Re: [Algorithms] Bounding Box calculation

[Johnson, M. <Mat...@am...>]

Everything is a tradeoff. What's the goal? AABB alone makes for a pretty
coarse collision detection system. If you want a quick initial rejection
test, use spheres.=20

P.S. Tutorial on Principle Component Analysis:
http://csnet.otago.ac.nz/cosc453/student_tutorials/principal_components.
pdf

Matthew

> So why not use an AABB in model space instead, and then turn it into
> OOBB as part of model orientation? That's as fast as it gets to

That makes sense; it's actually exactly what I've got at the moment as
my
first-cut temporary solution. I was thinking to try some alternatives in
case there's something better than an AABB, but still pretty cheap.
AABBs
can be pretty bad in worst cases, eg a pencil on the diagonal in model
space. OTOH, it would take a worst-case artist to make a pencil on a
diagonal in model space... :)