Sub-pixel positioning

[Adrian Page]

On Tuesday 16 March 2004 7:08 pm, Boudewijn Rempt wrote:
> Whichever would be easier & faster... I'm betting on a simple data
> structure for small bits of images, but perhaps we need to explore both
> alternatives, and then choose whichever is best.

The simple data structure is probably going to give the best performance, it 
just means writing a bit more code to get it going, I guess.

> It was meant to elicit some discussion. The advantages of rendering each
> layer to rgba before layer compositing are clear: it means you can
> basically combine anything with anything, if needs be even, eventually,
> karbon layers or whatever. The disadvantage is that I'm not sure what
> combining, say, cmyk, in rgba space is going to compared to combining as
> cmyk.

Perhaps we should keep to a single colour space if there's only one in use, 
and only convert to RGBA when there are mixed layers present. The little that 
I've found on the web seems to suggest that the same formulas can be used to 
blend CMYK channels as RGB.
http://www.penguin.at0.net/~fredo/files/PDF-DOCS/PDF_Transparency.pdf talks 
about these issues.

> > However, once we're in RGBA land, it's
> > easy going. I don't see any problem with doing this before the alpha, and
> > I would suggest it be done before any more of the compositing functions
> > are fixed, as they're simpler in premultiplied form. I'll happily
> > volunteer to do this. :-)
>
> Ah -- then there's not much choice, is there -- whoever codes the code gets
> to call the tune.

That's true in a practical sense, although personally I'd be happy to code 
whatever will give the best result, as that's the most important thing to me.

Adrian