That is a surprise.
In all my tests I had no memory problems with gdal_retile. Sometimes I had
some memory problems with gdal_merge. The biggest input file for gdal_retile
was a 80 GB Erdas image (not compressed) and it works fine.
I have never testet with an ECW File and my tests run on a Linux Box or AIX
Box.
Perhaps the problem is in the ECW gdal driver. You can try to transform your
ecw file in another format using gdal_transform.
Gilles Bassière writes:
Christian Müller wrote:
gdal_retile.py definitely creates georeferenced tiles, I developed this
utility.
I did some experiments reprojecting the single tiles while retiling, but
it simply does not work. As a consequence I canceled the reprojecting
support.
The solution for your problem is:
1) Use gdal_merge to create the big image
2) reproject with gdal_warp
3) use gdal_retile to create your new tiles
I did a lot of experiments to avoid creating the big picture without
success.
I understand. In my case, gdal_retile.py eventually crashed with a
Python MemoryError. I assume it is unsurprising with a 7.5GB input for a
single Python process.
Even Rouault writes:
Le Monday 22 March 2010 19:35:47 Gilles Bassière, vous avez écrit :
Hi,
The problem with this workflow is that it creates non-georeferenced
tiles (I don't know if this is the intended behaviour of this command).
gdal_retile.py *does* create georeferenced tiles. I've just verified
it. What make you think the contrary ?
_______________________________________________
gdal-dev mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/gdal-dev
_______________________________________________
gdal-dev mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/gdal-dev
--
Gilles Bassière - Web/GIS software engineer
http://gbassiere.free.fr/
_______________________________________________
gdal-dev mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/gdal-dev
