Apache-Talk @lexa.ru 

Inet-Admins @info.east.ru 

Filmscanners @halftone.co.uk 

Security-alerts @yandex-team.ru 

nginx-ru @sysoev.ru 




      :: Filmscanners
Filmscanners mailing list archive (filmscanners@halftone.co.uk)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[filmscanners] Re: 8bits vs. 16bits/channel:cantheeyeseethedifference

on 3/17/03 7:42 PM, Frank Paris at frankparis@comcast.net wrote:

>> -----Original Message-----
>> From: filmscanners_owner@halftone.co.uk
>> [mailto:filmscanners_owner@halftone.co.uk] On Behalf Of
>> Austin Franklin
>> Sent: Monday, March 17, 2003 6:52 PM
>> To: frankparis@comcast.net
>> Subject: [filmscanners] RE: 8bits vs. 16bits/channel: can
>> theeyeseethedifference
>> Hi Frank,
>>> That's what PS
>>> actually does? Of course that IS dithering, not aliasing.
>> But it doesn't do that.  It simply chops off the lower 8
>> bits.  That is not dithering or aliasing.  Converting 0x1234
>> to 8 bits is simply 0x12.
> Hmmm. We seem to have a disagreement here. I wonder who is right?
>> Now, if you are processing some data, and have to "split the
>> difference", happening to arrive at 127.5, that is
>> quantization error, not dithering.
> True, if that's what PS does. The other guy seems to disagree with you,
> though.
> Frank Paris
> frankparis@comcast.net

Wow you guys really got going!

First, the spacial resolution issue.  The only limitation is that you
can't see a .5 gray tone by only looking at one pixel.  My comment
of for this is just that can't resolve fractional grays a full
resolution.  There's no resolution difference of gray differences
above 1 LSB.


Demonstrating how PS works in converting 16-bit to 8-bit is easy.

First create an 8 bit gray, 3x1 inches, 360ppi or your favorite.
Let's make 3 inch squares of gray, values 0, 1, 2.  (To select colors fill
in the RGB value all the same.

It all looks black but there's 3 different patches.
Do an Auto Levels to see the 3 patches, then Undo.

Convert to 16-bit,  still 3 patches, 3 gray.  (do Auto Levels)

Now lets make the middle value an in between.   Do Adjust Levels
and slide the Right Output marker from 255 down to 127.  (its not exactly
.5 but its close.

So now you've got 3 patches, 3 grays of value 0, .5, 1.  Its hard to
see the .5, the eye dropper will just look at the top byte.  But feel
free to do an Auto Levels and Undo to see the 3.

So what happens now when we convert to 8-bit?  The 0, 1 are pretty
straightforward, but what about the .5?  Well do it and see.

Well they all look black.  Try this make a small marquis on each square.
Do a Histogram, look at the Mean.  Very close to 0, .5, 1

Now do an Auto Levels.  Three very different patches.  Remember that
you are looking at these with more that 100 times normal contrast.
Anyway, it sure looks like a random dither pattern.  You obviously
can't tell what PS's algorithm for dithering, but the fact that the Mean
is so close to .5 means its pretty good.


So does any of this matter?  Alex's original patches say maybe, sometimes.
What about real pictures? Well say you had a 16-bit scan of a very gradual
sky gradient that went from 120 to 125 over a 5 inch span.  Without this
PS dithering you could easily end up with 5 bands across that sky as it went
from 120, to 121, ... 125 when you converted to 8-bit for printing.
With the PS dithering you will have as smooth a gradient as your
printer can print.


Roy Harrington
Black & White Photography Gallery

Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 


Copyright © Lexa Software, 1996-2009.