[cairo] Image Format / Pre-Multiplied Alpha Related Feature Request

[Tavmjong Bah]

On Thu, 2016-01-28 at 17:18 -0500, William Kappler wrote:
> Thanks for the suggestion.
> 
> I'm not dealing with a single color, but that did give me an idea
> that perhaps I could render color without alpha and then alpha on its
> own, and just blend the data together on my end. I'm not sure if that
> will work or not, I need to play with it.
> 
> A floating point format would be great, though, if that's possible at
> some point in the future. It would potentially have utility beyond my
> specific usage.

We (Inkscape) would love to see a floating point format to avoid issues
such as banding in gradients.

Tav

> :-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-
> William E. Kappler II
> LinkedIn · Blog · Project Website
> 
> On Thu, Jan 28, 2016 at 3:11 PM, Bill Spitzak <spitzak at gmail.com>
> wrote:
> > Non-premultiplying of the destination buffer would involve changing
> > all the compositing math.
> > 
> > For instance OVER changes from A+B(1-a) to (A+Bb(1-a))/(a+b-ab)
> > 
> > This almost always is more complicated, which is why premultiplied
> > is almost always used in rendering and compositing. Also in most
> > cases the background image has to be stored in floating point, as
> > the multiply and division here will get you the same inaccuracy you
> > are having with premultiplied.
> > 
> > If your Cairo rendering is a solid color (ie paths and fill and
> > fonts, but all using the same color) then you can get a non-
> > premultiplied version by saving only the alpha channel and
> > replacing the color with the solid color. This may cover a lot of
> > the uses you want for texture maps. You may also be able to bleed
> > out the color from alpha=1 pixels into neighboring transparent
> > pixels, I have used this as a method to remove premultiplied
> > inaccuracy.
> > 
> > A floating-point format for Cairo would be a good solution however,
> > as it would be accurate, whether or not it is premultiplied.
> > 
> > Non-premultiplied source images are supported in most cases by
> > choosing the correct compositing operations and using the alpha as
> > a mask.
> > 
> > 
> > On Thu, Jan 28, 2016 at 1:40 AM, Lawrence D'Oliveiro <ldo at geek-cent
> > ral.gen.nz> wrote:
> > > On Thu, 28 Jan 2016 02:39:17 -0500, William Kappler wrote:
> > > 
> > > > The issue is not any external format like OpenEXR, but that so
> > > far as
> > > > I am aware, there is no way to extract non-pre-multiplied
> > > information
> > > > out of Cairo.
> > > 
> > > If you really wanted to, you could trade off resolution for bit
> > > depth
> > > by doing oversampling.
> > > --
> > > cairo mailing list
> > > cairo at cairographics.org
> > > http://lists.cairographics.org/mailman/listinfo/cairo
> > > 
> > 
> > --
> > cairo mailing list
> > cairo at cairographics.org
> > http://lists.cairographics.org/mailman/listinfo/cairo
> -- 
> cairo mailing list
> cairo at cairographics.org
> http://lists.cairographics.org/mailman/listinfo/cairo