[argyllcms] Re: relativ volume of gamut -- vrml comparison?
- From: Gerhard Fuernkranz <nospam456@xxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sun, 06 Jan 2008 00:31:12 +0100
Preben Soeberg wrote:
> Can AdobeRGB expand the camera's gamut?
> Can somebody explain what happens?
>
> timage -t -x -s timage_txs.tif
>
> Gamut for the camera profile:
> tiffgamut -v 6071_F.icc timage_txs.tif
> Total volume of gamut is 1062744.608376 cubic colorspace units
>
Before talking about your particular numbers, IMO the basic question is,
what's your understanding of a "camera's gamut"? How do you define the
gamut of an _input device_, like a camera or scanner?
Please find below a citation from a discussion in the newsgroup
sci.engr.color, with a statement from me and Graeme regarding this issue.
Regards,
Gerhard
------------------------------------------------------------------------
> Gerhard Fuernkranz wrote:
> [...]
> I basically do not like to use the term "gamut" in conjunction with
> input devices like cameras.
>
> For an output device it's pretty clear: The device/medium can reproduce
> only a particular set of XYZ colors (which is a subset of the visual
> gamut), and this set is the gamut of the device/medium.
>
> But how should we actually define the gamut of a camera? Should we
> simply define it as the set of colors which the camera can "see"?
>
> A typical camera has sensor sensitivities, which basically cover the
> whole visible spectrum (but they are usually not colorimetric and don't
> fulfill the Luther-Ives condition).
>
> Such a camera will "see" any visible color, and will return _some_ RGB
> values for each visible color. The camera may even "see" spectra outside
> the human visible range, and return some RGB values for these spectra as
> well.
>
> So the gamut of such a camera would be the full visible gamut (or even
> larger), if we would define the gamut simply as the set of colors which
> the camera can "see".
>
> But is this really helpful? IMO it isn't.
>
> The limitation of a camera is rather not, that it would not "see" some
> human-visible colors (as said above, typicaly cameras can basically
> "see" more or less all these colors), but the actual limitation is
> rather the "color-blindness" of the camera, caused by observer
> metamerism.
>
> This means, we will be able to find pairs of human-visible spectra,
> which are metameric from the camera's point of view (i.e. the camera
> will "see" the _same_ RGB values for both spectra), although the two
> spectra are clearly distinguishable by the human vision as different
> colors.
>
> So eventually the camera cannot distinguish some colors, which humans
> _can_ distinguish, and thus it isn't possible to establish an
> unambiguous RGB to XYZ mapping for the camera, which would be correct
> for arbitrary human visible spectra. Regardless which (linear or
> non-linar) RGB to XYZ mapping/characterization we try to establish for
> the camera, it will be correct for some spectra, but there will always
> exist other spectra, for which the mapping is incorrect :-(
>
> Thus I have rather no idea, how the actual limitation of a camera to
> discriminate colors could be simply described in terms of a "camera
> gamut". Just describing the set of colors the camera can "see" is IMO
> not sufficient and helpful.
>
> I could possibly imagine something like: First define a set of all
> spectra we intend to capture with the camera, and then define the gamut
> as the subset of these spectra, for which a RGB to XYZ mapping with an
> error below a given threshold is possible. Of course, according to a
> definition like this one, the gamut of many cameras is possibly no
> longer so huge (and it will of course depend on the admitted error
> threshold and the given set of spectra, which are desired to be captured
> with the camera).
>
> I'm not sure, whether a generally admitted (and useful) definition for
> the "gamut of input devices" exists, does it?
>
------------------------------------------------------------------------
> Graeme Gill wrote:
> [...]
> It's a controversial subject. I think your summary was a good one
> Gerhard. By a strict definition, no, an input device doesn't have
> a gamut. People do use "gamut" and input devices together in a loose
> sense through, since they have primary co-ordinates.
>
> Peter Morovic and Jan Morovic presented a paper "Spectral
> discrimination of color input devices" at CIC14 this year,
> addresses some of the issues, and they refer to something they call
> "device desciminability gamut".
>
> The following paper "Capture Color Analysis Gamuts" by Jack Holm,
> a title that raised mutterings of "input devices don't have
> gamuts" amongst some of the attendees :-), actually addressed
> the apparent spectrum locus that a camera perceives, and how
> this compares to the visible locus. Summary :- a camera may
> generate XYZ values outside the visible spectrum locus for
> certain spectral input.
>
> Graeme Gill
Other related posts:
