Re: Usable Tones in Different Gammas
Re: Usable Tones in Different Gammas
- Subject: Re: Usable Tones in Different Gammas
- From: Don Hutcheson <email@hidden>
- Date: Sun, 10 Dec 2000 13:01:58 -0500
DELTA-E CAUTION:
It is dangerous and incorrect to use Delta-E 1.0 as a definition of the
finest necessary tonal resolution. Delta-E 1.0 roughly defines the minimum
acceptable error9 between two samples separated by a gap of white, black or
any other color, but it does NOT define the minimum detectable difference of
the human eye. Far from it. If you display or print two TOUCHING color or
grey patches with an L* difference equivalent to 1.0 and no gap or border
between them, the difference is dramatically evident.
TONE LEVELS AT DIFFERENT GAMMAS:
Regarding the number of tones available at different display gammas, some of
the opinions recently posted here show novel theories but unnecessary alarm.
Most displays will show between 200 and 256 discrete levels at normal gamma
settings of 1.8 to 2.2. The only real danger is quantization loss in the
video LUTs.
In theory the closer you set your video LUTs to the display's natural gamma
(usually about 2.5 for a CRT) the fewer levels will be lost by the LUTs. In
practice no gamma will completely eliminate tonal stepping for the reasons
outlined under 'Eight bits is not enough', but choosing 1.8 or 2.2 seems to
produce the smoothest ramps on a Mac screen. The good news is that profiling
your monitor at a different gamma from the image file should have little or
no effect on display smoothness if the CMM links the profiles in 16 bits
(which most do.)
BEST GAMMA:
If you're serious about quality, especially in transparency scans or live
captures, the optimum gamma for scanner, camera, display and retouching
space is in the 2.2 to 2.8 range (nominally 2.5). Why? Because the gamma of
the human eye is roughly 2.4 - 3.0 (depending on brightness) hence a similar
gamma will distribute the limited 256 tone levels in an 8 bit file most
uniformly throughout the range of VISIBLE TONAL STEPS in the original. This
means that the file will capture what we see in the original's highlights,
mid-tones and shadows with equal fairness, giving equal latitude for
post-processes such as correcting under-exposure or over-exposure.
More importantly, this gamma range produces a file that is equally adaptable
to any output device, not just a desktop printer. The fact that most
printing devices have a theoretical gamma of about 1.8 is relatively
unimportant, as the transformation from 2.x to 1.8 mostly compresses shadow
tones, which we can barely see. Going from gamma 2.2 to 1.8 or vice-versa
loses only about 20 levels out of 256, but if you set Photoshop 6's 'Convert
to Profile' dialog to 'Use Dither' you'll never see the loss, as the CMM's
16 bit precision is effectively preserved in the 8 bit file (bravo Adobe).
EIGHT BITS IS NOT ENOUGH:
For the record, perfectionists in the pre-press industry have known for
years that 256 levels are not enough to simulate true continuous tonality,
even on an offset press. The proof of that is a simple Photoshop vignette.
No one would expect one made WITHOUT dithering to print smoothly.
Therefore the real solution to all these problems - gamma, banding,
contouring, quantization - is to use 16-bit per channel devices, files and
software. If you can't get true 16 bit images, even 10 bits per channel is
still four times better than eight.
*************************************
Don Hutcheson
Hutcheson Consulting
(Color Management Solutions)
11 Turnburry Rd
Washington, NJ 07882
Phone: (908) 689 7403
Fax: (908) 689 5305
email@hidden
*************************************