2 | | I spent quite some time verifying the backmap generation in both Python and C++. These seem consistent for the aligned orientations that I used for the unit tests above - WENS, EWNS, WESN, EWSN. The outputs of GDALGeoLocTransform are consistent with input data in all those orientations. |
3 | | |
4 | | Any other leads or ideas? Could there be any issues in center of pixel / left edge of pixel in any of the resampling routines or routines that chunk up data based on values return by the geox/geoy to pixel/line output from GDALGeoLocTransform? Any other parts of the code that rounds fractional pixels up or down assuming some inherent orientation or where geotransform needs to be adjusted to ensure a half pixel shift? |
| 2 | I spent quite some time verifying the backmap generation in both Python and C++. These seem consistent for the aligned orientations that I used for the unit tests above - WENS, EWNS, WESN, EWSN. The outputs of GDALGeoLocTransform seem mostly consistent with input data in all those orientations. Will keep digging. |