• Advertisement

Archived

This topic is now archived and is closed to further replies.

new bumpmapping ?

This topic is 5151 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Advertisement
That looks SO impressive.

I''ve always found bump mapping only works with moving lights sources or heavy specular lighting, but with this it actually looks displaced from the plane...very effective.

Share this post


Link to post
Share on other sites
That does look impressive. And like most great graphics techniques it seems so obvious once its explained!

Share this post


Link to post
Share on other sites
It''s rather displacement mapping than bump mapping IMO.
The fragements are displaced, so...


-* So many things to do, so little time to spend. *-

Share this post


Link to post
Share on other sites
Well it displaces the bump-map and the texture map so effectively the fragments you end up with are displaced versions of the fragment you would've got without the bump-map and texture map displacement.

I also agree is looks very impressive, nice and simple to implement as well

[edited by - Monder on January 13, 2004 2:29:01 PM]

Share this post


Link to post
Share on other sites
hmm.. now i just have to figure out whats wrong. that demo is looking flat for offset mapping and the combination looks like pure bump mapping. fragment program loads without any errors, the lighting part is working well but somehow the new tex coords are the same as the original ones *confused*.. did i mention its a pain to debug vertex/fragment programs? ,-)

Share this post


Link to post
Share on other sites
This is completely OT, but a little while ago I thought it could be worth making a version of Mesa GL that allows you to easily debug fragment and vertex programs, I believe Mesa GL supports ARB_fragment_shader and ARB_vertex_shader, so all you need to add is the bit that allows you to debug shader code, it''d be even better if it could debug Cg or GlSlang and link straight into VC++, I don''t think I''ve got the skill to do it though, but if someone did do it, it could become a very useful tool (especially as vertex and fragment programs become longer and more complex).

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
It is NOT displacement mapping - that physically offsets your surfaces. This simply perturbs your lookup into your texture maps (you have to perturb all lookups, be they bump, colour, specular, luminousity, etc). It gives the effect of parallax by compressing opposing faces into almost impossible to see pixels.

This is one damned impressive technique and will no doubt find its way into many commercial engines. Bump mapping made great - I always hated it.

Share this post


Link to post
Share on other sites
Trienco, I tried the demo myself and got the same result as you. Offset mapping looks flat, and the combination looks just like bump mapping. I''m running it on a Radeon 9800, so it should support everything the demo needs. Wierd.

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
The difference is obvious if you move in close (hold RMB) and pan around (MMB)

Share this post


Link to post
Share on other sites
yeah, you have to move to see it. and go close..

its espencially great if you DON''T have the bumpmapping (the lighting), and you get close to the brick wall, and look around..

it looks really impressive.

its indeed a very cheap effect. but the resulting gain in image quality is imense. thats what we always ask for, not? cheap and good at the same time.




If that''s not the help you''re after then you''re going to have to explain the problem better than what you have. - joanusdmentia

davepermen.net

Share this post


Link to post
Share on other sites
The next step is to allow the dispaced bumps to self shadow from arbitrary lightsources (cheaply - something better then horizon mapping)... Very good technique.

Share this post


Link to post
Share on other sites
quote:
Original post by treething
Trienco, I tried the demo myself and got the same result as you. Offset mapping looks flat, and the combination looks just like bump mapping. I''m running it on a Radeon 9800, so it should support everything the demo needs. Wierd.


why do i get the feeling that it might have to do with the fact, that we are using the same card? guess i''ll really have to take it apart and check step by step where things are going wrong.

Share this post


Link to post
Share on other sites
Very nice,really looks like displacement mapping!
but it''s already known from 1980 or so,isn''t it?
(texture coordinates displacement,part of cheap but good tricks with bump mapping)

(googled : probably isn''t. But lots of "abstract" blabla pdf''s i don''t have time to read )
Computer graphics aren''t explored very well, and it''s possible that it''s new...

I got idea: render to billboards that have this mapping. Then you could rotate by some angle till you need to re-render again.
Interpolate(blend) between 2 views. Trick to render very detailed geometry at high framerates.

Heh,i already had idea doing it myself for landscapes,for far planes...

Regards,

Dmytry Lavrov.

