[Gdal-dev] Resampling with gdalwarp

[Frank Warmerdam]

Chris G. Nicholas wrote:
> The real test is, as always, change detection. :-)
> 
> And I think (scientists out there, please correct me!) nearest-neighbor is as reasonable as anything. 
> 
> Now, it would be interesting if GDAL could have a flag where one could take 
 > the time/computation hit and do the transformation matrix for *every* pixel...
 > (like the MODIS reproject tool; see:
 > http://lpdaac2.usgs.gov/landdaac/tools/ modis/info/MRT_Users_Manual.pdf)

Chris,

When you talk about doing the transformation exactly, are you talking
about resampling or the geometric transformation?  The geometric transformation
can be done exactly for each pixel (as far as the underlying transform goes)
if the error threshold is set to zero.

I am not sure what it would mean to do a resampling transformation exactly.
Would this mean using a convolution matrix as large as the image?  I am not
really very savvy about convolution kernels.

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