[Openicc] beyond 8-bit precission [was: new version of xcalib]

[Bob Friesenhahn]

On Tue, 8 Mar 2005, Glenn Randers-Pehrson <glennrp> wrote:
>
> Samsung claims that their latest plasma display delivers 10,000:1
> contrast
> ratio.  I don't know if that is believable though, and naturally you'd
> need
> to drive it with 14-bit samples (maybe less with an appropriate gamma
> curve)

Contrast ratio numbers quoted by home entertainment display vendors 
are almost always on/off measurements.  This means that they turn all 
of the pixels off, take the low measurement, and then turn all of the 
pixels on and take the high measurement.  This results in much higher 
reported contrast levels (often 10X) than the ANSI white/black squares 
method.  Due to necessary biasing (no, you can't just turn off all the 
transistors!), background noise, internal reflections, and 
interference from other pixels, it is much more difficult to obtain 
high contrast ratios when displaying actual image data.

Lately we have seen contrast enhancement tricks like adding irises to 
projectors so that consistently darker images may be brought into the 
usable span of the display element.  For Plasma, it is likely that 
special circuitry is added to strap each element to ground for the 
special case of "black".

Regardless, if a human can see 300:1, and there are only maybe 200 
usable levels in 8 bit grayscale data, I am not surprised that Kai-Uwe 
can see each level.

Bob
======================================
Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/