Not really.  We generally assume that alpha is associated (i.e. premultiplied, 
though I hate that term), and that's really the sensible way to represent 
images IMHO.  Among the many problems with "un-premultiplied" color (I dare you 
to tell me in physical terms or units what that represents) are that it can't 
represent all the situations we care about.  As an extreme (but totally useful) 
example, consider an emissive but transparent gas, which might have a color 
like (.5, .5, .5, 0).  What's that as an un-premultiplied color?

I'd like to hear more about the blend modes that need non-premultiplied colors. 
 Let's determine if that's really true, and if so, whether there is a sensible 
way to have the equivalent blend using the usual premultiplied colors.


On Jul 11, 2012, at 6:01 PM, Stefan Stavrev wrote:

> I am working on blend modes and I need non-pre-multiplied color values.
> 
> Is there a way to "stop" pre-multiplication when the image is read? I don't 
> know
> how that works.
> 
> Or will I have to divide with the alpha values? This could be expensive 
> though,
> imagine doing this for high-res image sequences, for each image in the 
> sequence,
> for each channel.

--
Larry Gritz
[email protected]


_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org

Reply via email to