In povray.unofficial.patches Zeger Knaepen <zeger.knaepen@student.kuleuven.ac.be>
wrote:
> small bright objects don't disappear in AA and focal blur looks more realistic.
> In general, the new clipping system is a more realistic and a more logical
> approach.

  You are wrong.

  Firstly, antialiasing and focal blur are two different things.
Focal blur can keep the post-clipping, but with antialiasing post-clipping
is not a good solution.

  Secondly, post-clipping antialiasing is not more correct than the
pre-clipping version. Basically both are wrong. However, post-clipping
is more problematic than pre-clipping.

  Pre-clipped antialiasing causes some little annoyances in a few cases
(eg. if you try to make a subpixel-sized bright star) but in general it
gives high-quality smooth non-pixelated images. While it can be as an
annoyance, the resulting images have good visual quality (with regard
to antialiasing).
  Post-clipped antialiasing causes problems in a multitude of images
very frequently. You only have to have a surface which is much brighter
than 1 and a sharp edge adjacent to a dimmer surface (which is quite
a common situation) to get ugly pixelation of the edge. Pixels much
brighter than 1 are completely normal, not modelling flaws.
  Just look at The Kitchen image carefully to see the ugly post-clipping
artifacts (these were not taken into account in judging because it
was decided that these artifacts can be got ridden of in future
versions of POV-Ray).
  There are cases were post-clipping causes horrible artifacts which
basically destroy the image, while pre-clipping results in a smooth
and extremely nice-looking image (one such image I have seen is a
scene with wine glasses which have lots of small bright sparkles
due to how the glasses reflect and refract light; the result with
the pov3.6 was completely horrible while the result with pov3.5
was very nice and photorealistic).

  Note however, once again, that I'm not saying pre-clipping is *correct*.
I'm just saying that it gives much better results in the vast majority
of cases while post-clipping gives horrible results in many.

  What would be the most physically correct interpretation of
higher-than-1 pixels?
  Think about what happens in photography: Having areas of the image
which are much brighter than what the film can register is completely
normal. There's nothing abnormal there and this effect is in fact
used very frequently in photography: Just think about a nature photo
with the Sun visible: The Sun will be much brighter than what the film
can register, specially if the exposure time has been set so that the
rest of the photograph will have a good illumination.

  So what happens with these ultrabright areas?
  Answer: Color bleeding.

  In photography ultrabright areas will bleed their brightness to
surrounding areas of the film. In the above example the Sun will
look bigger than it really is and its brightness will usually fade
gradually (instead of getting a sharp edge).
  This is exactly what happens with all spots in the image which are
brighter than what can be registered: They will bleed their brightness
and look bigger on film than what they really are.
  Thus if you photograph for example a very small source of light in
an otherwise normal lighting and exposure conditions, it will look
like a much bigger light than it really is.

  Now, *that* would be the most physically correct solution for
brighter-than-1 pixels.

  It is still to be decided what kind of solution future versions of
POV-Ray will have, but even if a color-bleeding algorithm is not yet
implemented, pre-clipped antialiasing is still a much better solution
than the post-clipped one.

-- 
plane{-x+y,-1pigment{bozo color_map{[0rgb x][1rgb x+y]}turbulence 1}}
sphere{0,2pigment{rgbt 1}interior{media{emission 1density{spherical
density_map{[0rgb 0][.5rgb<1,.5>][1rgb 1]}turbulence.9}}}scale
<1,1,3>hollow}text{ttf"timrom""Warp".1,0translate<-1,-.1,2>}//  - Warp -

Post a reply to this message