[gdal-dev] Warping onto an image with only GCPs

Anton Korosov anton.korosov at nersc.no
Fri Dec 16 09:56:20 EST 2011


I'm not sure but probably my question is related to the one below.
I'm trying to mosaic one MODIS images onto another with GDAL:

-b 1 /data/modis1.tif

-b 1 /data/modis2.tif

gdalwarp modis1.tif modis2.tif

Bu the results looks weird - the first image is on top of the second one 
but it is very small, distorted and in the top-left corner (though they 
cover approximately the same area).

Do I understand right that such operation would work only when the 
target image georeference is defined by GeoTransform and not by GCPs?

Can I also ask if it is a big work to include ability of GDAL to warp 
onto grid with GCPs (or RPC or geolocation array) ? And if it is planned 
in some future?

Thank you!

On 12/15/2011 03:47 PM, Knut-Frode Dagestad wrote:
> Hi,
> Hróbjartur Þorsteinsson explained to me a near-hidden feature of GDAL,
> namely to reproject one image onto the coverage of another.
> Given two images:
> imageA.tiff
> imageB.tiff
> Then an empty image can be created from image A with e.g.
> $ gdal_translate -ot Float32 -scale 0 0 999 999 -a_nodata 999
> imageA.tiff domainA.tiff
> And then imageB can be warped onto the extent of this image simply with:
> $ gdalwarp imageB.tiff domainA.tiff
> domainA.tiff will then contain the data from imageB, exactly co-located
> with imageA.
> The problem is that this only works when imageA contains a geotransform,
> and not when it contains only GCPs.
> So my question is:
> Is there any workaround to warp data from one image onto another image
> which is geolocated by GCPs only?
> With the -to option of gdalwarp it is possible to pass some parameters
> to GDALCreateGenImgProjTransformer2(). In the documentation of this
> function I find the following option:
> "METHOD: may have a value which is one of GEOTRANSFORM, GCP_POLYNOMIAL,
> GCP_TPS, GEOLOC_ARRAY, RPC to force only one geolocation method to be
> considered on the source dataset."
> http://www.gdal.org/gdal__alg_8h.html#94cd172f78dbc41d6f407d662914f2e3
> But there is no corresponding option to force use of GCP_POLYNOMIAL for
> the *destination* dataset. I guess this is a bad sign?
> Thank you for any help!
> Knut-Frode
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/gdal-dev

More information about the gdal-dev mailing list