[Gdal-dev] RE: GCPs and Thin Plate Splines - C++

Frank Warmerdam warmerdam at pobox.com
Sat Mar 10 14:27:22 EST 2007

chris pozsgai wrote:
> Frank,
> I have been successful in using GDALGCPsToGeoTransform, but
> I would like to have zero error at the GCPs.  I was hoping to use
> TPS transformer for this purpose.  My thought was to first get
> the GDAL GCP transformer working, then to get brave and try
> the TPS transformer.  I was unfortunately not able to get good results
> with the GCP transformer, and got nothing when I used the TPS transform
> method.  (My geotransform had parameters on the order of 10e67 in it.)
> I guess my main question would be if this is the correct approach to the
> problem.  If it is, I'll keep working at it.  If not, I can live with 
> using the
> least squares transform used in GDALGCPsToGeoTransform.


Ok, to isolate a point that might be investigatable, you report that
you do not get good results with the GCP transformer.  Could you elaborate
on that?

Note that the thin plate spline transformer is quite unstable mathematically
(in my opinion).  If you don't have good gcp coverage, or for various other
hard to predict reasons, the TPS results can go extremely screwy.  I use it
with caution.

Best regards,
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org

More information about the Gdal-dev mailing list