Questions for the Techies on PS3 and displacement mapping

btdvox

Newcomer
I just wanted to know if most games are using this or rather 3d bump mapping I know oblivion used it on the 360 version and i would assume that ps3 would but i just havent been able to confirm it at all, also i saw somewhere that nintendo patented it?? or something. Im not positive. any help would be apprciated thanks
 
I expect Oblivon didn't use displacment mapping or 3D bump mapping... I expect you talking about parallaz mapping AKA offset bump mapping and if you Unreal PR machine virtual displacement mapping...

In which case then yes at least one (HS) and I expect many other PS3 titles will use it...
 
I though the after the last GDC (where it seems like it could be disp from the bad vids) Evolution studios transitioned to using some kind of dynamic parallax mapping in Motorstomr instead of disp as the terrain geometry doesn't seem to deform anymore?

I've been wanting to see if it's gone completely, used in spots or if my eyes just decieve me altogether...I guess the demo will be available soon enough.
 
I expect Oblivon didn't use displacment mapping or 3D bump mapping... I expect you talking about parallaz mapping AKA offset bump mapping and if you Unreal PR machine virtual displacement mapping...

In which case then yes at least one (HS) and I expect many other PS3 titles will use it...

Deano: Any chance you can link me to a quick tutorial on the differences between displacement/3D-bump/offset-bump/parallax mapping? The two middle ones, in particular, I don't know the difference between, and this seems to be a recurring theme now. PEACE.
 
I though the after the last GDC (where it seems like it could be disp from the bad vids) Evolution studios transitioned to using some kind of dynamic parallax mapping in Motorstomr instead of disp as the terrain geometry doesn't seem to deform anymore?

I've been wanting to see if it's gone completely, used in spots or if my eyes just decieve me altogether...I guess the demo will be available soon enough.
The terrain looks like it still deforms in the last set of vids. The demo seems to be running on a hard-packed surface, but the trailer showed enough mud and dirt to show the deformations are still there. PEACE.
 
The terrain looks like it still deforms in the last set of vids. The demo seems to be running on a hard-packed surface, but the trailer showed enough mud and dirt to show the deformations are still there. PEACE.

The most recent stuff we heard was that graphically they were using parallax occlusion maps for the mud tracks and Cell was calculating the effect of ruts in realtime. I don't believe they are actively deforming the terrain/poly data any more.

While real deformable terrain would be great, I am guessing it is significantly cheaper to go the short cut and still get a great effect (probably better effect as you can do more with lower resource requirements). If they don't let you look too closely and/or at flatter angles it would be hard to tell the difference if they have done a good job of rut intersection. Interestingly Sega has announced deformable terrain for the new Sega Rally, so it will be interesting if they have real deformable terrain or simulate the effect with a texture mapping technique.
 
Modelling those ruts in realtime would quickly add huge amounts of polygons I think. Perhaps they could model a basic groove and add parallax to that, so the wheel sits in the rut more effectively, but I'm not expecting fully modelled dirt-tracks this gen.
 
Any chance you can link me to a quick tutorial on the differences between displacement/3D-bump/offset-bump/parallax mapping? The two middle ones, in particular, I don't know the difference between, and this seems to be a recurring theme now. PEACE.
You probably don't realize the differences between the latter 3 because there aren't any differences per se. Well, there are a host of similar techniques that go by various names, and they all get confused with each other. First of all, real "displacement mapping" means actually changing the geometry, which means if you want it to look reasonably good, then your geometry density had better be damn good in relation to your map resolution.

Parallax mapping is the more generic name if you go by the literature, but it's also called offset mapping by some. It's simply the idea of transforming the view vector into tangent space and using that as the basic direction along which you change texture coordinates according to a height map value. So (depending on the direction of offset) you basically take height map value * some scale / TangentView.Z = some value, and some value * TangentView.XY * scale factor relating to texture size = the UV offset. If you use it in combination with bump/normal mapping, it might be called offset-bump or 3d-bump especially, but I don't think that's really necessary since it's essentially a precondition for all texture-based operations including normal mapping.

Some people will also take into account that once you've offset the texture coordinates, the height in the height map is different from what you originally sampled, so they repeat the process a few times (doing fixed-point iteration) or do some ray stepping along the view vector to get a more "correct" offset position by determining where the view vector intersects the height map, and this might be called "steep parallax mapping" or "relief mapping" depending on what they do to get that result, and the results are certainly better, but it's obviously a lot more costly, and so you have to question whether it's worth it or not.

