Voxels and Parallax Mapping

fearsomepirate

Dinosaur Hunter
Veteran
Call me crazy, but I think the end result of parallax mapping looks awfully similar to voxels. But from what I understand, it's computationally expensive and doesn't actually alter the silhouette of the object. While I realize voxel terrain engines died an unloved death after Novalogic abandoned them and are too memory-intensive to serve as an engine paradigm, has anyone ever thought of the possibility of using voxels as an effect to be used as a texture effect in conjunction with normal maps? If you've already got the code to handle voxels, it seems like it wouldn't be too hard to extract their heights out of the data already encoded in a normal map (or you could use your regular boring height maps), which would keep your data throughput from going nuts. And I've read that the blockiness we saw in Comanche and Delta Force games can be smoothed out using bilinear filtering and/or Marching Cubes.

So is this anything anyone's thought of looking at, or is it computationally even more expensive than parallax mapping? Or would it just be such a coding nightmare that it wouldn't even be worth doing, since we've already got parallax mapping down?
 
fearsomepirate said:
But from what I understand, it's computationally expensive and doesn't actually alter the silhouette of the object.

Parallax mapping is cheap. It comes at 2 extra ALU instructions and one TEX.
 
Well, the 2-instruction parallax mapping hack only looks good with very smooth heightmaps. The more advanced (and expensive) techniques are essientially using exactly the same technique as the ol' heightmap voxel renderers of yore.
 
Humus said:
Parallax mapping is cheap. It comes at 2 extra ALU instructions and one TEX.

It certainly seems a lot more involved than that. One TEX and 2 extra ALU instructions sounds more like normal mapping and not parallax mapping unless I totally misunderstood the article that satein posted.
 
trinibwoy said:
It certainly seems a lot more involved than that. One TEX and 2 extra ALU instructions sounds more like normal mapping and not parallax mapping unless I totally misunderstood the article that satein posted.
The article refers to parallax occlusion mapping, which is an extremely expensive extension to parallax mapping to increase its quality. It can apply to the exact same art assets as parallax mapping.

Uttar
 
Uttar said:
The article refers to parallax occlusion mapping, which is an extremely expensive extension to parallax mapping to increase its quality. It can apply to the exact same art assets as parallax mapping.

Uttar

Ah, thanks. So in regular parallax mapping implementations how do they make sure that the "occluding" surface sample is not overwritten by the "occluded" one?
 
trinibwoy said:
Ah, thanks. So in regular parallax mapping implementations how do they make sure that the "occluding" surface sample is not overwritten by the "occluded" one?
They don't :)

Basically regular parallax mapping sort of flattens out whenever occlusion would have been a problem. Regular offset mapping doesn't really do raytracing, but it's sort of hard to describe exactly how it works. In a way the occluded surface moves so it's never occluded. You can check out this thread for Humus' demo: http://www.beyond3d.com/forum/showthread.php?t=8498&highlight=offset. Go close to the ground, and you'll see what I mean.
 
Mintmaster said:
They don't :)

Basically regular parallax mapping sort of flattens out whenever occlusion would have been a problem. Regular offset mapping doesn't really do raytracing, but it's sort of hard to describe exactly how it works. In a way the occluded surface moves so it's never occluded. You can check out this thread for Humus' demo: http://www.beyond3d.com/forum/showthread.php?t=8498&highlight=offset. Go close to the ground, and you'll see what I mean.

Thanks!
 
Back
Top