Input raw image
output despike image.
The DESPIKE algorithm consists of two tests. The first test searches for all
"suspicious" pixels in each 3x3 pixel area by comparing its difference from
its median to the mean difference for all 9 pixels:
8
---
S \
|Di-Mi| > --- / |Dj-Mj| + To (1)
9 ---
j=0
The second test is applied if the central pixel Do is "suspicious":
|Do - Eo| > S x A + To (2)
where
Eo is an estimate of the "correct" value of Do,
A is a scene activity measure.
In both tests, the right side of the inequality represents a threshold used
to determine whether the difference is significant. In both cases, the
threshold is computed as a linear function where the scale and offset constants
are specified via the SCALE and TOL parameters.
See help file for further details, including suggestions for selecting these
parameters.
See help for SCALE parameter.
When this is specified, ony positive spikes are allowed. A motivation might be if you know that the source of the noise always increases the DN.