home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!uknet!mucs!lilleyc
- From: lilleyc@cs.man.ac.uk (Chris Lilley)
- Newsgroups: comp.graphics
- Subject: Re: Is there a mathematical relationship RGB -> wavelength?
- Message-ID: <5760@m1.cs.man.ac.uk>
- Date: 14 Aug 92 13:32:23 GMT
- References: <92225.171906POLOWINJ@QUCDN.QueensU.CA> <dangold1.713711275@vincent1.iastate.edu> <1992Aug13.233117.21185@cco.caltech.edu>
- Sender: news@cs.man.ac.uk
- Reply-To: lilleyc@cs.man.ac.uk (Chris Lilley)
- Organization: Department of Computer Science, University of Manchester UK
- Lines: 108
-
- In article <1992Aug13.233117.21185@cco.caltech.edu>
- nyet@cco.caltech.edu (n liu) writes:
-
- >dangold1@iastate.edu (Daniel M Goldman) writes:
-
- >>In <92225.171906POLOWINJ@QUCDN.QueensU.CA> Joel Polowin
- <POLOWINJ@QUCDN.QueensU.CA> writes:
-
- >>> > For curiosity's sake, I was wondering if it is possible to take a set
- >>> > of RGB color fractions and come up with an approximate wavelength of
- >>> > the observed mixture. [...]
-
- >> In dealing with real light, we run into metamers, colors which look
- >> the same under one light source, but look like different colors
- >> under a different light source. [...]
-
- >Basically, the three coordinate system was made up since we have 3 mostly
- >independent color receptors.
-
- No, it was made up because matching studies showed that many colours
- could be matched with only three lights. All colours can be matched
- with any three colours, if you allow them the clever property of
- negative contributions.
-
- This does not relate directly to the cones in the retina.
-
- >RGB space manages to cover most (but not all)
- >of the color gamut we are able to see.
-
- Actually , a typical monitor covers less than a third of the colours.
- But it covers the most common, less vivid ones.
-
- >Keep in mind that our color sense is
- >extremely crude, i.e. we can't tell the difference between pure magenta
- >and a spectrum that contains a spike at blue and a spike at red.
-
- Yes, all colours are seen as a single sensation.
-
- >This is because
- >our receptors are stimulated "identically" in both cases.
-
- Not necessarily. There is a lot of processing going on, and a lot of
- psychological effects to consider.
-
- >CIE did a huge survey
- >with a bunch of subjects (does this look the same distance in blue to this
- >sample as the distance in blue between these other samples? etc.) and came
- >up with
-
- A set of matching functions for the three lamps they were using (the
- amount of each lamp required to match each single-wavelength colour
- (down to 1nm bands) throughout the visible spectrum. They transformed
- these to another set of imaginary primaries which do not ever have to
- be mixed in negative amounts to match any visible colour.
-
- >an average cone response for each of the three cones.
-
- No, but I see why you said this.
-
- >(foley and van damn has some stuff on this i think).
-
- Very little, and that not very clear. FvDFH is a great book for c.g
- geometry and stuff, but the section on colour is IMHO weak.
-
- >The CIE colorspaces are a very close
- >approximation to these responses.
-
- The CIE colourspace is *exactly* these responses. Given the measured
- spectrum of a colour, the matching functions are used to compute the
- amount of each CIE primary required to match each band of wavelengths
- (by multiplication, ie a weighting scheme).
-
- Adding these amounts over the whole spectrum gives the amount of CIE
- primaries to match the given colour. This is the CIE XYZ model,
- published in 1931
-
- >In any case, you can look up these response
- >curves and do a bit of math by superposing the responses to a particular
- >wavelength and then convert to RGB (i forget what the standard wavelengths
- >are for RGB...
-
- Pretty close, see above. As you note, colours are indivisible
- sensations so by this point the actual wavelengths have become lost.
- If you think of it as information reduction - 10nm samples (say) from
- 380nm to 730 nm is a lot of info, boiled down to just 3 numbers (XYZ).
- The wavelength info has been lost by this point.
-
- RGB do not correspond to individual wavelengths. They are not spectral
- (monochromatic) colours. Also, each RGB device has *different* colours
- of RGB (in some cases, very different. The NTSC monitor is a good
- example).
-
- Given a colour in XYZ, and the spec of your monitor, and the gamma, an
- XYZ value can be converted to RGB for any desired monitor.
-
-
- >nyet@cco.caltech.edu
- >nyet@aerospace.aero.org
-
-
- --
- Chris Lilley
- ---------------------------------------------------------------------------
- Technical Author, ITTI Computer Graphics and Visualisation Training Project
- Computer Graphics Unit, Manchester Computing Centre, Manchester, UK
- Internet: lilley@cgu.mcc.ac.uk Janet: lilley@uk.ac.mcc.cgu
- Voice: +44 (0)61 275 6095 Fax: +44 (0)61 275 6040
- ---------------------------------------------------------------------------
-