[Bf-cycles] (un)Premultiplyed render passes

[Brecht Van Lommel]

Here is a graph showing how the passes fit together:
http://wiki.blender.org/index.php/Doc:2.6/Manual/Render/Cycles/Passes#Combining

As you can see, for example the total diffuse component is:
diffuse_color*(diffuse_direct + diffuse_indirect).

Of these I would say it is actually the other way around, the diffuse_color
pass is "premultiplied" with the alpha value, and it's the diffuse_direct
and diffuse_indirect that are "normalized" to not be influenced by alpha.
(Actually if the alpha and diffuse_color are zero then diffuse_direct and
diffuse_indirect are also zero, because computing them would have no effect
on the combined pass and would make rendering slower, so it's not exactly
so simple).

Only one of the two is "premultiplied", otherwise combining them would give
a double "premultiplication". It's not a matter of inconsistency, it's just
that the passes contain a different kind of data.

Brecht.

On Fri, Feb 28, 2014 at 4:41 AM, Adriano Oliveira <adriano.ufrb at gmail.com>
wrote:
> BUILD: b5d6f14
>
> I don't know if this is a bug or not, but I am rendering dufuse and glossy
> passes (direct, indirect, color) and I have noticed that some are
> premultiplied (likre indirect) and others are not (like color).
>
> This seams to be inconsistent.
>
> Adriano A. Oliveira
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20140228/1940c31b/attachment.htm