home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.graphics
- Path: sparky!uunet!stanford.edu!ames!data.nas.nasa.gov!wk207.nas.nasa.gov!uselton
- From: uselton@wk207.nas.nasa.gov (Samuel P. Uselton)
- Subject: Re: Is POV optically accurate ?
- References: <1992Nov12.133317.5833@genes.icgeb.trieste.it>
- Followup-To: comp.graphics
- Sender: Sam Uselton
- Organization: NAS, NASA Ames Research Center, Moffett Field, California
- Date: Thu, 12 Nov 92 18:13:51 GMT
- Message-ID: <1992Nov12.181351.23378@nas.nasa.gov>
- Lines: 58
-
- In article <1992Nov12.133317.5833@genes.icgeb.trieste.it> oberto@genes.icgeb.trieste.it (Jacques Oberto) writes:
- >I am trying to reproduce one of Newton's experiments with POV.
- >Would a simulated white light beam hitting a glass prism with the
- >right angle be scattered into a rainbow spectrum ?
- >Are POV and other ray-tracers optically accurate in that respect ?
- >Has anybody tested the properties of 'raytraced' light other than
- >reflection and refraction ?
- >
- >Any comment appreciated !
- >
- >Jacques
- >
- >
- >--
- >Jacques Oberto <oberto@genes.icgeb.trieste.it>
-
- I don't know POV................BUT
-
- If it is in the tradition of the standard ray tracers, it starts at
- the eye, shooting rays through pixels into a scene. Essentially this
- is tracing ray paths in reverse. Generally speaking, ray paths are
- reversible, so this is fine. There are, however, difficulties. When
- a ray hits an object, in addition to reflected and refracted rays,
- rays to the light sources must be generated to determine shadows,
- intensity of illumination, etc. If one of these light source rays
- hits a refractive object, it may no longer be headed for the light
- source once the refraction is done.
-
- What one would LIKE to do is shoot the light source ray in a direction
- (there may be several correct choices) which will result in a ray
- pointed at the light AFTER the refraction. Guessing what direction(s)
- this might be is HARD. One solution is to shoot lots of illumination
- rays in various directions (especially from a surface that is not
- perfectly specular, use the BRDF as the distribution being randomly
- sampled => distribution ("distributed") ray tracing).
-
- Another solution is to do the ray tracing from the light, but then
- most of the rays won't get to the screen so the effort is wasted.
- There are various schemes in the literature for getting around the
- difficulties, but my guess is that they are not in most public domain
- code.
-
- Another difficulty probably ignored in most PD code, is that in order
- to break out the spectrum, the refraction angle must be calculated on
- a wavelength dependent basis, generally with a single ray turning into
- several rays to properly sample the spectrum.
-
- We (Redner, Lee & Uselton ++)have done an image of the experiment I
- believe you are describing. It was accepted into the SIGGRAPH slide
- set and distributed last summer at the conference. It does involve a
- forward ray tracing phase, and some stuff to remember these results in
- a way that can be used in a traditional backward ray tracer. Harder
- than it looks.
-
- Hope this helps.
-
- Sam Uselton uselton@nas.nasa.gov
- employed by CSC working for NASA (Ames) speaking for myself
-