[opendtv] Re: All colors of rainbow in new display concept

  • From: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Mon, 11 Sep 2006 15:17:53 +0200

Hello, 

(I've just returned from my vacation.) 

Yes, this seems intriguing. If you can make 2 arbitrary rainbow colors 
of arbitrary intensity then you can mix them to all colors on the 
line that connects them, and then you can essentially make ALL visible 
colors. E.g. you would make white by adding cyan + red, or blue + yellow. 

There is a big catch though: the 2 primary colors would be generated 
from the white input (back)light at very low efficiency, because their 
spectra are very narrow. For this reason alone this is a bad idea ! 

A grating might be a less selective band pass filter, but then the 
rendered colors, especially the greens, would be not so pure. 
A typical example of this is the colors that you see in a soap bubble. 


What Don Munsil says about xvYCC is absolutely true. It is no trouble 
at all for current transmission systems to encode a wider color gamut. 
By reducing the amplitude of Y' you can make the gamut arbitrarily 
large. (But that's just another way of saying that you could apply 
a saturation gain factor to the CbCr signals...) 
xvYCC is based on standard Rec.709 colorimetry and signal amplitudes, 
so the added cost of wide gamut is next to nothing ! This is why it 
was so easy to get it standardized (again) as IEC 61966-2-4. 

There is a very nice Powerpoint document about this xvYCC standard: 

        "The world with widegamut color imaging"
        - xvYCC: Extended-colour space for video applications -
        Authors:
        Naoya KATOH
        Yoshihide SHIMPUKU
        Takehiro Nakatsue
        Color Rendering Community (CRC)
        Sony Corporation
        (Nov.7/2005)

I'm hoping to have a discussion about xvYCC with Charles Poynton, 
tomorrow over dinner after the IBC has ended. 


The real cost is of course in the display, where those wide gamut 
colors must actually be rendered. Today's trends are LED backlights, 
wide-gamut CCFL backlights, and multi-primary displays (by Genoa). 
It's easy enough to cover the entire gamut of surface colors, and 
I question the need to go beyond that. 


Camera design for wide color gamut should be trivially easy. There 
is never any need to go beyond 3 primaries, because the behaviour 
of the human eye is defined entirely by only 3 primaries. They are 
called LMS, for low-medium-short wavelengths. Any optical filter 
has non-negative gain for all frequencies, therefore it represents 
a color primary that must lie outside the horseshoe of spectral 
colors. Then the only requirement of an accurate color camera is 
that its 3 primaries are linear combinations of the LMS primaries, 
or of the XYZ (1931) primaries (which are themselves a linear 
combination of LMS). So an XYZ camera would be just fine, though its 
color space is larger than necessary and its color S/N will not be 
optimal. Then a 3x3 matrix transforms this space to the Rec.709 RGB 
primaries, and this (!) is where negative RGB values will occur. 
In a standard gamut camera, these negative values are removed by 
clipping (or any other form of gamut mapping). We only have to 
remove this clipping circuit to obtain a wide color gamut camera.. 


xvYCC defines how these negative values must be carried through the 
transmission chain. That's rather the trivial part of it all, it 
is only digital video signal processing. The color space conversion 
matrix for a wide-gamut RGB display will make these values less 
negative. The display can render only positive values, of course. 

The real question is: how useful is it all ? The original NTSC 1953 
color gamut (the one that is NOT in use right now !) is actually 
quite nice. Attempts to go beyond that will mostly lead to a loss 
of efficiency, and not to much user benefits. Just my 2 cents, YMMV. 

Best regards, 
-- Jeroen

+-------------------------------+------------------------------------------+
| From:     Jeroen H. Stessen   | E-mail:  Jeroen.Stessen@xxxxxxxxxxx |
| Building: SFJ-5.22 Eindhoven  | Deptmt.: Philips Applied Technologies |
| Phone:    ++31.40.2732739     | Visiting & mail address: Glaslaan 2 |
| Mobile:   ++31.6.44680021     | NL 5616 LW Eindhoven, the Netherlands |
| Skype:  callto:jeroen.stessen | Website: http://www.apptech.philips.com/ 
|
+-------------------------------+------------------------------------------+

Other related posts: