Nieminen Juha <war### [at] sarakerttunencstutfi> wrote...
>   A more general approach:
>   Add a new keyword, eg. 'no_antialiasing'. When applied to an object,
then
> a ray hitting this object will stop the supersampling immediately. Should
> be easy to implement.

This could end up more complicated than you might think.

First, radiosity sampling methods generally re-use points from one pixel
calculation to the next.  This would probably not be a problem, though (I
think POV implements something like a dirty-pixel buffer that determines if
it needs to trace the sample or not).

Second, what do you do if you can see an object _through_ another object.
What if the clear blocking object is a crystal?  What if it is, instead, a
media container object?  I have been wondering how to handle this and
similar questions when dealing with post-processing.  There will be ways to
tag objects for post processing, and then let the filter find out if a
tagged object was hit.  But how should those clear media containers be
handled???  And what about not-so-perfectly-clear, yet mostly clear,
objects, like a wine glass.  The other thing to worry about for me is depth.
Which depth should be returned: the first object hit, or the first object
that you'd actually _see_?

-Nathan

Post a reply to this message