LOD Question

You need to adjust your error metric such that it will depend on more than just the distance from the camera.

In most terrain LOD systems, the error metric considers the visual error of removing a vertex, something like (y_of_vertex - interpolated_y_without_vertex) / distance. Unforetunately, this is not quite enough, as the error you''d get is not large enough to tell you that after a few more LOD reductions your whole mountain is going to dissapear. In other words, the error from each individual LOD reduction may be acceptably small, but their sum may well be too large.

If you have the option of preprocessing your tarrain, you could precalculate the total error accross all LOD reductions for each vertex, store it with the terrain, and work from there. If you don''t have that option, you could try to approximate this by calculating the error metric for two or three LOD reductions instead of only one. Although such calculations may be difficult, depending on what type of LOD you''re using and how it is implemented.

Another possible addition to the error metric, is the angle between the face normals and the camera view direction. The larger the angle, the less detail you need, perhaps even exponentially less. The idea is that you keep your silhouettes almost intact, but reduce the rest of the detail as much as you can.

You could also try rendering large distant objects (like mountains) as impostors. Although that may be problematic if you allow the camera to go high above the tarrain...

Michael K.,
Co-designer and Graphics Programmer of "The Keepers"

---

We come in peace... surrender or die!