[Gdal-dev] projection coordinates support

[Frank Warmerdam]

April Shimitz wrote:
> Frank,
> 
> We've been having some issues with accuracy when using the
> Get/SetGeoTransform methods to pass our corner coordinates through GDAL.
> The transformation coefficients retrieved with GetGeoTransform are not as
> precise as the values given to SetGeoTransform. I noticed your note on the
> GDAL doc. that you are not completely happy with this mechanism either. 

April,

Could you point me to this comment in the docs?  It doesn't ring a bell.
I am not sure why there would be a problem with the accuracy (precision?)
of Get/SetGeoTransform() assuming the image is actually sampled on an
exact grid in the image projection.

 > Do you have any plans for improvement of this in the near future? Since LAS
> only needs the functionality to store the coordinates at the extents of the
> image, we'd like to have the ability to do that more directly. We have been
> considering developing a new class to store/retrieve the corner coordinates
> and would like to move on that now. Our plan for the class is a simple
> object to contain the exact projection corner coordinates, removing the
> usage of an intermediate transformation equation, thereby returning to the
> caller the exact values that were set in the object. This class would be
> useful to a number of other format drivers as well I am sure (including
> ones we plan to develop in the future), so I'd like to implement it as part
> of the GDAL core. Let me know if we should move forward with this plan or
> if you have any ideas/input.

I think that if the GeoTransform isn't an exact way of describing the image
georeferencing you should be using the GCP API.  This allows you to associate
an arbitrary number of control points with the image.  Each control point
associates a georeferenced position with a given pixel/line location.  In
your case these might be corner locations.  For other images GCP coverages are
used to approximate more complex underlying georeferencing models.

Let me know if you need pointers on the GCP API of if you know of a reason
why it isn't adequate to represent the corner coordinates you want to offer.
I am hesitent to introduce any additional georeferencing mechanisms if the
existing ones can be used.  It is already hard to write a GDAL application
that understands the two existing georeferencing mechanisms (geotransform
and gcps).

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    | Geospatial Programmer for Rent