Hi Onkelpoe,
Interesting. You may wish to have a close look at the normals map. To do this, select the 'terrain normals' tab, then select the '
View->Set view area...' menu item. Enter the X coordinates from 2972 to 3172, and the Y coordinates from 7068 to 7268. This will centre the view on the seams at x = 3072, y = 7168. This is what I find:
If I pan around, I find another set of seams crossing at x = 2048, y = 7168:
What this means is that the normal map has seams repeating at intervals of 1024 pixels. Assuming there's no bug in L3DT's normal map generator (which I don't immediately discount), this suggests there are subtle steps in your heightmap every 1024 pixels. Let's have a look in the heightmap (back at the first field of view, centred at x = 3072, y = 7168):
Nothing to see here, but then again, greyscale palettes aren't particularly good at showing subtle differences. What if we select the generally overlooked 'View->Display effects->Show gradient colouring' menu option? This calculates the difference from one pixel to the next, and overlays the magnitude of the changes as a colour using a rainbow palette:
Lo and behold, there are the subtle seams.
Has this heightfield ever been separate 1024x1024 pixel tiles? If so, what file format was used, and how were they joined together? If not, do you mind telling me which application was used to make the 8192 dataset?
I don't have a great solution for removing these seams, but you can try the '
Operations->Heightfeld->Edge blending->Blend tile edges' menu item. Set the X and Y tile size to 1024, and the blend depth to something small (e.g. 8 ). This will remove the step, but I just noticed it has the unfortunate side-effect of making the pixels on either side of the seam equal, so you trade steps into little flat strips. Hmmm....will have to think about it further.
Best regards,
Aaron.