HDR + per pixel shadowing

I was thinking about how HL2 source and others are introducing HDR gaming, and how Doom and FEAR use Completely Dynamic per pixel Shadowing. Although I think Monolith does it differently than Id.

Is it possible to combine both techniques? Have completely natural Dynamic HDR that Generates completely natural and dynamic Shadows that react to the surrounding amount of light, your location etc. In other words a very close to life like lighting model that covers all the bases.

Can today’s top of the line hardware do this kind of thing? Can DX9 or OpenGL do this kind of thing?

 
HDR have little in common with any sort of shadowing. If you want a very exact representation of a real-life shadow, you have to deal with complex raytracing and remember that the real world shadows are generated from non-point light sources.

//btw, there are offline renders, that can generate shadows like that, I think.
 
The use of HDR is completely independent of the use of any of a number of shadowing techniques. There's no reason you can't mix them.

The problem, of course, lies in that once you move to HDR, the problems with lack of proper calculation of diffuse light will become more apparent.
 
Chalnoth said:
The use of HDR is completely independent of the use of any of a number of shadowing techniques. There's no reason you can't mix them.

The problem, of course, lies in that once you move to HDR, the problems with lack of proper calculation of diffuse light will become more apparent.

Can you tie the control of shadowing to the HDR?

Is HDR dynamic at all? OR is it faking you out with really high resolution texturing type techniques.
 
All that HDR is is the support of colors outside the usual [0,1] range, combined with an algorithm that maps the high-range format to [0,1] (tone mapping).

All that shadowing is is the prevention of light from being applied to a surface.

So they aren't really that related, at least in a very basic sense. HDR, of course, can have profound effects on how light is reflected. But attempting to combine the two effects to get something new and different? I'm not seeing it.
 
I'm implementing this as my final-year dissertation right now. Got Visual C++ 2005 fired up and mid-compile as I type this sentence :D

The thing you want to bare in mind with this technique is that it will murder all but the most hardy of GPU's. Shadow volumes or maps can be pretty expensive on their own - throw in the high bandwidth requirement of HDRI and you're not gonna have much space/time left for anything else ;)

My application doesn't really do much yet and it's chewing up 44mb of VRAM @ 1024x768 without any models/textures loaded :oops:

Jack
 
I don't know as much about this subject as i'd like...so bear with me if my idea sounds stupid.

HDR is a 'multipass' effect as I like to think of it (in that the same scene is 'rendered' multiple times to generate the effect. I would think that in order to tie-in per-pixel shadowing with the games HDR effect, you would need to use a similar multipass effect for all the shadows in the game...otherwise you'd end up with a nice dynamic HDR effect and shadows that may not be as dark or as bright as the HDR lit scene requires...

Am I making any sense?

Basically, I look at it as, if this were feasible, the shadows would need to dynamically change in tandem with the HDR lighting to produce the correct shaded output. I also have this notion in the back of my head that to acheive such a thing with per-pixel shadows would require a hefty chunk of performance (to..'multipass' the shadows.).

So anyway, random thought...
 
It's not an "effect" but rather a capability of the hardware to process and output data in more than canonized 8 bpc - the ability to apply operations over MUCH wider field gradient of data (and store the final/intermediate results in far sutable precise format).
Think about, how ugly was 16-bit dithered output (mostly on multi-pass effects) compared to 24/32-bit mode in the Golden age of 3Dfx.
 
Sobek said:
Basically, I look at it as, if this were feasible, the shadows would need to dynamically change in tandem with the HDR lighting to produce the correct shaded output. I also have this notion in the back of my head that to acheive such a thing with per-pixel shadows would require a hefty chunk of performance (to..'multipass' the shadows.).

So anyway, random thought...
But shadows aren't painted onto a scene. Shadows are there to block light from being applied to surfaces. So as long as your lighting algorithm is HDR in nature, the shadowing algorithm automatically is.

The only drawback is that in scenes with high dynamic range, light reflections lighting up areas become more prominent. So not having additional attempts to go for global illumination will start to look worse.
 
See in the console games subforum the latest Ghost Recon Advanced Warfighter single player videos to see a good example of HDR + Shadowing utilisation.
 
Last edited by a moderator:
fellix said:
Think about, how ugly was 16-bit dithered output (mostly on multi-pass effects) compared to 24/32-bit mode in the Golden age of 3Dfx.
Well, 3Dfx's solution was pretty awful but there was another that looked near perfect. :p
 
Razor1 said:
ambient occlusion does the job well. peudo radiosity with shadow calculations all in one :)

Ambient occlusion cannot handle direct light sources and thus no realistic shadows either. It is used to 'measure' the amount of soft, diffused light that can reach the surface and thus you'll need a few direct lightsources for most lighting situations, except maybe an overcast day.
 
Simon F said:
Well, 3Dfx's solution was pretty awful but there was another that looked near perfect. :p
...just like the INT10 output in present days - why bother wasting resources over FP16 if it can be done with slightly hacked integers. :D
 
Wasn't 3DFX's actually 22bit or whatever hack they managed to do?

Hey Chalnoth, if the definition of HDR is a larger range than [0,1] then Quake 3 has HDR since it was capable of [0,2] before dithering the output. ;)
 
Back
Top