Hi guys and gals,

I recently wrote this plugin for my own amusement; someone else might find a reasonably use for it.

It probably does something very similar to the 'antialias' plugin for GIMP 1.2, except this is for GIMP 2.0 (2.X?) and uses a different algorithm (I haven't compared results).

http://icculus.org/~aspirin/gimp/auntiealias/

--Adam

Matt wrote:

Is it possible to use this under win32?

It's entirely possible, if someone wants to compile it up for win32. I'm not equipped to do so myself.

--Adam

Hi there!

Michael Natterer wrote:

yesterday, i stumbled across this mail, and it looks as if this plug-in is doing exactly the right thing to lineart images.

I "ported" it to 2.4 API and codingstylized it a bit:

http://mitch.gimp.org/auntiealias.c

Since we consider including it, I have some questions:

- The licensing would be changed to GPL, would that be ok?

It wouldn't be my first choice as I took pains to non-taint it in the first place, but re-licensing to GPL is one of the things implicitly permitted by the existing license so if you must, do so.

- Do you have any link to that Scale3X algorithm? I would be nice to include an URL in the help.

This was *probably* the page I used as reference: http://scale2x.sourceforge.net/algorithm.html though to be fair that's just a page of pseudocode not especially clearer than the plugin's. :)

- Is the behavior on images with alpha intentional? (it only affects opaque areas, but does not antialias between opaque and transparent parts)

It refuses to antialias between not-entirely-transparent and entirely-transparent to avoid the possibility of giving some partial opacity to undefined colours. This was intentional but heavyhandedly implemented; there is likely to be a more delicate compromise.

- Do you maybe have a newer version around?

I'm afraid not; it was a one-day fire-and-forget experiment. Since there's some interest I'll have a quick hack to change the moronic subsambling kernel right now (n.b. that's not related to the alpha thing) and if the results are better I'll submit a patch.

Cheers,
--adam