> I'm not sure whether emmisive, specular and diffuse lighting might give
> a different result here.
Hmm, I don't think things are that dramatic... Admittetly, the following
thoughts apply only to local (per-texel) image differences, but
the "big picture" shoudln't be worse of.
The specular term of a local reflection model normally does not depend
on the object surface color, it merely emulates a highlight. Diffuse lighting
could amplify a visual error locally if the local contribution of all light
sources is larger than 1 (I assume emissiveness is just a constant light
contribution term, similar to ambient). For a given wavelenght:
diffuse_exact = min(1.0, lossless_surface_color*light_contributions)
diffuse_lossy = min(1.0, lossy_surface_color*light_contributions)
It follows that
(diffuse_exaxt - diffuse_lossy) <= (lossless_color -
I.e., the difference image is scaled by the light source contributions.
So, only if an area is strongly lit (contribution > 1), artefacts will become
more noticeable than in an unmodulated view of the image.
Also keep in mind that most image differences will probably be in the same
ballpark as the discretisation error; so the same problem presents itself with
lossless textures, but probably in a less structued and therefore less
noticeable manner. You just can't get dynamic range where there wasn't, same
goes for spatial resolution.
Don't take my musings as gospel, though :-)
Flightgear-devel mailing list