[postgis-users] Raster Processing and ST_Transform

[Pierre Racine]

> First problem, after reprojection, all the chunks have different alignements ("The
> two rasters provided do not have the same
> alignment") which is unfortunate. It would be nice to be able to ensure that
> when reprojecting all the objects get resampled into a same alignment (am I
> right that an "alignment" is a combination of a origin, pixel size and skewness?)
> so that they can be processed together in succeeding steps.

Alignment is the combination of any arbitrary pixel corner on a raster grid, scale (what some call pixel size) and skewness. We are discussing about that with Bborie. He might come up with a slight redesign of ST_Transform. Right now you can achieve what you want with ST_Resample if I understand well.

> I was hoping in my next steps to calculate slope and go from there, but I'm
> stopped at the transform step. Thoughts? While a general utility to re-sample
> any collection of rasters into the same basis would be good, multiple re-
> samplings are not great, so inherently re-sampling operations like st_transform
> should probably have a way to force the output alignment.

I think we all agree on that. BTW You will have to ST_Union everything before passing it to ST_Slope as all our neighboring functions do not transparently work on a tiled coverage. For now, calling ST_Slope on a tiles coverage should result in tile border effects. The way to get all the neighboring functions to work on tiled coverage is still an open question.

Pierre