[Proj] Pixel Shifts When Warping

[Martin Desruisseaux]

Hello

Le 17/05/15 06:44, Yuta Sato a écrit :
> Do you have any idea on which is more precise: gdalwarp or MRT? Or
> there is any way in gdalwarp to get the same image as MRT gives?

I do not know for this case. But generally speaking I think that the
question of projection library accuracy is not yet solved. The
/Geospatial Integrity of Geoscience Software/ (GIGS) test suite is one
attempt to answer that kind of question, but it is not yet widely
implemented as far as I know. We have a partial implementation in GeoAPI
[2] and we tried those tests on Proj.4 through JNI bindings. I found a
large difference between Proj.4 and GIGS for "Lambert Conformal
(Belgium)" projection, but I'm not sure that I didn't made any mistake
with Proj.4 parameters.

More recently I found an other difference between Proj.4 and Circés (the
official software from French mapping agency for coordinate
transformations in France) for the transformation from "Nouvelle
Triangulation Française" (EPSG:4275) to WGS 84 (EPSG:4326). I got datum
shift in opposite directions, but again I'm not sure that I did no mistake.

I feel that for answering your question with more confidence, we would
need to push forward the GIGS test effort or something equivalent. This
is a little bit different than the tests provided by the projects
themselves, which sometime tend to be anti-regression tests rather than
test for correctness.

    Martin


[1] http://www.iogp.org/geomatics#2521115-gigs
[2] http://www.geoapi.org/geoapi-conformance/index.html

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20150517/da51d3a7/attachment.html>