Even, Checked the code and it seems that the criteria for a nodata (ND) pixel to be interpolated is that its distance to a valid pixel to be less than dfMaxSearchDist (rasterfill.cpp:803)
This means that a ND pixel outside the valid image boundary will be marked to be interpolated if it is close enough to the valid image area. The current value for dfMaxSearchDist in the geolocation array backmap is hard-coded to 3 pixels, which creates the "border" effect I mentioned in issue #4707. I could try for instance removing the pixel from the filling mask if in the last scanned line a direct pixel neighbors' LastY is nNoDataVal but I'm not sure that will be a general solution. Fred PS. I submitted a DRAFT PR with the WIP changes in the backmap implementation and had no idea that this would start all CI workflows. Please let me know if this is a problem, I may delete the PR and submit a new one when finished. _______________________________________________ gdal-dev mailing list [email protected] https://lists.osgeo.org/mailman/listinfo/gdal-dev
