PDA

View Full Version : New "Mesh" Compression for 3d Graphics


Bonedaddy
06-17-2004, 05:29 AM
This just hit /., thought I'd post here (and it was done at my alma mater!):

From slashdot:

A new algorithm that uses successive approximations of detailed models to get significant compression has been revealed by researchers at The University of Southern California. Just as MP3s remove high frequencies we can't hear, this algorithm removes the extra triangles in flat or near flat surfaces that we can't see. Experts in the field are giving this work high praise and imply that is will be immediately applicable to 3D modeling in games, movies, CAD and more.





Sounds like it might make for some much better LOD and reduce functions... :)


Link (http://www.usc.edu/dept/engineering/news/2004_stories/2004_06_15_desbrun.html)

Ollarin
06-17-2004, 06:07 AM
Wow! This is awesome! Will be big breakthrough for game developers i reckon. :bounce:

By the way. What does "LOD" stand for? :blush: (Not really fluent in acronyms. :D

gdimmrt
06-17-2004, 07:11 AM
LOD = Level of Detail.

ie: in a game engine, as the geometry becomes further away with respect to the camera, the geometry is replaced with a lower poly version as to speed up rendering while still visually feasible.

Bonedaddy
06-17-2004, 07:59 AM
Now that I think about it, it'd probably make much more manageable meshes out of scan data. Currently, say, face scans are at least in the hundreds of thousands of polys, and getting them down into a workable order, so one could animate them, is difficult (the Matrix guys didn't even try; they went for a displacement-based animation capture method). If this could sensibly bring down polycounts on scan data (esp. LIDAR stuff), it could be a lifesaver for render times, and result in more real-world accurate models (and thus renders).

Ollarin
06-17-2004, 08:55 AM
Ahh...Cool. Thanks guys! :thumbsup:

playmesumch00ns
06-17-2004, 09:06 AM
Originally posted by Bonedaddy
(the Matrix guys didn't even try; they went for a displacement-based animation capture method).

Yeah, that's if you believe the conference papers. Their reflectance-captured cloth and skin shaders also looked remarkably like a bunch of blinns too... :)

The simple examples they show on the web page look kinda cool, but for most applications it's far better to get a human modeller to do it. For film work our modellers generally just draw polygons on top of the scan to create beautiful low-poly looping quad meshes that can be subdivved. Then they apologise profusely about the two triangles they couldn't get rid of. Bless:)

Bonedaddy
06-17-2004, 01:48 PM
Originally posted by playmesumch00ns
for most applications it's far better to get a human modeller to do it.

Ack. That's kind of icky work. Well, I say give it a shot, it couldn't hurt. It may even work well in concert with human modelers, to take into a low-poly form, and let artists resculpt the edges to make loops and whatnot... :)

-J

playmesumch00ns
06-17-2004, 06:10 PM
Yep. Anything that saves time's great!

CGTalk Moderation
01-18-2006, 01:00 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.