Re: 16 bits = 15 bits in Photoshop?
Re: 16 bits = 15 bits in Photoshop?
- Subject: Re: 16 bits = 15 bits in Photoshop?
- From: Marco Ugolini <email@hidden>
- Date: Mon, 18 Apr 2005 01:01:10 -0700
In a message dated Sun, 17 Apr 2005 13:40:38 EDT, Dan Margulis wrote:
> Marco Ugolino writes,
Well, Count Ugolino was actually a character in Canto 33 of Dante's Inferno.
But that was him. I am someone else. MY name is Ugolini. Different guy.
> The Richard Pryor line is my standard response in large group sessions
> when people ask me to show histograms. I've used it for years.
Me too. Funny how lots of people remind me too of Pryor's line.
> Lying eyes cannot be trusted to honestly evaluate neutral colors displayed on
> a monitor. With that single exception, there is no one in this industry who is
> remotely as associated with the concept of using one's lying eyes as I am.
I'll try to translate: you seem to say that it's better to trust a measuring
instrument to calibrate a monitor, but, except for that one instance, for
everything else one ought to trust his own eyes. Am I warm?
You must certainly be aware that the eyes can also deceive you. In color
management (forgive me for mentioning this expression, so distasteful to
you) I don't think we would be going very far by looking at the individual
patches of a printed test chart and guessing their L*a*b coordinates...
> By contrast, those who say there's an advantage rely on histograms, and--as
> in your case--pure faith. If I may summarize the logic of yourself, Bruce,
> and Gariba (who I contacted off line, and found he had never tested anything
> either):
>
> 1. I personally work in 16-bit.
> 2. I am satisfied with the results.
> 3. I do not see any of the defects that I believe would be there
> if I used 8-bit.
> 4. Therefore, 16-bit is clearly better.
Don't take this the wrong way, but I do not enjoy condescension, yours or
anyone else's. And I am certainly not a man of faith, not even the religious
kind (if I may be so bold as to say so publicly). I believe in what I can
see and prove. The caricature of low-caliber logic that you expound is not
even worth addressing among professionals. The level of discussion on this
board ought to be maintained higher than that.
> This approach does *not* employ the lying eyes. It uses the imagination,
> which is much more mendacious. Your, and Bruce's and Gariba's, comparison is
> not between two competing images, which your lying eyes could make for you. It
> is between your own image and an *imaginary* bad picture that your mind has
> concocted for you.
> [...]
> Assuming on faith that 16-bit is better is just as bad as me assuming on
> faith that my color corrections are always best. We both need our lying eyes
> to tell us whether we were right or not, and for them to do that, we have to
> have a real image to compare to, not something our imaginations conjure up.
OK, this is getting silly. I will give you examples, if that is what you
really want, and not just an overdose of hyperventilation.
Let's start with a synthetic image. I will provide detailed steps so that
anyone can do this for themselves. And I do apologize if this is going to be
a bit long: I certainly hope that you will all find it worth your time.
1) In Photoshop, press Command-N to create a new file; select Grayscale
mode, 16 bits, white background, Gray Gamma 2.2, Square Pixel Aspect Ratio,
1500 W x 600 H pixels (5" x 2" at 300ppi). Call the file "Grayscale-16
bits." Click OK.
2) Press D to set the foreground to black and the background to white.
3) Select the Gradient tool, in "Foreground to Background" mode. Deselect
Reverse, Dither and Transparency from the top bar. Create a gradient going
from left to right.
4) Duplicate this image of the gradient (Image -> Duplicate). Call the
duplicate "Grayscale-8 bits." Convert it to 8 bits (Image -> Mode -> 8
Bits/Channel").
5) Stay in the "Grayscale-8 bits" image. Add a new Curves adjustment layer.
Set 3 points in the curve (I am using the 0-255 values, not the 0-100 ones):
- Point 1: Input 17; Output 43
- Point 2: Input 56; Output 118
- Point 3: Input 211; Output 228
Click OK.
6) In the Layers palette, duplicate this Curves adjustment layer twice, so
that there will be a total of 3 identical Curves adjustment layers above the
gradient image.
7) In the Layers palette, click on any one of the 3 Curves adjustment layers
and drag it to the "Grayscale-16 bits" file. Here as well, duplicate it
twice to have a total of 3, same as in the 8-bit file.
Please see the results at my web site:
http://home.mindspring.com/~marcoug
Click on the "Gradient-16 bits.jpg" image, and you will see that the 8-bit
image exhibits clear signs of posterization. These are screen shots from my
monitor, at 100%. The 16-bit file shows no posterization whatsoever (though
some very slight banding is visible, which may be due to the LUT values
generated by my monitor profile).
At the same URL, you can also go ahead and see for yourselves what happens
to another grayscale image, a 16 bit scan I made some time back from a black
& white film negative. The "Mike-16 vs 8 bits.jpg" image shows (at 100% here
as well) the results of an identical set of Curves adjustment layers in both
images. Please take my word that the only difference is that the image on
the left is in 16 bits, and the one on the right in 8 bits, and that
everything else is exactly the same. Since the dithering/noise introduced to
a scan by the film grain of a negative tends to hide posterization and
banding, I applied the despeckle filter once to both images. The true test
of 16 vs 8 bits is in images that are very smooth. And digital images are
smoother than film-based ones, let's remember that. The use of the despeckle
filter takes us one small step towards a smoother image, reducing the
deceptive masking effect of noise.
The last image ("Mike-Blurred-16 vs 8 bits.jpg") shows what happens to each
picture when I apply a Gaussian Blur filter of 10.0 radius. The 8 bit
picture shows its barely hidden ugliness and finally falls apart, but the 16
bit one stays together very nicely, without displaying any signs of banding
whatsoever.
Moral of the story:
If you want to work in 8 bits, be aware that there is only so far you can go
before you had better watch out for posterization and banding. That, of
course, is assuming that you care enough to do that. If circumstances force
you to apply a heavy hand to your image, it would be foolish at least not to
CONSIDER starting your work in 16 bits. Still, if you insist that it makes
no difference worth your time, then this is no longer about reason or logic,
but about religion. And I won't touch that one with a ten-foot pole.
Thank you all for your patience.
--------------
Marco Ugolini
Mill Valley, CA
_______________________________________________
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