Re: Does it make a difference
Re: Does it make a difference
- Subject: Re: Does it make a difference
- From: Graeme Gill <email@hidden>
- Date: Wed, 26 Oct 2005 14:02:08 +1000
Mark Rice wrote:
Yes, it makes a big difference. Because gamma is a logarithmic?? (I am not
sure if it is logarithmic or some other function), a number of values tend
to bunch at the top or bottom of the scale. Since I referred to the LVT film
recorder, I will use transparencies as the medium for an example. With a
normal gamma function, and a 26 step gray scale that the LVT used for
calibration (this is burned in my head, as we had four LVT's and a Lightjet
that I calibrated 5 times per day!!) the darkest 5 steps of the gray scale
appeared as solid black, even though a densitometer could detect subtle
differences between them. Using the "equal brightness scale", or L*, the
step above solid black appeared to be visibly lighter than the blackest. The
3rd step appeared visibly lighter than that, and so on. At the lightest part
of the scale, the 25th step appeared visibly darker that the 26th step.
That is the point of using the L* scale - to imitate the gradation steps in
a real image, so that one can visibly detect what is happening.
L* is exactly the same kind of function as used for the "gamma" curve of
sRGB, and differs only slightly from a straight power curve.
See <http://www.srgb.com/srgbgammacalculation.pdf>.
Choosing a "normal" gamma of 2.45 will give a similar
curve to L*. L* is a curve that approximates a perceptually
uniform brightness scale. There have been several experiments
to derive such curves, and each experiment seems to arrive
at broadly similar curve that is different in detail.
Note that the ideal curve depends on the viewing conditions -
perhaps in your setup, L* is closer to perceptually uniform
for your particular viewing conditions than typical gamma values
you've tried.
sRGB: (Close to a straight gamma of 2.2)
if (x <= 0.03928)
y = y/12.92
else
y = ((0.055 + x)/1.055) ^ 2.4
L*: (Close to a straight gamma of 2.45)
if (x <= 0.08)
y = y/9.03296
else
y = ((0.16 + x)/1.16) ^ 3.0
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden