Archived

This topic is now archived and is closed to further replies.

LoreKeeper

ROAM + QuadTrees ???

Recommended Posts

Somebody mentioned something on that... is there a point to using QuadTrees with the ROAM algorithm? How does ROAM benefit? thanks A-Lore

Share this post


Link to post
Share on other sites
*GrinZ* - BoRReL... you are a funny one...

I know that QuadTrees are great for a basic terrain engine - but the ROAM algorithm already caters for everything that QuadTrees offer, plus it adds LOD.

So why would I want to use QuadTrees in conjunction with ROAM?

Henri aka A-Lore

Share this post


Link to post
Share on other sites
A simple quadtree doesn''t support LOD, but there are many LOD methods out there that use quadtrees(Rottger). I can''t see any reason why you would use quadtrees with roam. I don''t know why you''re worrying about it.

*** Triality ***

Share this post


Link to post
Share on other sites
AFAIK, ROAM *only* does LOD reduction. There is no frustum culling. You can use a quadtree to determine which patches are visible, and then use ROAM on those.

There is also a method of using quadtrees in the ROAM algo itself, instead of bintrees. Cant remember exactly how they are supposed to tesselate, but it is potentially faster than using bintrees.

An alternative to ROAM is to use geomipmaps - easier to code and much easier to optimise with todays hardware.

Share this post


Link to post
Share on other sites
Actually the ROAM algorithm >does< include frustum culling (go read the paper again - the section on Wedges is where frustum culling comes to play) infact the ROAM frustum culling culls horizontally and vertically, which a Quadtree-culling cannot provide.

Somebody once mentioned quadtrees and ROAM working together... but I don''t remember what the purpose was...

Ah well... back to work... trying to get a great system going.

Henri aka A-Lore

Share this post


Link to post
Share on other sites
The Quadtree is a data structure that allows you to quickly
reject children via a bounding hierarchy. If the parent is not within the view frustum then the children cannot be. However the structure could just as easily be used to compress images.

Quadtrees can easily be extended to horizontal,vertical or
any dimension it is just how you view the data structure.

Please do not put these algorithms into a bucket of "Yes that
is a landsscape rendering data structure".

ROAM + BSP TREE + QUADTREE. Wow that must be really good if
it includes all those trees. Data structures represent data
do not make them magic or mystical beings

Otherwise you end up not seeing the wood for all the trees




Share this post


Link to post
Share on other sites