Opened 13 years ago
Closed 9 years ago
#3798 closed defect (invalid)
gdalwarp resampling errors with "cubicspline" and "lanczos"
Reported by: | maxbuch | Owned by: | warmerdam |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | Algorithms | Version: | 1.6.3 |
Severity: | major | Keywords: | resampling lanczos cubicspline gdalwarp |
Cc: | antonio |
Description
Use of cubicspline or lanczos results in a (1,1) pixel offset in the output, with respect to the input. None of the other options exhibit this behaviuor.
Machine & OS: Centos 5 linux on a x86_64, 8 core Gdal version: 1.6.3 Python version : 2.5.4
My test case was a resampling of Radarsat2 from a VRT file pointing to the original "product.xml" to produce a GeoTiff.
The command line I used was : gdalwarp -r lanczos -of GTiff -tps -et 0.020000 --config \ GDAL_CACHEMAX 100 -wm 100 -co TILED=YES \ sgf/0510-ortho-img.vrt sgf/0510-ortho-lanczos.tiff
These files are all in WGS84 Geographic coordinates and had a pixel size of about 0.00014 degrees.
Registration of the images with all resampling modes was assessed with OpenEV against the original.
Change History (3)
comment:1 by , 13 years ago
Priority: | high → normal |
---|
comment:2 by , 12 years ago
Cc: | added |
---|
comment:3 by , 9 years ago
Resolution: | → invalid |
---|---|
Status: | new → closed |
Closing in lack of test data. Labeling as invalid which in this case means "invalid as a bug report".
do you have any test data (and possibly small) to reproduce this ?