RE: 8 bit vs. more bits
RE: 8 bit vs. more bits
- Subject: RE: 8 bit vs. more bits
- From: "Mark Rice" <email@hidden>
- Date: Sun, 17 Apr 2005 23:04:04 -0400
- Organization: Zero One
I understand your argument, Robert, but most 'tests' for bit depth involve a
ramp of colors, not a square inch where every pixel is a different color - I
can't imagine what that test would be used for.
If you have a monitor with 1024 lines, and you attempt to display a subtle
green to hue shift green, you may or may not see quantization on the screen
- largely depending on the dithering algorithm.
I have been working in digital imaging since 1985, first as a color lab
owner, and then developing the first digital camera for Sinar in 1987. I
worked with the people at LVT on their prototype film recorder, developing
look-up tables for them.
Output to Ektachrome as a continuous tone device is probably the most
difficult test of any output device. Even though the device operated
internally in 12 bit space (4096 levels per color), this was only done for
log to linear linearization purposes. It was all converted to 8 bit on
output, I suspect because of limited computer power at the time. The worst
case was a very soft gray shadow against a white background. Quantization
was totally invisible on a monitor, but strikingly visibly on the Ektachrome
output (because of the high dynamic range of Ektachrome) The Lightjet 2080
also had this problem. Paper exposing devices, such as the Lambda and the
Lightjet show much less of the problem because the dynamic range of
reflective output is so much lower.
We had to introduce noise to make the Ektachrome shadows come out
"smoothly", but the problem was never visible on reflective output.
-----Original Message-----
From: Robert L Krawitz [mailto:email@hidden]
Sent: Sunday, April 17, 2005 9:14 PM
To: email@hidden
Cc: email@hidden
Subject: Re: 8 bit vs. more bits
From: "Mark Rice" <email@hidden>
Date: Sun, 17 Apr 2005 20:46:33 -0400
There certainly have been a lot of specious and non-scientific
arguments on this subject lately.
I would like to establish several things:
1. Viewing a monitor to detect differences in 8 bit vs. more bits
is a thankless task. Have you ever asked yourself, how can I get
the 16.7 million colors promised by 8 bit (or 24 bit for 3 colors)
from only 1024 lines of video resolution? YOU CAN'T!!!. We simulate
it with dithering.
You are correct in that a computer generated color ramp, especially a subtle
one, is the most difficult test. Photos usually have noise in them, which
effectively, is dithering.
Here are 4 images in tiff format, zip compressed, 1280x1024:
ftp://ftp.zero1inc.com/tests/greenrampdithered.tif,
ftp://ftp.zero1inc.com/tests/GreenrampUndithered.tif ,
ftp://ftp.zero1inc.com/tests/Green_to_purple_Dithered.tif ,
ftp://ftp.zero1inc.com/tests/Green_to_purple_undithered.tif
The green ramps go from 100R 200G 100B to 100R 200G 90B. I have tried it
both dithered and undithered in Photoshop. Both look very smooth on my
screen, and measuring values on the screen shows 1 bit shifts about every
inch.
A worse test case is this - the green to purple ramps go from 100R 200G 100B
to 100R 0G 100B, both dithered and undithered. These have apparent color
problems on the screen, but the problems look like somewhat random stripes -
I suspect they are related to monitor imperfections, as they are not
measurable with the eyedropper.
As for the acid test of 0-255 Green, go to
ftp://ftp.zero1inc.com/tests/GreentoBlackundithered.tif. When viewed on the
screen at 100%, it appears to have quantizing, or banding. However, if you
zoom in to about 400% the banding disappears. I interpret this as a monitor
problem. True quantizing should show up more clearly when zoomed in.
So I still maintain that a monitor tells us nothing about color quality.
Mark Rice
email@hidden
www.zero1inc.com
The acid test here is displaying a green ramp from 0-255. Can you see the
transitions? When I've tried it, the answer is yes, and to me at least that
means that the resolution, even at the output, is too coarse.
2. Can a color printing device display 16.7 million colors? Not
likely. The Durst Lambda, at 400 dpi, would need to make a print
3,479 FEET long in order to use all 16.7 million colors! So
dithering is also used in printing devices.
Again, the real question is whether you can see a stairstep effect.
Not to mention that that calculation is incorrect -- at 20,000 points/square
inch (the Lambda is continuous tone, right?) it would take about 850 square
inches to print 16.7 million pixels, each of a different color. But that's
a red herring -- the issue is whether you can see transitions between
particular colors that you care about, not how many different colors can be
printed on a print of a particular size.
Both of the above statements are true, and I suspect that is where
Dan M's reasoning coming from. If one takes a perfectly exposed
digital photos, or makes a flawless scan utilizing the entire 256
bit range, than no higher bit depths are necessary.
I doubt very much that that's what Dan means. Photos aren't usually the
acid test for high bit depths, although I've taken a few photographs where
you can see the stairsteps in the sky at 8 bits of precision. A better test
would be a logo with a gradient from, say, a medium green to a light green,
where the range of colors you care about is very small but it's essential to
have transitions that are perceptably perfectly smooth. The example of the
sky, of course, is a gradient.
--
Robert Krawitz <email@hidden>
Tall Clubs International -- http://www.tall.org/ or 1-888-IM-TALL-2 Member
of the League for Programming Freedom -- mail email@hidden
Project lead for Gimp Print -- http://gimp-print.sourceforge.net
"Linux doesn't dictate how I work, I dictate how Linux works."
--Eric Crampton
_______________________________________________
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