In article <38989C51.C2809A76@unforgettable.com>, 
inq### [at] unforgettablecom wrote:

> Hmm.. could this be done algorithmically by using trace from a lot of
> really odd angles to determine how "open" a certain point would be? (ie,
> a convex surface could be hit from anywhere, but a deep concave one, or
> a groove between two convex surfaces, could only be hit from angles
> which approach perpendicular.)

That is one of the methods I have been thinking of using for curvature. 
Another would be to test a bunch of points in a sphere centered around 
the evaluation point, and returning the percentage which are inside the 
object. Or shooting rays at parts of the surface close to the evaluation 
point, and using their positions and normals to calculate the 
curvature.(The last one is probably the most accurate, but is the one I 
wasn't able to complete)

For the visibility pattern, a single ray(or multiple rays with a 
specified jitter amount) would be shot at a specified point or along a 
specified direction.

I will get to work on these and see if they are at all useful...

-- 
Chris Huff
e-mail: chr### [at] yahoocom
Web page: http://chrishuff.dhs.org/

Post a reply to this message