[Fwd: Re: [gdal-dev] Warping an already georeferenced image with control points

Jan Hartmann j.l.h.hartmann at uva.nl
Sat Feb 7 15:10:36 EST 2009


No, this doesn't seem to work. Although the coordinate system *is* 
retained when I add -a_srs to the gcps, the boundaries of the input file 
are discarded: they are reset to the pixel dimensions of the file. Is 
this a bug?

Jan

Frank Warmerdam wrote:
> Jan Hartmann wrote:
>> Hi,
>>
>> I am trying to georeference old maps in two steps: first 
>> georeferencing them from the old projection to a modern one, and then 
>> ameliorating the georeferenced image by rubber-sheeting it on the 
>> base of control points, mostly triangulation points for which the 
>> coordinates in the old and new maps are known.
>>
>> The second step causes trouble: if I add  control points to the 
>> georeferenced image with gdal_translate -gcp , the projection 
>> information gets lost.
>>
>> Yet, AFAICS it should be possible: if I use gdaltransform -s_srs 
>> EPSG:28992 -t_srs EPSG:28992 -gcp ... , I can use control points with 
>> coordinates in EPSG:28992 and get a correct transformation.
>>
>> So: regarding this behavior  of gdal_translate, to discard projection 
>> information when control points are added: is it a bug or is it the 
>> expected behavior? In the last case, how can I rubber-sheet an 
>> already georeferenced map?
>
> Jan,
>
> If you use the -a_srs switch of gdal_translate in combination with the
> -gcp switch to add gcps, the GCPs should be recorded as being in the
> coordinate system described by -a_srs.  Because there is no reason to
> believe that the current image coordinate system is related to the
> coordinate system of the GCPs, no effort is made to use the existing 
> image
> coordinate system as the default GCP coordinate system.
>
> By the way, setting GCPs with a GCP coordinate system on an image is 
> likely
> to not work with most GDAL supported formats, and the coordinate system
> related aspect may also be supported in a variable fashion depending 
> on the
> format driver's implementation.
>
> Hopefully things work well with the GeoTIFF format, which I assume you 
> have
> been using.
>
> Best regards,




More information about the gdal-dev mailing list