Re: [Algorithms] Octree PVS

[Charles B. <cb...@cb...>]

There are several papers that do things similar to this.  Unfortunately,
they're all at home, but if you look in recent Siggraphs and read the
visibility stuff, you'll find that actually a majority of current guys
are doing a cell-based PVS.  Basically they divide their world into a
uniform grid (but you could use a non-uniform grid, aka an octtree), and
then do things like form "extended projections" (that's a keyword to use
in Google) on the 6 faces of the cell to find what other cells are
visible, or what other objects are visible if you prefer.

At 12:15 PM 1/8/2001 -0800, you wrote:
>Has anyone thought of generating a PVS for octrees similar to how a PVS for
>BSPs are done? I was thinking you could take a polygon(s) that represents
>the face(s) of a cube and run them through all of the geometry between each
>octree leaf and clip against it.  If any of the polygon makes it through,
>then that leaf is visible, and you store the bit.
>
>I am about to try to implement this, but wanted a quick check to make sure I
>am not heading down a dead end road :)

--------------------------------------
Charles Bloom           www.cbloom.com