I've often found in practice that simple parallax mapping or a few (3 or 4) fixed-point iterations plus a little bit of extra geometry to accentuate shape (assuming that you're not trying to simulate small variations on an otherwise flat surface, but something that has some substantial shape to it) often gives the most convincing results.
 
Modelling those ruts in realtime would quickly add huge amounts of polygons I think. Perhaps they could model a basic groove and add parallax to that, so the wheel sits in the rut more effectively, but I'm not expecting fully modelled dirt-tracks this gen.

From the screens I have seen it appears the that the wheels sink (clip) just below the surface of the mud. I cannot say that they are not creating a basic rut and augmenting the terrain mesh in realtime, but it seems to me that they just have the wheels clip below the surface in muddy areas. I guess we will get more definitive answers when the game is released.
 
I'd agree with everything ShootMyMonkey says except I'd keep 3d-bump-mapping for the old style blinn bump-mapping but using 3D bump-map texture itself. Used a bit back in the day for bumpy marble etc.
 
Oh im very sorry yes was talking about offset mapping- What are the differences between displacement, and offset mapping? It looks great.
 
btw that lair shot is wicked- and yes the floor thats what i mean- oblivion used alot of it- but it didntlook as good as that pic!!!
 
Lair has parallax mapping, click here to see it.

That picture is both beautiful and kind of shows the problem with the technique: with deep parallax you get visual depth, but the actual poly surface is still flat and when objects (people, vehicles, etc) go across them they tend to walk across the poly surface and not along the depth of the parallax map. ATI's Toy Shop video, which also has excellent stonework, shows the same issue when the rain hits the rockwork on the ground, with appearant splashes above the cracks. Not a huge deal, but a good reason not to use parallax mapping that is either really deep (if objects come in contact with it), when gamers can look nearly flush to the poly plane, or can get really close in general. Of course in a fast paced game you don't really have time to notice these things and the general approximation looks good in motion.
 
That picture is both beautiful and kind of shows the problem with the technique: with deep parallax you get visual depth, but the actual poly surface is still flat and when objects (people, vehicles, etc) go across them they tend to walk across the poly surface and not along the depth of the parallax map.
the other big problem is that like all new effects, it's overdone, so you get stones laid with gullies 6" deep and 4" across and if the surface were really like that, soldiers would be tripping up all over the place!
 
btw that lair shot is wicked- and yes the floor thats what i mean- oblivion used alot of it- but it didntlook as good as that pic!!!
Oblivion used "standard" parallax mapping while that Lair shot looks like steep parallax mapping to me. Though you can't say for sure from just 1 screenshot - in fact, you can't say for sure without seeing the code ;)
 
You probably don't realize the differences between the latter 3 because there aren't any differences per se. Well, there are a host of similar techniques that go by various names, and they all get confused with each other. First of all, real "displacement mapping" means actually changing the geometry, which means if you want it to look reasonably good, then your geometry density had better be damn good in relation to your map resolution.

Parallax mapping is the more generic name if you go by the literature, but it's also called offset mapping by some. It's simply the idea of transforming the view vector into tangent space and using that as the basic direction along which you change texture coordinates according to a height map value. So (depending on the direction of offset) you basically take height map value * some scale / TangentView.Z = some value, and some value * TangentView.XY * scale factor relating to texture size = the UV offset. If you use it in combination with bump/normal mapping, it might be called offset-bump or 3d-bump especially, but I don't think that's really necessary since it's essentially a precondition for all texture-based operations including normal mapping.

Some people will also take into account that once you've offset the texture coordinates, the height in the height map is different from what you originally sampled, so they repeat the process a few times (doing fixed-point iteration) or do some ray stepping along the view vector to get a more "correct" offset position by determining where the view vector intersects the height map, and this might be called "steep parallax mapping" or "relief mapping" depending on what they do to get that result, and the results are certainly better, but it's obviously a lot more costly, and so you have to question whether it's worth it or not.

I've often found in practice that simple parallax mapping or a few (3 or 4) fixed-point iterations plus a little bit of extra geometry to accentuate shape (assuming that you're not trying to simulate small variations on an otherwise flat surface, but something that has some substantial shape to it) often gives the most convincing results.
Hey, thanks for that. That clears up a lot of confusion I had over the different techniques. I get the math too. My eyes always used to glaze over when it came to vector math for 3D. I viewed it as a programmer problem. ;) But your explanation with those terms made it easy to follow. PEACE.
 
that lair screenshot doesnt use standard parallax mapping eg like kameo
its something better
its not quite as good as displacement mapping though
my vote would be a variation of Steep Parallax Mapping as well
btw lair looks quite good, what were all the complaints about it for?
 
Back
Top