Here's a render by my own modest little ray tracer/photon mapper LiAR isn't a raytracer. What you're seeing is three car models from Live for Speed with team skins of PHD-Motosports of which I'm the "official sponsor", in the wonderful Uffizi gallery.
The body surface is using the famous Ashikhmin & Shirley BSDF. The glass of the windows and head lights use some modified fresnel equations. Because the glass is modeled as a single intersection, I had to cramp the double air-glass-air interface + the attenuation by the glass medium in one BSDF. In a nutshell:
where and are the resulting reflectance and transmittance, is your usual fresnel reflectance, , the transparency of the glass medium (Beer's law) for orthogonal incidence, and the internal angle of the refracted rays (Snell's law).
Would a double interface not give more realistic effects?
Sure, but the model doesn't model it that way. So, that's why the "hack". As long as it is thin enough (but not too thin), and you're away far enough, you shouldn't see much difference.
Dunno if you know about it, but the Radiance ray tracer used a model like this for their glass, e.g. some kind of closed form equation for a series of bounces of light in a pane of glass.
Dunno if you know about it, but the Radiance ray tracer used a model like this for their glass, e.g. some kind of closed form equation for a series of bounces of light in a pane of glass.
You may be referring to something analogous to the Fabry-Perot equation:
Basically each internal reflection adds to the main reflection by a phase, and the infinite series of reflections is a geometric one, which can be summed. The wikipedia article is probably a the quickest overview.
Hmm.. i'm not sure if the width of the glass panes is small enough, and the surface regular enough, for phase-related effects like interference to come into play.
I think the Radiance ray tracer just dealt with the result of a series of reflections using fresnel terms, without considering interference.
Well, the derivation is simple enough for me not to expect I would be the first. So I'm not surprised Radiance has something similar, in contrary. Apart from them having split up the thing for TE and TM waves (which is of course the right thing to do), it doesn't differ from what I have. They also seem to apply Beer's law for medium attenuation, just like I do.
Indeed, the approximation assumes incoherent sum of the terms: the pane is thin enough for the reflections/transmissions to have local (coincident) appearance to the camera, but not thin enough to have any phase correlated effects. The light source is assumed to be incoherent as well.
Still, maybe I will write it down someday, or maybe not =)
Bramz
You need to log in in order to reply to this topic.
