Hi,

> For normal photographs that's great - for textures that get scaled,
> projected, sheared (sp?), lit, ... the uses assumptions dodn't hold
> anymore.

Why should projection, shearing, scaling be a problem? Wouldn't every JPEG
image displayed on your computer screen would look lousy when looking at your
display from an angle, if this was generally true? 

The problem here is resampling (as indeed happens with texture mapping). If
you sample your textures incorrectly, things might look worse than expected,
but nobody says it isn't possible to do things correctly. 

This is why I don't like the article you referenced; it sure has some valid
points, but they are presented in a manner which is a bit across-the-board.

Lighting can be more of a problem, but areas are more often dimmed than
fully lit, so I'd assume the the difference (i.e. error) between lossless and
lossy texture is also dimmed. 

The one thing that really looks crappy is JPEG-compressed normal maps.

> An extreme example: when you use a very high compression rate you'll see
> the blocking artefacts. So you use a not so high compression and are
> hapy with the result. If you zoom into the picture you'll start to see
> the blocking again as the pixels got large enough.

If you zoom in even further, you will single out individual pixels. 

> So JPEG isn't usefull.

Well, I think it is. Is surely isn't optimal, but things don't look nearly 
as bad as one would assume after heaving read "JPEG's are evil, too" :-)

> One solution that might work could be wavelets. (This is where JEPG2000
> gets interesting again). But the wavelets used would need to be choosen
> carefully.

Maybe the gwic library would be worth a look...

bye,

 Manuel

_______________________________________________
Flightgear-devel mailing list
Flightgear-devel@flightgear.org
http://mail.flightgear.org/mailman/listinfo/flightgear-devel
2f585eeea02e2c79d7b1d8c4963bae2d

Reply via email to