There are some significant errors in the manner of spacular projection from
material surfaces that is evident in ALL of the "hall of fame" and related
renderings. I know some work was done in the 90s to facilitate crystaline
surface materials, but there is no evidence of that now. Moreso, there are
problems with the majority of renderings compared to both human and machine
vision characteristics. Some of these issues are almost unbearable to view.

The lack of an environmental medium in the default rendering is also a concern,
someone needs to encourage "air" to be introduced as a bounding object.

Regarding the specific problem:

In almost all of the renders, pixle regions that eminate from the same source
material are typically generating contiguous blocks of uniform color,
regardless of the structure of the object. This has always been a major
problem, but here, it is a glaring error. Every single HOF render has the same
problem, the textured stone and other procedural materials are the most blatant
issue.

When comparing the output of pov to either conventional camera or human vision,
one may argue that it is "more accurate" for scientific representations like
atomics or non-physical views, however, humans and machines rely on the
perceptual differential to distinguish objects in their field of view.

Most common digital cameras compensate for diverse light by alternating pixle
weights to average the target color. Human eyes rely on both this quality as
well as time based fluctuations regardless of how minute.

POV simply doesnt... This really needs to be addressed, the accuracy of
rendering is severely off due to the lack of surface properties and
environmental interplay.

Historically there was a concern with the rays not including the full set of
material and optical deviations throughout the entire path, along with some
issues with processing speed that struck down quality of render for the ability
to process it on a conventional machine.

Now, however, this limitation is more of the problem than a gain, there are few
 major problems in the rendering, and it is quite clean, but the lack of signal
qualities on the optical trajectory blatantly eliminates accurate interference
and surface qualities and generates huge blobs of visually convoluted and
highly distasteful pixles...

Simple retention of the resulting rays and appropriate generation of the final
pixle field is not a complex fix, however, these blobs of perfectly smooth
color are simply impossible in the real world and highly distracting in human
perception.

The problems/issues with environmental characteristics would also be fixed with
this solution, the interplay between optical characteristics and the
manipulation through a constant medium (like air/etc) requires the same
processing as the corrections.

This would solve a lot of the issues with pov's obsession with perfect
crystals... I've never seen glass that has a perfect lense characteristic.
Easiest fix is to functionally generate the actual glass crystals and let
povray beat on it for a few days. (not to mention air and fluids)

There is a more simple solution that would reduce the problems with unrealistic
rendering and fix a lot of the medium-based rendering issues without having to
deal with atomic-physics scale data sets.

Let me know if anyone has worked on this issue, especially the crystaline
surface generation and procedural textures, and spaculation of the resulting
image to generate a perceptually clean output.

Humans simply do not fare well perceiving distances without air ;)

Contiguous color and quality is close to impossible in the real world, you
naturally freak out when you encounter such, so... lets take a hint from those
yucky off-color lime green and orange pixles in the digital photographs that
somehow and mysteriously blend perceptually into the best image to view, and
fix the rendering system to remedy this problem.

Try to reverse model your renders, especially from a real source environment, to
see how far off the render really is... on a lot of this surface problem, its
close to 10% skew on any ray.

-Wilfred
Wil### [at] Gmailcom

Post a reply to this message

Wilfred wrote:
> When comparing the output of pov to either conventional camera or human vision,

So is there a difference between POV-Ray 3.6 and 3.7, or is this a feature
request? If so, please post it in porvay.general and not this group.

	Thorsten, POV-team

Post a reply to this message