Re: RGB color: enhancement request

On 02/08/07, Theo Zweers wrote:

> Don Cox wrote:
>
> ...snip...
>
>>; We really need CMYK monitors.
>>
>>; I remember some talk of an experimental one some years ago, but it
>> never appeared.
>>
>>; If you look at the gamuts of RGB and CMYK on a CIE colour chart,
>> they do
>> not exactly overlap: each includes colours that cannot be shown in
>> the
>> other. The Lab colour space does have a larger gamut which includes
>> all
>> the colours of the other two - but there are no Lab display devices.
>>
>
> Take a look at: http://www.hdmi.org/pdf/HDMI_Insert_FINAL_8-30-06.pdf
>
> Figure 3 is about xvYCC (extended gamut YCC).
>
> So with HDMI 1.3, the interface between your screen and DVD/PC/etc.,
> there is talk about Deep Color, xvYCC. Less banding figure 1, but also
>; all the colors in the world (as matter of speaking).
>
This is about increasing the number of bits per channel.

We have to distinguish clearly between the number of colours and the
gamut.

For example, you can change greyscale from 4 bit to 8 bit or 16 bit, and
this will increase the number of shades of grey from 16 to 64k. But the
image is still greyscale.

However, if you go from an 8 bit grey image to an 8 bit color image
(typical GIF image), there are still 256 colors, but they cover a wider
gamut.

xvYCC increases the number of colours but only slightly increases the
gamut (every little helps, of course). Naturally they will sell it as a
colossal improvement.

The curve around the outside of a CIE diagram represents the pure,
monochromatic, single-wavelength colours. These can only be produced by
either using a spectroscope (which throws away almost all the light) or
a laser source.

Given three lasers at well spaced positions in the spectrum, say R G and
B, you could display most of the possible colors - those within that
triangle of points on the diagram. Adding a fourth primary, say a Cyan
laser, would give you all the colours within that quadrilateral. If the
R and B are not at the extreme ends of the spectrum (400 and 700 nm),
you would not be able to display the most saturated purples and magentas
- the colors that lie along the straight line at the bottom.

In practice, a scanning-laser display could cover almost all the visible
gamut of colours. How many distinct colours would be shown _within_ that
gamut depends on the number of bits per channel.

There is no equivalent in dyes or pigments of a laser. All the practical
colorants are relatively unsaturated (so the claim I saw that xvYCC can
reproduce all the Munsell colors is not very impressive).

Regards
--
Don Cox
doncox%40enterprise.net">doncoxenterprise.net

Messages in this topic (0) Reply (via web post) | Start a new topic

.

Don Cox wrote:
> On 02/08/07, Theo Zweers wrote:
>
...snip...
>> Take a look at: http://www.hdmi.org/pdf/HDMI_Insert_FINAL_8-30-06.pdf
>>
>> Figure 3 is about xvYCC (extended gamut YCC).
>>
>> So with HDMI 1.3, the interface between your screen and DVD/PC/etc.,
>> there is talk about Deep Color, xvYCC. Less banding figure 1, but also
>;> all the colors in the world (as matter of speaking).
>>
>>
> This is about increasing the number of bits per channel.
>
> We have to distinguish clearly between the number of colours and the
> gamut.
>
> For example, you can change greyscale from 4 bit to 8 bit or 16 bit, and
> this will increase the number of shades of grey from 16 to 64k. But the
> image is still greyscale.
>
> However, if you go from an 8 bit grey image to an 8 bit color image
> (typical GIF image), there are still 256 colors, but they cover a wider
> gamut.
>
> xvYCC increases the number of colours but only slightly increases the
> gamut (every little helps, of course). Naturally they will sell it as a
> colossal improvement.
>

It goes to the full colors that monochromatic colors can represent. That
is an increase of colors!
However you are saying that this only a part of the color world, it's
not 2 dimensional (as this CIE diagram in the PFD shows), but it's
3-dimensional.
(I took a look at: http://en.wikipedia.org/wiki/Gamut)
So they bring us not all the colors in the world as I thought (but still
a bit better than we now see on our screens, and really better than my
own 18 bit screen).

> The curve around the outside of a CIE diagram represents the pure,
> monochromatic, single-wavelength colours. These can only be produced by
> either using a spectroscope (which throws away almost all the light) or
> a laser source.
>
> Given three lasers at well spaced positions in the spectrum, say R G and
> B, you could display most of the possible colors - those within that
>; triangle of points on the diagram. Adding a fourth primary, say a Cyan
>; laser, would give you all the colours within that quadrilateral. If the
> R and B are not at the extreme ends of the spectrum (400 and 700 nm),
>; you would not be able to display the most saturated purples and magentas
> - the colors that lie along the straight line at the bottom.
>
> In practice, a scanning-laser display could cover almost all the visible
> gamut of colours. How many distinct colours would be shown _within_ that
>; gamut depends on the number of bits per channel.
>
>
> There is no equivalent in dyes or pigments of a laser. All the practical
> colorants are relatively unsaturated (so the claim I saw that xvYCC can
> reproduce all the Munsell colors is not very impressive).
>
> Regards
>

Thanks for your explanation!

Theo

Messages in this topic (0) Reply (via web post) | Start a new topic

.

Theo:

> ...So they bring us not all the colors in the world as I thought (but still
> a bit better than we now see on our screens, and really better than my
> own 18 bit screen).
>

Of course, a lot of this is impractical. The number of colors we can
distinguish under laboratory conditions using highly specialized tests
and the number we can actually distinguish in real life around us is
quite different. I no longer remember the practical limits, but I
believe it's in the low 5-digit area. The quest to get natural color in
the various forms of photography is largely one of adjusting the biases
of various ways of recording color. But, at the same time, people tend
to like some of these biases and that's part of the "art." In
traditional photography, some people preferred Ektachrome and others
Kodachrome. I took to using Japanese brands when I was vacationing in
Japan because they had a way of catching the colors people used in
painting homes and businesses, particularly the greens.

Digital cameras of all types have their own biases as you'll see in the
detailed test reports on the Internet. To what extent exploding the
gamut or any other measurable dimension of color would eliminate that,
I've no idea. But color matching between media is a toughie. Consider,
for example, that many motion pictures are now being archived in digital
formats, and those digital prints are likely to be the standard
reference for all future uses. Consider that in making those archive
copies, there's a human being making decisions about what the colors
should be and making that decision based on what (s)he sees on a monitor
with its own peculiarities. (Hopefully also based on a reference print
of the film from an original negative which hasn't deteriorated
significantly.)

But then those matters have never been "perfect." Even in the
traditional film environment, there's a laboratory and people called
"color timers" making similar decisions.

HB

Messages in this topic (0) Reply (via web post) | Start a new topic

.