On Sunday 25 February 2007 19:07, Felix Nawothnig wrote:
> Hi.
>
> Just some crazy idea I decided to throw in - how about hardware
> acceleration for the scale2x, etc. family:
>
> http://scale2x.sourceforge.net/

There's been some discussion about texture compression some time ago. 
S3TC (was it called that?) seems to be the standard, but it's patented, 
so if we can avoid it we should. I've had some ideas on this in the 
back of my head, but haven't had time to experiment. They're similar to 
this though.

Basically, my idea was that it may be possible to predict how to split 
up a pixel into four subpixels from the surrounding pixels. What I 
planned to do was to make a programme that would record various 
statistics of each 3x3 block of pixels as well as the correct 
subdivision of the centre pixel (I'd be working with a half-size 
version of an existing image, so that I'd have ground truth). I'd feed 
it a bunch of textures, and then I'd try to find correlations.

This algorithm looks like it could be the result of such a process. I'm 
not sure that it's directly applicable to texture compression though: 
the example images are all low-resolution 'pixeled' sprites; these days 
textures are more like high-resolution photographs. But perhaps 
something similar could be done, and if it's generic (or similar) 
enough then this scale2x stuff could use the same circuit.

Lourens

Attachment: pgpYjGieHMy4O.pgp
Description: PGP signature

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to