In paragraph 4.2 (Page 5 and 6) they have an idea that solves texturing issues by calculating the mipmaps in a special way. This seems to solve the repetition pattern that is often observable when using texture splatting. To be true, I don’t understand much of what they say there… maybe someone can enlighten me?
As I understand article’s authors, they don’t try to beat repeating pattern, but they simply fade total mip-map color to an average one. So, texture on far mip-levels looks like having one solid color.
I don’t think, that this might be a general advice. Using their scheme in fractal terrains is quite good, and the details are achieved due to shading and sufficient tesselation.
But for general terrains, where texture information is often necessary at far distances, this solution is bad, IMO.
My main problem is finding a good texturing algorithm for terrain, without going for a full virtual texture implementation. What I can think of is using splatting at near distances and fading to some blurry texture stretched over the whole terrain at the distance.
And yes, adding Wang would be nice. But doing a texture to wang tile processor is far beyond my skills.
The main idea is, to make level editing and texturing for designers as easy as possible and have a fairly moderate download size for final levels. So splatting is the way to go, I think. Virtual textures are nice on paper, but they place to much burden on the designers…