RE: [GD-General] software for artwork

[Tom F. <to...@mu...>]

Yep - we premultiply, palettise, and then un-premultiply again. Works well.
We've also spent quite a lot of effort finding good palettisers.
Kohonon-network based ones are good, and weused that for the last project.
The new one is a mad mix of various perceptual techniques and stuff (we have
a chap fascinated by this stuff), works in an arbitrary number of
dimensions, but is very very slow. We plan to use the old kohonen one for
quick previews, and then overnight crunch away at the textures with the
slow-but-superb one.

But we never found any decent out-of-the-box ones.


Tom Forsyth - Muckyfoot bloke and Microsoft MVP.

This email is the product of your deranged imagination,
and does not in any way imply existence of the author.

> -----Original Message-----
> From: Javier Arevalo [mailto:ja...@py...]
> Sent: 13 November 2002 09:11
> To: gam...@li...
> Subject: Re: [GD-General] software for artwork
> 
> 
> Mickael Pointier <mpo...@ed...> wrote:
> 
> > Standard quantization algorithms create artefacts, but thats not
> > visually to disturbing when applied to color components, 
> surprisingly
> > if you extend these algorithm to 4d, the final result is really,
> > really crappy... 
> > 
> > I think the reasons are multiples, the first beeing that when A tend
> > toward 0, you do not really care about the true values of RGB
> > components (they are almost invisible at this point), so a lot of
> > colors are used in semi translucent tints, that could have been used
> > in nearly opaque areas. 
> 
> Yeah that's one of the places where premultiplied alpha 
> textures help...
> 
>   Javier Arevalo
>   Pyro Studios