View Full Version : Microsoft realtime lighting research
08-29-2002, 04:55 AM
Microsoft recently posted these on it's research site:
Keep in mind that these solutions are semi-precomputed, and most likely require a great deal of memory. make sure your jaw doesnt drop too far when you see the hdri :eek:
I think I broke my jaw...
08-29-2002, 07:57 AM
oops, i forgot to mention this was presented at siggraph 2002.
I looked through the video a few more times, and it seems like they are using some sort of finite element rendering technique (i.e. Gathering and Radiosity). I say this because they talk a lot about transfer vectors, and finite element rendering algorithms excell at solving energy transfer problems.
The caustics threw me off a bit though...finite element techniques are not too keen on calculating that, but then again it's a partially pre-computer solution. :surprised
Also, as a side not, the hardware this was demonstrated on was most likely an ati r300 (aka ati 9700). It's the only directX 9 compliant hardware available.
08-29-2002, 08:05 AM
ok heres a paper explaining the techniques demonstrated in the video:
researcher's homepage (obvious right? :applause: )
It's too late right now to read through these (3 AM), I'll take a look in the morning and try to make some sense of it all.
08-29-2002, 10:07 AM
Holy mother of crap! Microsoft are good for something after all!
Right I'm having a read of these now...
can you imagine what games are going to look like in 3...5...10 years time?
I can't wait to play Doom 9!
08-29-2002, 10:33 AM
Kay I get lost in that pretty quickly, my maths is not really up to speed
far as I can figure they do a compile-time GI pass of the model lit by an infinite sphere, then store where the light 'bounces' from adaptive sampling points as vectors.
Sounds great, but surely this technique has got to be a memory black hole?
08-29-2002, 12:55 PM
whoohaa, just watched it, amazing stuff, still gotta read the papers
08-29-2002, 04:28 PM
I am sure that we will have the necessary system requirements with lots of memory till it will be available to public....
does this mean OpenGL is going to go away?
08-29-2002, 07:24 PM
grey- not at all. openGL 2.0 will have the ability to do all of these nice things. OpenGL will allways have a wider appeal anyways, as it's not platform dependant.
08-31-2002, 01:22 PM
:: the research is amazine as well as excellent. DX 9 seems like it's getting closer to all the promises MS has made for enhancing Windows 3D abilities. I have a feeling that by DX 10 some of the realtime 3D feature will be available and running on the gigherts machines we have now.
Remember, software developers have not completely exploited all the power of current gigahertz processors, therefore Microsoft's research seems be aiming for this. So good to see! :thumbsup:
08-31-2002, 02:14 PM
Grrr. I've just spent the last 20 minutes searching, without success, for a cool related technology that I was looking at a few months ago.
The technology I'm thinking of took a series of photos of an actor's face, while colored lights were shone at it from hundreds of different angles.
They then generated a 3D model of the face, and used cunning interpolation techniques to allow you to move virtual lights and see the effect on the 3D face in realtime.
You could download a video, or play with an app that allowed you to load in multiple heads, move lights around, change the colors and so on. It was really, really cool to have a photorealistic head, and see photorealistic lighting effects changing in realtime in response to your actions.
Anyone got a link to what I'm talking about?
08-31-2002, 07:08 PM
Originally posted by Catfish
Anyone got a link to what I'm talking about? [/B]
Hi Catfish, I think you're talking about the "Acquiring the Reflectance Field of a Human Face" presented at siggraph 2000 by Paul Debevec, Tim Hawkins, Chris Tchou, Haarm-Pieter Duiker, Westley Sarokin, and Mark Sagar.
This is the direct link to it:
Hope this is what you are talking about, and yes, it does look pretty interesting :)
08-31-2002, 09:28 PM
That's the one - thanks.
The demo is worth the large download, for anyone who's interested.
01-13-2006, 04:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.