[GRASS-user] Filtering high "outliers" in Landsat reflectance imagery?

[Nikos Alexandris]

Nikos A wrote:
> > >> Advanced users, may I seek for some recommendation on filtering Landsat
> > >> reflectance outliers?

Nikos A (replying to self):
> > > (Thanks Yann)
> > > Reflectance is a unit-less ratio, ranging from 0 to 1.0.  So I can
> > > flatten everything >1.0 to 1.0.
 
Markus N:
> > I would run r.neighbors with a 3x3 or 5x5 moving window and averaging
> > or likewise. The question is always: are they outliers or data?
 
> Wild guessing: too high to be data!? 
> Nikos

Coming back to this! I think that Markus' suggestion was/is "correct". 

However, at the time of having to decide what to do, I simply flattened them. 
They were only a few though, not a real issue.

Anyhow, just for completeness an "official" definition of reflectance,  here 
is one copy-pasted from 
<http://landsat.gsfc.nasa.gov/references/glossary.html>:

--%<---
reflectance - A measure of the ratio of outgoing to incoming radiation 
calculated by converting a radiometrically calibrated image to an innate 
characteristic of the target being observed.  Calibrated at-satellite spectral 
radiance is converted to unitless reflectance by separating out the 
atmospheric component of the reflective band radiance and assuming that the 
target is a Lambertian reflector, re-radiating incident solar radiation 
equally in all directions. In general, reflectance is a function of incident 
angle of the energy, viewing angle of the sensor, spectral wavelength, and 
bandwidth, and the nature of the object. Also see planetary albedo, 
bidirectional reflectance and atmospheric correction. (Source: Dr. John 
Barker)
--->%--

Nikos