Warp wrote:

>   As a completely personal opinion, I think *any* kind of post-processing
> of the rendered image is cheating, or at least can be used for cheating.
> 
>   For example, the old IRTC allowed resizing the rendered image as a post-
> processing step. Some people abused this allowance to cheat and get a much
> higher-quality antialiasing than what the renderer was able to produce all
> by itself (by rendering a very large image and then using a 2D software to
> scale it down by averaging pixels, resulting in high-quality antialiasing,
> which was *not* the product of the renderer).

my personal opinion on that matter: This example sounds perfectly
acceptable to me and could also be used to argue the point that it
should be allowed to do more post-processing.

If the above procedure actually has a better ratio of quality
per render time than other anti-aliasing methods one might be
allowed to ask: Why isn't the method built into the renderer?
Maybe it wasn't felt worth the development time because it is
so simple to do as a post-processing step? That would lead to
a somewhat circular argumentation.

Also, if I now build my own patched version of the renderer which
implements this anti-aliasing method, is it suddenly ok to use it
because it is now longer a separate post-processing step? What
about using MegaPOV which also has some post-processing?

Most global 2D effects could conceivably be implemented in SDL
with multipass renders and evaluating pigments of the input image.
Does this make it any more or less cheating?

My recommendation would be to allow any kind of post-processing
which is of a "general" nature, that is, a global effect which can
be meaningfully applied to any image. Naturally this excludes any
manual editing of portions of the image, say, using a brush.

However, I think it should then be mandatory that the image
description lists all post-processing steps which were used.

 > This principle would of course allow extreme #3 to be submitted.
 > However, it's then up to the judges to vote it down because there
 > was no real 3D modelling involved.

Similarly, people could vote done images due to excessive
post-processing if they feel someone abuses relaxed rules.

Post a reply to this message