[GRASS-user] New addon: r.denoise. Denoise topography including SRTM

[Frank Broniewski]

On Wednesday 16 September 2009 12:47:52 John A Stevenson wrote:
> Hi Frank
>
> Frank Broniewski wrote:
> > I want to try your module with my srtm data, but I always get out of
> > memory messages. Maybe I am just to ambitious with my region settings :-)
> > ,
>
> How big a region are you using?  Have you had any luck testing it on a
> smaller region?
>
> I have only ever run out of memory on a 24 million point region, but can
> happily (if slowly) denoise it if I halve the resolution (~6 million
> points).  I have 4Gb RAM.
>
> > but could
> > you provide some hints on general memory usage?
>
> I've emailed Xianfang Sun who wrote the denoising part, (I just wrote
> the GRASS interface) and will let you know what he says.
>
> > Or should I program a script for a moving window so I get smaller
> > regions? Are there any problems then on the connecting borders, e.g.
> > differnently smoothed surfaces at the edges
>
> Each of the cells moves individually based on their neighbouring points,
> so the edges would be smoothed differently.  The magnitude of the
> differences would be very small, but many algorithms e.g. slope, shaded
> relief are very sensitive to them, so I think that they would show up.
>
> Cheers
>
> John

Hi John,

thanks for your answers so far. The region was quite large, a good part of 
europe and I only have 1GB of RAM. I chose a much smaller subregion and the 
algorithm worked without a problem. It was really nice to see that I was able 
to calcualte a good flow direction grid from a former problematic region now 
with reasonable results.

If there are differences at the borders, I will have to see what I can do 
about it. Some low memory swith like the r.watershed module has would be nice 
... but I have no clue how to implement this. But I haven't investigated yet 
if the mdenoise script has some more options.

Thanks

Frank