Share this post


Link to post
Share on other sites
This thing allows to use much less polys to get same image errors. I mean,that if we assume that there''s actual surface,we could get smal color error by using simple bumpmaps,and to get smaller _coordinate_ error,we could use this trick. That allow to reduce "ideally needed" polygon count,without reducing the quality. I mean,that for 1/256 color error w/o bumpmapping we need really lots of poligons, with bumpmapping we need less polygons but introducing some pixel coordinate erro ,and with this "cheap trick" to get 1-pixel error we need even lesser triangle count,by orders of magnitude.
Very good thing.

Share this post


Link to post
Share on other sites
as very busy on many things yesterday, didn''t take time to really study the tech, indeed it''s not displacement mapping.

Looks great on the pics, no hardware to test it, so don''t know if it has any aliasing problems.


-* So many things to do, so little time to spend. *-

Share this post


Link to post
Share on other sites
quote:
Original post by Dmytry
I got idea: render to billboards that have this mapping. Then you could rotate by some angle till you need to re-render again.
Interpolate(blend) between 2 views. Trick to render very detailed geometry at high framerates.



Actually, I saw a talk at GDC 2002 about doing exactly this. They were using it to render volumetric clouds. And I believe their work was based on work from a previous GDC, or maybe it was some European conference, about doing something similar for rendering vast amounts of vegetation.

It is a good idea though!

-John

Share this post


Link to post
Share on other sites
quote:

Actually, I saw a talk at GDC 2002 about doing exactly this. They were using it to render volumetric clouds. And I believe their work was based on work from a previous GDC, or maybe it was some European conference, about doing something similar for rendering vast amounts of vegetation.

It is a good idea though!

-John



Without "cheap trick" based displacement mapping you need to re-render too often,imo. And that displacement mapping allow much bigger angle limits.
I also had some ideas about doing something similar to that displecement mapping,but not so simple. My idea was to allow biger angle limits for _very_ sharp objects like trees or even terrain panoram(i mean,i wanted to draw medium-far panoram that if you move,panoram looks like 3d,and that it will be possible to render photorealistic panoram in non-realtime).I''m still developing that.

For clouds there should be something very different from displacement mapping because clouds are semi-translucent.

----
http://dmytrylavrov.narod.ru/voxfoge.htm

Share this post


Link to post
Share on other sites
quote:
Original post by Trienco
why do i get the feeling that it might have to do with the fact, that we are using the same card? guess i''ll really have to take it apart and check step by step where things are going wrong.



I use a radeon9800 pro (catalyst 3.10) and the demo runs fine, and the effect is clearly visible. Its a very interesting effect indeed

Share this post


Link to post
Share on other sites
As was suggested I moved closer to the surface and I do indeed see the effect, and it is indeed nice.

Offset mapping on its own still looks flat, its just that the texture coordinates are distorted. I guess thats what you''d expect though.

Share this post


Link to post
Share on other sites
yep.. moving in _very_ close *g* now lets figure out a dirty hack how to use the height only as "depression" and discard fragments that result in invalid tex coords. if there''d be a way to decide per fragment if the view vector intersects the heightfield or not one could even get the silhoutte right

Share this post


Link to post
Share on other sites
quote:
Original post by Maximus
quote:
Original post by Trienco
why do i get the feeling that it might have to do with the fact, that we are using the same card? guess i''ll really have to take it apart and check step by step where things are going wrong.



I use a radeon9800 pro (catalyst 3.10) and the demo runs fine, and the effect is clearly visible. Its a very interesting effect indeed


ditto for me on a 98xt

Share this post


Link to post
Share on other sites
quote:
Original post by Dmytry
For clouds there should be something very different from displacement mapping because clouds are semi-translucent.



Sorry, I wasn''t very clear in my earlier post. The talk I saw had nothing to do with displacement mapping or offset mapping. It was about using the technique of rendering to billboards and then re-rendering them only after their angle to the camera had changed enough.

If I recall correctly, they were using this technique to avoid the overdraw every frame of rendering many slices of the volumetric data. Or maybe they were splatting particles for their clouds? I forget now.

Anyway, I just figured I''d mention the talk in case you wanted to look up their paper for any ideas on the general technique.

-John

Share this post


Link to post
Share on other sites

  • Advertisement