Re: [css-color] . 12-Bit deep color support.

On 10/12/2013 10:57 AM, Daniel Beckstein wrote:
> Summary of this message:
>
> -          Maintain the #FF FF FF notation
> -          Additional xRBG notation, covering numbers till 1023, for simple
> 10/12-Bit  support
> -          Handle the 10 to 8 Bit fallback, with rounding down to the next 8
> Bit value
> ___________________________________________________________________
> I see that the ability to specify an exact color for deep color spaces won't
> be so important,
> because the human eye can just distinguish between 255-500 different shades,

This is not quite true but then, the above statement is ambiguous. True 
color (24-bit) has at least 256 shades of red, green, and blue. 
Likewise, you have 256 shades of a greyscale.

If you are talking about a particular hue, then you have less shades of 
such a hue as you move away primary colors (red, green and blue) and the 
secondary colors (aqua, fuchsia and yellow). For each of the primary and 
secondary color sets (6 hues), you have over 30,000 shades of for each hue.

> so deep color would matter most in transitions and in images (maybe drawn on
> the canvas element) ,
> where the computer generates the shades in between, the web-develper just
> specifies the borders.
>
> That`s why I actually don't see a need in changing the old notation #FF into
> something like #FFF,
> unlike I proposed beforehand,
>
> but there could be a need for an
> -          extended xRGB/xRGBA notation with numbers up to 1023 (covering
> 10-Bit)
> -          and the option to address 12-Bit colors in the form of  1022.25
> -          with a maximum of almost exactly 1000, instead of 255 it is
> very easy to break down the percentage of saturation 100 -> 10%

This does not make sense. Saturation is independent from sRGB color. 
Saturation is a component of HSL color.

> This format could then be used by the ambitious web developer to easily
> address the exact colors.
>
> Using the recent rgb notation covering 10 Bit works well (254.25),
> but covering 12 Bit will become quite an effort (254.0625).

There is no reason to do this. All you are doing is deceasing the 
lightness and saturation of the RGB components. A sRGB color cube has
16,777,216 colors. The human eye can discriminate up to ten million 
colors but only half of all perceived color can be seen displayed in 
sRGB. Why do you need places for over 68 billion colors?

> The fallback from 10 Bit to 8 Bit,for simplicity, should always be rounding
> down (IMO)
>
> 254.25 -> 254
> 254.75 -> 254

Are you talking about 'Rec. 2020' color which has a wider gamut than 
Rec. 709 color [2]? Depending on the available or envision display 
devices, the difference is the level of luminance [3] or the level of 
luma [4]. It's the overall difference between the black level [5] and 
white points [6] of a display device.

One way to imagine luminance or luma is to take a photo of a set of 
traffic lights (red, amber and green) using sRGB. Now load that up on a 
sGRB display and compare the difference between the display device and 
the traffic lights. You will still see red, amber and green (though the 
red and green may be clipped) but up close, the traffic lights will be 
far more brighter color than what is on a display device.

It would be good to understand color depth [7] and appreciate what the 
below quoted text is talking about.

   | As the number of bits increases, the number of
   | possible colors becomes impractically large for
   | a color map. So in higher color depths, the color
   | value typically directly encodes relative
   | brightnesses of red, green, and blue to specify
   | a color in the RGB color model.



1. http://en.wikipedia.org/wiki/Rec._2020
2. http://en.wikipedia.org/wiki/Rec._709
3. http://en.wikipedia.org/wiki/Luminance
4. http://en.wikipedia.org/wiki/Luma_%28video%29
5. http://en.wikipedia.org/wiki/Black_level
6. http://en.wikipedia.org/wiki/Standard_illuminant#White_point
7. http://en.wikipedia.org/wiki/Color_depth#Direct_color


Alan


-- 
Alan Gresley
http://css-3d.org/
http://css-class.com/

Received on Tuesday, 10 December 2013 09:01:52 UTC