Re: filmscanners: Why not sRGB ?

["Anthony Atkielski" <atkielski.anthony@wanadoo.fr>]

Steve writes:

> So what's the colour gamut of the average human
> eye and how much variance is there between people's
> perception ?

I believe that so-called Wide Gamut RGB is the closest match for the gamut of
human vision (which is very large); unfortunately, it is not a close match for
much of anything else, so it is rarely a practical colorspace to use in a
typical workflow.

There is precious little variance in color perception from one individual to
another.  Just about any variation produces so-called color-blindness.  Note
that "color-blindness" is most often nothing more than a slight shift in the
sensitivity curve of a person's eye for one of the three additive primaries to
which the eye is normally sensitive (usually red, if I recall correctly).  A
total inability to see color (achromatopsia) is very rare, and usually occurs
concomitantly with legal blindness (because it is usually caused by an absence
of cones in the retina, and without cones, it is impossible to see fine detail,
and blooming on the rods makes vision in bright light nearly useless in any
case).

> I bizarrely found during the colour blind discussion
> that I could change the hue of some of the colour
> charts such that I (CB) could very clearly see
> the correct number on the chart and so called "normal"
> people could see nothing but dots.

There is nothing bizarre about this.  When you shift the peak wavelengths of the
primaries used for vision, as it typically the case for the most common forms of
color-blindness, you do not necessarily see _fewer_ colors, you simply see
_different_ colors.  Some colors will look the same for you and for everyone
else (others who have normal vision, that is).  Some colors will look different
to you, but the same to normal people.  And some colors will look the same to
you, but different to other people.

Some arrangements of the tristimulus values might be more or less optimal than
the normal arrangement, but most simply shift the colorspace of vision, rather
than making it any larger or smaller in gamut.

> None of them match up - each has some colours that
> are not seen by other devices/people.

Quite so ... and you usually have to work with a colorspace that matches
everything in your workflow as closely as possible, or, if that isn't practical,
a succession of colorspaces that are successively more restrictive.

> Perhaps we should be amazed that we ever get a
> good match.

I agree.  Then again, we don't often get a good match.  At best, sometimes we
can only get an adequate match, and then we depend on human visual perception to
correct things so that colors look as they should (that's why fluorescent lights
don't look green to human beings, for example).

---

Maris writes:

> I don't know, but I suspect that the 4-color general/
> business application inkjets also print colors outside
> of the sRGB color space ...

Just about everything has colors outside sRGB.  My understanding is that sRGB
was largely based on the NTSC colorspace, which is probably about the worst
colorspace in the known universe, originally designed to accommodate the
space-age features of consumer electronics of the mid-20th century.  It's really
dismal.

These days I use mostly Adobe RGB in Photoshop.  It's so hard to calibrate and
match anything on today's desktop computers (without spending a fortune on
hardware and software) that I don't stress myself by trying to go much beyond
that.