Spatial Aperture comprising adjacent points on the
same line and on adjacent lines above and below
the central pixel.
The spatial filter in PREFIX uses a median filter based on an aperture of 5 pixels shown above.
Median filters operate by rank ordering a set of points selected from a median aperture. The outer points end
up at the extremes of the list with the median value in the middle.
Although median filters can be effective at suppressing impulse they are also effective as gaussian noise
reduction filters.
The spatial filter operates by resolving the spatial median and then verifying if this is of similar value to the
current pixel. By varying the comparison threshold, the spatial filter can be tuned to reject noise but still
preserve picture transitions and textures.
The spatial filter is controlled by two settings that vary the comparison threshold and effectively set the balance
between the level of noise suppression and detail preservation.