[Bf-committers] General thoughts on * mapping (Re: Re: parallaxmap error)

[Nick Winters]

On Fri, 17 Dec 2004 23:52:11 -0500, trip <trip at spymac.com> wrote:
> 
> On Dec 17, 2004, at 8:06 PM, GSR - FR wrote:
> 
> >
> > Also interesting is relief texture mapping (solves the edge problem
> > that otherwise would require the expensive true displacement approach)
> > http://www.inf.ufrgs.br/%7Eoliveira/pubs_files/RTM.pdf
> 
> 
> Woa..!
> 
> That would be perfect for cutting down render times and game scenes.
> But what are the CPU usage for this ? Would it be real time or just
> another different way to use CPU cycles? And what of the fuzz around
> some of the edges as you can see in some of the picturees ??
> 
the "fuzz" is most likely a image rescaling error.  Some pixels are
fully opaque, others are fully transparent.  When the image is
resampled [resized for your display] it isn't interpoloated like in
opengl, so the edges remain either fully opaque or transparent.  You
can re-create the same problem in blender by rendering something, then
zooming out from it in the render display.

the cpu change of techniques like this is mainly in that you can
reduce the number of polygons transformed and rasterized, yet still
have the detail of like subsurf level 5 [1024 times the number of
polys of no subsurf, 64 times the number of polys of subsurf 2].  The
idea is that the cost of these techniques [on the texture] is much
lower than simply rendering that many polygons.  On some graphic cards
this isn't the case [all graphics cards today have special hardware to
transform polygons, where few have special hardware to allow such
complex texture effects].