# [gdal-dev] GDAL Cubic Interpolation

5 messages
Open this post in threaded view
|

## [gdal-dev] GDAL Cubic Interpolation

 Hi,I subsampled a GLOBE image of elevation data by a subsampling factor of 12 to a resolution of 0.1 deg x 0.1 deg on integer multiples of 0.1 deg. The original GLOBE image is: Latitude from +90-1/240 deg to -90+1/240 deg in 1/120 deg steps, and Longitude from -180+1/240 deg to +180-1/240 deg in 1/120 deg steps. I understand that the cubic interpolation method in gdalwarp accounts for the subsampling factor, so 48x48 adjacent pixels influence the target pixel for a subsampling factor of 12. Does gdalwarp assume the original image is a sphere, where 48x48 adjacent pixels influence all target pixels, or a rectangle, where a limited number of pixels influence the target pixels at the top, bottom, left, and right edges of the subsampled image? If gdalwarp assumes the original image is a rectangle and uses a modified interpolation method at the four edges of the subsampled image, the modified method affects two rows and two columns at the top, bottom, left and right edges of the subsampled image. If gdalwarp assumes the original image is a rectangle, is there a command option so gdalwarp considers spherical wrapping when interpolating? Alternatively, I assume I could augment the original GLOBE image with the appropriate additional rows and columns on all four edges. If I did that, how do I tell gdalwarp the extent of the augmented input image?Kind regards,Harvey _______________________________________________ gdal-dev mailing list [hidden email] https://lists.osgeo.org/mailman/listinfo/gdal-dev
Open this post in threaded view
|

## Re: GDAL Cubic Interpolation

 Harvey,   > > I subsampled a GLOBE image of elevation data by a subsampling factor of 12 > to a resolution of 0.1 deg x 0.1 deg on integer multiples of 0.1 deg. The > original GLOBE image is: Latitude from +90-1/240 deg to -90+1/240 deg in > 1/120 deg steps, and Longitude from -180+1/240 deg to +180-1/240 deg in > 1/120 deg steps. > > I understand that the cubic interpolation method in gdalwarp accounts for > the subsampling factor, so 48x48 adjacent pixels influence the target pixel > for a subsampling factor of 12. > > Does gdalwarp assume the original image is a sphere, where 48x48 adjacent > pixels influence all target pixels, or a rectangle, where a limited number > of pixels influence the target pixels at the top, bottom, left, and right > edges of the subsampled image?   A rectangle, with fallback to bilinear resampling at edges   > If gdalwarp assumes the original image is a > rectangle and uses a modified interpolation method at the four edges of the > subsampled image, the modified method affects two rows and two columns at > the top, bottom, left and right edges of the subsampled image. If gdalwarp > assumes the original image is a rectangle, is there a command option so > gdalwarp considers spherical wrapping when interpolating?   Nothing directly straightforward. Your below plan is the a reasonable solution.   > Alternatively, I > assume I could augment the original GLOBE image with the appropriate > additional rows and columns on all four edges. If I did that, how do I tell > gdalwarp the extent of the augmented input image?   You'll have to recreate a properly referenced input image, for example with   gdal_translate -ullr upperleftx upperlefty lowerleftx lowerlefty   Even   -- Spatialys - Geospatial professional services http://www.spatialys.com_______________________________________________ gdal-dev mailing list [hidden email] https://lists.osgeo.org/mailman/listinfo/gdal-dev
Open this post in threaded view
|

## Re: GDAL Cubic Interpolation

 Alternatively,  use GMT's grdsample with -fg option to force the knowledge (if it's not already in the file) that the Earth is round.http://gmt.soest.hawaii.edu/doc/5.4.2/grdsample.html Harvey,   > > I subsampled a GLOBE image of elevation data by a subsampling factor of 12 > to a resolution of 0.1 deg x 0.1 deg on integer multiples of 0.1 deg. The > original GLOBE image is: Latitude from +90-1/240 deg to -90+1/240 deg in > 1/120 deg steps, and Longitude from -180+1/240 deg to +180-1/240 deg in > 1/120 deg steps. > > I understand that the cubic interpolation method in gdalwarp accounts for > the subsampling factor, so 48x48 adjacent pixels influence the target pixel > for a subsampling factor of 12. > > Does gdalwarp assume the original image is a sphere, where 48x48 adjacent > pixels influence all target pixels, or a rectangle, where a limited number > of pixels influence the target pixels at the top, bottom, left, and right > edges of the subsampled image?   A rectangle, with fallback to bilinear resampling at edges   > If gdalwarp assumes the original image is a > rectangle and uses a modified interpolation method at the four edges of the > subsampled image, the modified method affects two rows and two columns at > the top, bottom, left and right edges of the subsampled image. If gdalwarp > assumes the original image is a rectangle, is there a command option so > gdalwarp considers spherical wrapping when interpolating?   Nothing directly straightforward. Your below plan is the a reasonable solution.   > Alternatively, I > assume I could augment the original GLOBE image with the appropriate > additional rows and columns on all four edges. If I did that, how do I tell > gdalwarp the extent of the augmented input image?   You'll have to recreate a properly referenced input image, for example with   gdal_translate -ullr upperleftx upperlefty lowerleftx lowerlefty   Even   -- Spatialys - Geospatial professional services http://www.spatialys.com_______________________________________________ gdal-dev mailing list [hidden email] https://lists.osgeo.org/mailman/listinfo/gdal-dev