BJ> Alien battle ship commander
I think it’s time I got new business cards 
k> Bump mapping to me is dot products with a normal map.
Yes. Unfortunately, to any artist, or anyone involved in offline CG/rendering, bump mapping means a grayscale height field. Most artists (even technically studly ones!) would not know what a normal map was if they opened it in Photoshop.
Thence, the language confusion.
“Projected light textures,” anyone?
Mmm, well I read my normal maps in as heightfields anyway, and convert them to normal maps internally - I can’t understand why people bother with those photoshop plugins etc. You’re left with limited precision options, and storage options if you import them pre-cooked, in my opinion. If your clients hardware has to have them range compressed, then that’s no problem…etc.
It surprises me that Doom3 is running so fast with stencil shadows on top of all the other effects. It’s got to be movie sequences, right?
It was running on an R300 board, thats why it was running nice and smooth.
There are plugins for Max and Maya to take a high detail object, optimize it and generate normal maps for it. I’m assuming that is what they are doing, why waste time doing the same thing in the engine.
My $0.02
John.
Originally posted by knackered:
[b] Eh? To perturbate means to ‘offset’. That’s a different trick.
Bump mapping to me is dot products with a normal map. Why are people still talking about the offset trick in this day and age, and on this forum? It looks crap - forget it. On hardware incapable of dot3, don’t even bother trying to emulate it.
[/b]
It was a siggraph paper (2001 I beleive). It wasn’t just about that, but sort of an introduction to their research. Something about using the vertex normals and perturb them. Is that it?
V-man
Originally posted by john_at_kbs_is:
[b]There are plugins for Max and Maya to take a high detail object, optimize it and generate normal maps for it. I’m assuming that is what they are doing, why waste time doing the same thing in the engine.
My $0.02
John.[/b]
I wasn’t talking about geometry reduction using normal maps - just general bump mapping.
bumpmapping is not doing a dotproduct with a normal map. its mapping some bumpiness onto a surface and use some technique to let the bumpiness get visible. the most loved way today is simple dot3 on a normalmap, but there are plenty others. in fact, it just depends on the lightingequation/solution you use, and as todays hw is not very programable yet all we can use is some dot3 and some dot3^x for specular… that’ll change… and you don’t need a normalmap. what you need are normals. these you can get from a heightmap as well… and the actual bumps are represented in a heightmap. the surfacenormals of the bumpmap are in the normalmap…
don’t be that restricted or you will not see that much use in the upcomming shader hardware…
Here is the original related paper for this detail reconstruction using a bump map:
http://www.cs.unc.edu/~olano/papers/aps/APSlossy.pdf
I’m surprised it hasn’t been posted already.
On the stencil volumes vs displacement mapped geometry, they all use the simplified mesh (no displacement). It’s been posted already (by me and I assume others) that the visible silhouette of an object is already defined by the simplified geometry, and the shadow volume as projected onto a surface would be no worse than the appearance of the object silhouette in the scene.
Carmack said in a .plan (I think the latest one) that, displacement mapping would screw up the stencil volumes. I described this problem at the start of the massive doom3 thread some time ago and it upset someone ;-).
[This message has been edited by dorbie (edited 07-29-2002).]
Originally posted by davepermen:
[b]bumpmapping is not doing a dotproduct with a normal map. its mapping some bumpiness onto a surface and use some technique to let the bumpiness get visible. the most loved way today is simple dot3 on a normalmap, but there are plenty others. in fact, it just depends on the lightingequation/solution you use, and as todays hw is not very programable yet all we can use is some dot3 and some dot3^x for specular… that’ll change… and you don’t need a normalmap. what you need are normals. these you can get from a heightmap as well… and the actual bumps are represented in a heightmap. the surfacenormals of the bumpmap are in the normalmap…
don’t be that restricted or you will not see that much use in the upcomming shader hardware…[/b]
You speak with such authority, dave. I said bump mapping TO ME is … etc. etc.
I don’t really care what it means to you - I only care that my fellow programmers and artists in my team know what I mean (psst! [whisper]we use dot3[/whisper])
There’s also a really good demo of geometry simplification with normal maps…go to here:-
http://research.microsoft.com/downloads/
and click on the “Texture Mapping Progressive Meshes” link.
knackered,
I was trying to answer dabeav’s original question on making the low res mods look like the ultra high res mods…
John.
I think of a normalmap as being a texture that contains rgb values that represent the xyz values of a normal. Each normal from each triangle is interpolated onto the texture to their corrisponding places(UVmap data). You can think of it as pre-calculating the normals accross a polygons face. Doing this with a highres model increases the normalmaps detail. With this method the normals can point in all directions that could resemble a sphere.
A heightmap as far as I’m conserned is a black and white image where white is the highest point and black is the lowest. Before you can use a heightmap with the DOT3 bumpmapping method you have to convert it to a normalmap. The problem with this method of bumpmapping is that your normals can only create a sort of half dome.
Anyway, that’s how I’ve interpreted everything I’ve read on this subject.
[This message has been edited by WhatEver (edited 07-29-2002).]