What's the most efficient terrain rendering algorithm for high-resolution datasets?
Author
A lot of time passed since I last had a look at terrain rendering algorithms. What is the current de facto standard for huge terrain sets (possibly planet scale) and modern GPU's?
Depends on what you want to use it for. Is it just visualization? Also interaction? Care about hardware requirements? But in general, I would say any algorithm that builds on the Clipmap technique and (possibly) also implements advanced compression. Here are some articles to look at:
Terrain Rendering Using GPU-Based Geometry Clipmaps (Hoppe, Asirvatham)
Terrain Rendering using Spherical Clipmaps (Clasen, Hege)
Seamless Patches for GPU-Based Terrain Rendering (Livny et al)
A Heightfield on an Isometric Grid (McGuire, Sibley)
Persistent Grid Mapping - A GPU-based Framework for Interactive Terrain Rendering (Livny et al)
http://www.gamedev.net/community/forums/topic.asp?topic_id=504549
Terrain Rendering Using GPU-Based Geometry Clipmaps (Hoppe, Asirvatham)
Terrain Rendering using Spherical Clipmaps (Clasen, Hege)
Seamless Patches for GPU-Based Terrain Rendering (Livny et al)
A Heightfield on an Isometric Grid (McGuire, Sibley)
Persistent Grid Mapping - A GPU-based Framework for Interactive Terrain Rendering (Livny et al)
http://www.gamedev.net/community/forums/topic.asp?topic_id=504549
Agreed, the Clipmap technique works very well with GPU's. We are using it together with wavelet compression of elevation data and with fractal refinement algo to render planet Earth (~150m elevation dataset, further refined down to centimeters). You can see some screens and videos in my journal.
http://www.gamedev.net/community/forums/mod/journal/journal.asp?jn=503094
Not much technical info in there, but as I've read ndhb's paper it shares some techniques - discrete LOD with the same LOD criteria, skirts for the overlaps (adjusted for particular tile so the stretching is minimized). Tiles of required level are uncompressed and/or refined on the fly on GPU.
However, we use R2VB instead of VTF because we (still) support ATI SM3 cards with their non-support of VTF.
regards,
Brano Kemen
http://www.gamedev.net/community/forums/mod/journal/journal.asp?jn=503094
Not much technical info in there, but as I've read ndhb's paper it shares some techniques - discrete LOD with the same LOD criteria, skirts for the overlaps (adjusted for particular tile so the stretching is minimized). Tiles of required level are uncompressed and/or refined on the fly on GPU.
However, we use R2VB instead of VTF because we (still) support ATI SM3 cards with their non-support of VTF.
regards,
Brano Kemen
Author
wow cameni, that's some impressive work!!! will this be a proprietary engine or open source?
you basically implemented what I tried to implement :)
congrats
you basically implemented what I tried to implement :)
congrats
Well, we'd like to make it to engine suitable for visualization, simulators and of course to make a game with it [wink]
Even though we are wondering what kind of game can really employ the enormous area Earth (or any planet) provides, in a way that not only the players won't get lost but also that it will be really utilized in there. We have some ideas, but we didn't fully regarded it because there's sooo much work on the engine yet.
But any suggestions are welcome [smile]
What do you want to achieve?
[Edited by - cameni on December 3, 2008 12:10:41 PM]
Even though we are wondering what kind of game can really employ the enormous area Earth (or any planet) provides, in a way that not only the players won't get lost but also that it will be really utilized in there. We have some ideas, but we didn't fully regarded it because there's sooo much work on the engine yet.
But any suggestions are welcome [smile]
Quote:you basically implemented what I tried to implement :)
What do you want to achieve?
[Edited by - cameni on December 3, 2008 12:10:41 PM]
Author
Quote:Original post by cameni
Well, we'd like to make it to engine suitable for visualization, simulators and of course to make a game with it [wink]
Even though we are wondering what kind of game can really employ the enormous area Earth (or any planet) provides, in a way that not only the players won't get lost but also that it will be really utilized in there. We have some ideas, but we didn't fully regarded it because there's sooo much work on the engine yet.
But any suggestions are welcome [smile]Quote:you basically implemented what I tried to implement :)
What do you want to achieve?
Planetary rendering, I played around with P-BDAM for a while. I have the entire, reworked SRTM 90 set (http://srtm.csi.cgiar.org/) sitting on my box, waiting to be rendered in its full glory in real time :)
Quote:Original post by pompokoPlanetary rendering, I played around with P-BDAM for a while. I have the entire, reworked SRTM 90 set (http://srtm.csi.cgiar.org/) sitting on my box, waiting to be rendered in its full glory in real time :)
We use the SRTM 90 set too, remapped and LOD-friendly compressed, but also plan to use the more detailed mountain datasets to achieve better shapes where it matters more. Hmm, now I see there is v4 of the SRTM 90 dataset, could be fixing some bugs in terrain data ...
I didn't read P-BDAM paper before though, I came to my present rendering somewhat iteratively [smile]
Author
Quote:Original post by cameniQuote:Original post by pompokoPlanetary rendering, I played around with P-BDAM for a while. I have the entire, reworked SRTM 90 set (http://srtm.csi.cgiar.org/) sitting on my box, waiting to be rendered in its full glory in real time :)
We use the SRTM 90 set too, remapped and LOD-friendly compressed, but also plan to use the more detailed mountain datasets to achieve better shapes where it matters more. Hmm, now I see there is v4 of the SRTM 90 dataset, could be fixing some bugs in terrain data ...
I didn't read P-BDAM paper before though, I came to my present rendering somewhat iteratively [smile]
Yes, v4 allegedly fixed a lot of the obvious issues. Are there any plans to go open source with your project, or do you need a helping hand?
Quote:Original post by pompokoAre there any plans to go open source with your project, or do you need a helping hand?Open source - certainly not in this phase. I shall PM you and we can talk about cooperation, though.
Author
Quote:Original post by cameniQuote:Original post by pompokoAre there any plans to go open source with your project, or do you need a helping hand?Open source - certainly not in this phase. I shall PM you and we can talk about cooperation, though.
Cool, appreciated :)
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
