GCP Transformer using thin plate splines

[cpegel]

Working with SAR images, the data is often not geocoded with a CRS+AffineTransform but with a grid of GCPs that describe the ground range geometry. For precise georeferencing, GDAL supports using thing plate spline (-tps option for gdalwarp or METHOD=GCP_TPS option of GDALTransformer), for example:

ds = gdal.Open(geotiff_path)
tr = gdal.Transformer(ds, None, ["METHOD=GCP_TPS"])
# forward transform:
tr.TransformPoint(0, row, col)
# backward transform:
tr.TransformPoint(1, x, y)

Rasterios GCPTransformer seems to use GDALs less precise GCP transformation and doesn't take any options to use TPS instead.

Is it possible to use thin plate spline transformations in rasterio? For now I use the above code to get the transformer using the osgeo bindings, but it is discouraged to import both osgeo and rasterio in the same python script.

[underchemist]

@cpegel rasterio only supports polynomial GCP transforms currently. Using TPS would require implementing the option to use GDAL's TPS specific transformer