Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site uwmacc.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!uwvax!uwmacc!curtis
From: curtis@uwmacc.UUCP (Alan Curtis)
Newsgroups: net.graphics
Subject: Re: Re: Constant-Time Ray-Tracing
Message-ID: <1432@uwmacc.UUCP>
Date: Tue, 27-Aug-85 13:37:03 EDT
Article-I.D.: uwmacc.1432
Posted: Tue Aug 27 13:37:03 1985
Date-Received: Thu, 29-Aug-85 23:17:45 EDT
References: <644@wdl1.UUCP>
Reply-To: curtis@uwmacc.UUCP (Alan Curtis)
Organization: UWisconsin-Madison Academic Comp Center
Lines: 12


In addition to searching the tree for the particular node (cube) which the
ray happens to land in, all objects in the nodes which are passed through
during the search must be examined for intersection, as they may potentially
overlap the cube in question.

I haven't tested octtrees, but I've found (emperically)
that a multi-dimensional binary tree
is faster than a quadtree for determining intersections of objects because the
number of unsuccessful tests is cut by about a factor of two.

                                          Alan Curtis