[GRASS5] [bug #3166] (grass) bad target region calculations with i.rectify order=3

[Markus Neteler]

On Wed, Apr 20, 2005 at 07:00:11AM +0200, Request Tracker wrote:
> this bug's URL: http://intevation.de/rt/webrt?serial_num=3166
> -------------------------------------------------------------------------
> 
> Subject: bad target region calculations with i.rectify order=3
> 
> Hi,
> 
> i.rectify (both 6.0.0, 6.1cvs) is creating rasters with incorrect regions and
> resolutions when doing a 3rd order transform. It works fine with 1st or second
> order transforms. 'i.rectify -c' (or lower order) is needed to get a good
> transform.
> 
> Starting with unreferenced simple XY map (PNG imported with r.in.gdal)
> and re-regioned to positive values with r.region, keeping res=1 pixel.
> 
> Target location is lat/lon (both wgs84 and none/intl datums fail)
> 
> e.g. eastern boundary ends up being +2 degrees east of what it should be,
> resolution is calculated from e-w/cols so that ends up very coarse...
> southern boundary is wrong too.
> 
> CRS_georef() in imagery/i.rectify/crs.c ?
> 
> more info available upon request.

Hamish,

you may compare with GDAL:

 gdal/alg/gdal_crs.c

- is the algorithm identical?
- does gdalwarp produce the same (error)

If GDAL performs better for above operation, its improvements
might be worth backporting.

Markus