[Bf-cycles] Cycles loads RGBA textures "premultiplied"?

[Sergey Sharybin]

Hi,

It is not "associated" it is "premultiplied" in context of blender and
would be nice if this terminology is followed in discussions.

OIIO is not really relying on any alpha treatment, it just gives you
whatever you're requesting to as far as i concerned. But what Cycles
expects from the images is also depends on the shader setup, so having a
look into setup and image which gives issue to you would help investigating
if it's some issues with setup or some bug or some lack of proper CM in
Cycles.

It is also to be mentioned, that all the CM work we did for image input in
blender is to be still re-done for Cycles. So it'll respect input color
space, alpha type and so on.


On Wed, Mar 11, 2015 at 2:40 AM, Troy Sobotka <troy.sobotka at gmail.com>
wrote:

>
> On Mar 10, 2015 1:44 PM, "Jason Clarke" <jason.clarke0 at gmail.com> wrote:
> > Why does it do this? Is there a way to disable it? It makes it
> impossible to store anything in a texture's alpha channel except actual
> transparency. Trying to store something like bump or roughness data in the
> alpha channel will ruin whatever is being stored in the RGB side.
>
> Technically I believe it assumes associated alpha. This is because
> associated alpha is the only correct method to represent light, and as
> such, OIIO relies on this implementation.
>
> That said, associated alpha doesn't prevent you from storing values in the
> RGB data with zero transparency. A set such as 13.5,0.0,0.0 with an alpha
> of 0.0 is perfectly valid, representing pixel with no occlusion and
> emitting light.
>
> I would think this has more to do with your chosen file format perhaps?
> Have you tried with EXRs?
>
> With respect,
> TJS
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>


-- 
With best regards, Sergey Sharybin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150313/9b7b809d/attachment.htm