[Gdal-dev] gdalwarp: difference between -tps and -order 1

[Jan Hartmann]

Hi,

I'm georeferencing historical maps that are not geometrically correct. 
When I warp them with control points to a modern map, using a first 
order rigid transformation (gdalwarp -order 1), the resulting, 
georeferenced map will keep its original form. This means that the 
position of the control points on the georeferenced historical map is 
different from those on the modern map. Warping with  a thin plate 
spline (gdalwarp -tps) results in a georeferenced map with the control 
points exactly on the same place on both maps, which is what I want of 
course.

I can see the difference between both methods visually, but I would like 
also to obtain the numerical coodinates of the control points after 
warping them with both methods. That way I can compute the differences 
between both methods and construct a distortion map. So essentially I 
would like to have a warping program that does not output a raster, but 
produces the coordinates of the transformed input control points. I 
already did this for the first order transformation (pretty elementary), 
but how could this be done for the thin plate spline transformation? 
Where should I look in the gdal codebase to extract the algorithm? 
Alternatively, if an expert would program this, how much would that cost?

Jan