[Celestia-developers] Re: lighting etc

[Constantine T. <con...@sh...>]

(I sent this from the wrong email address first, dunno if that post will 
get through, apologies if this is a double post)

At 11:38 PM 12/12/2005 -0800, you wrote:
>All of the experimentation I've done up to this point is relevant for both
>traditional (8 bit per channel fixed point) and floating point frame
>buffers. I'm trying to get rid of all the random parameters in the star
>rendering code; the current mapping of apparent magnitude to star disc
>size and brightness looks sort of right in a for a limited range of
>sensitivities, but it falls apart when you set the limiting magnitude to a
>large value, and the apparent brightness of stars isn't properly
>correlated with the apparent brightness of galaxies. The only way to
>address the problem is to start over from the physical light intensities
>and consider how these affect the simulated detector, and that's what I'm
>doing.

Honestly, I didn't understand much of what you said here.

First, I am still very uncomfortable with this 'modelling the human eye's 
response' lark - it seems very arbitrary and 'analog'. I'd rather you used 
an explanation along the lines of modelling a CCD's response such that the 
brightness of the object was a standard value. So Mercury would look as 
bright as Pluto because in Mercury's case you're modelling a brief exposure 
of the CCD, and in Pluto's case you're modelling a long exposure of the 
CCD, or something along those lines. The 'human eye reponse' opens up a big 
can of worms, including for example the fact that at low enough light 
levels a human eye wouldn't be able to distinguish any colour in the object 
being viewed. Either way I think a better approach would be to stop trying 
to simulate specific detectors and instead just try to simulate what is 
actually THERE (ie the structure, the general appearance etc) because the 
viewer is always going to be aware that he is watching a simulation on a 
monitor that can't possibly recreate the true appearance of what is out there.

Second, planetary appearance is determined by the photometric function of 
the surface being viewed - which is either Lambertian, Lommel-Seeliger (in 
the case of the moon and most rocky airless bodies), some combination of 
the two (i.e. a Buratti function which combines Lambertian and 
Lommel-Seeliger terms), Raleigh scattering (in the case of atmospheres), or 
if you want to get really detailed a complex Hapke function which 
incorporates all sorts of terms for the scattering properties of the 
regolith on the surface. The parameters for these functions are fairly 
well-known for most of the major bodies in the solar system, I think there 
are several scientific papers that list the Buratti parameters. Frankly, I 
think that doing anything regarding planetary appearance that doesn't 
attempt to simulate this is only going to take Celestia further from being 
a realistic simulator of the universe - Celestia MUST eventually 
incorporate these functions for planetary bodies to be credibly realistic. 
I can't think of any reason why these should be impossible to simulate in 
Celestia, unless there is something about them that requires specific 
raytracing routines that can't be implemented in OpenGL, in which case I 
guess we're screwed. I do know that there is at least one person out there 
- Bjorn Jonnson - who has created software that can render planets using 
these functions. I've linked to his pages to illustrate the appearance of 
surfaces with these functions several times on the forum.

Basically, 'instrument sensitivity' shouldn't matter for rendering planets 
in Celestia. You can't show them realistically anyway because a monitor 
simply can't reproduce the falloff of illumination with distance. I think 
what you should be focussing on is getting the photometric functions 
implemented in the code, and never mind the human eye response or CCD 
response or how bright it really is. That's already been comprimised by the 
limitations of the medium (heck, stars aren't even rendered remotely 
realistically when viewed close up - you're certainly not seeing a 'human 
eye response' when you look at the sun from Mercury in Celestia), trying to 
correct that is going to be futile I think.