Digital BW, The Print

Fellows,

if I open an image file in PS 5.5 or 6.01, which I saved in 8 bit 
grayscale modus before and proced only a little curve tweaking, 
for instance elevating the 3/4 tones by 3% and lowering the 1/4 
tones by 3% in order to get a little more contrast 
(image/adjust/curves) I always heavily destroy my histogram. 
That means multiple levels count zero while others count 
doubled values compared to their neighbours.
That never happens if I do the same adjustments in a 16 bit 
greyscale file.
Switching back from 8 bit to 16 bit doesn´t help.
Is that normal, or did I miss something.
Thanks for your advise.
Bernd

PS 
I use to scan in 16 bit greyscale or 48 bit RGB, but I save my 
"ready to print files" in 8 bit, so if I want to change them later I 
must handle the 8 bit file.

I don't think that's normal, but it's a good case for always scanning in Hi
bit and doing as much as you can there.

I think but am not certain that the next version of photoshop will allow
most everything to be done in Hi bit. Lets hope so.

Jerry



ruhrfoto@... wrote:

> Fellows,
>
> if I open an image file in PS 5.5 or 6.01, which I saved in 8 bit
> grayscale modus before and proced only a little curve tweaking,
> for instance elevating the 3/4 tones by 3% and lowering the 1/4
> tones by 3% in order to get a little more contrast
> (image/adjust/curves) I always heavily destroy my histogram.
> That means multiple levels count zero while others count
> doubled values compared to their neighbours.
> That never happens if I do the same adjustments in a 16 bit
> greyscale file.
> Switching back from 8 bit to 16 bit doesn\ufffdt help.
> Is that normal, or did I miss something.
> Thanks for your advise.
> Bernd
>
> PS
> I use to scan in 16 bit greyscale or 48 bit RGB, but I save my
> "ready to print files" in 8 bit, so if I want to change them later I
> must handle the 8 bit file.
>
>
> If you do not wish to belong to Digital B&W, The Print, you may
> unsubscribe by sending an email to:
> DigitalBlackandWhiteThePrint-unsubscribe@yahoogroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

The likelihood that you'd see the results of the combed histogram 
in your prints from such a minor change I think would be unlikely, 
even if you are seeing it in the histogram.

Converting from 8 to 16 bit won't help if you are making a single 
change, but if you are making multiple changes, it can prevent 
some of the round-off errors.

Usually, if I want to make modifications to a pre-existing scan, and 
they are less than trivial, I'll rescan (assuming the negative is still in 
good shape).

Tim
 
> if I open an image file in PS 5.5 or 6.01, which I saved in 8 bit 
> grayscale modus before and proced only a little curve tweaking, 
> for instance elevating the 3/4 tones by 3% and lowering the 1/4 
> tones by 3% in order to get a little more contrast 
> (image/adjust/curves) I always heavily destroy my histogram. 
> That means multiple levels count zero while others count 
> doubled values compared to their neighbours.
> That never happens if I do the same adjustments in a 16 bit 
> greyscale file.
> Switching back from 8 bit to 16 bit doesn´t help.
> Is that normal, or did I miss something.
> Thanks for your advise.


--
Tim Spragens
http://www.borderless-photos.com

Bernd,

To give you some idea of how bad a histogram can look and still 
provide a good print I have uploaded an image of three histograms to 
the "Files" section of the Group Homepage in the "Message Related 
Files" folder entitled "Barracks Photo Histograms"

These are from two photos in the "Photo" section of the homepage that 
I uploaded to illustrate message #17.

The first histogram is of the 8-bit grayscale tiff file (Barracks 1) 
made on a drum scanner by a service bureau. Not the greatest, as you 
can see, and leaving me with a lot of adjustment to do. (Yes I did 
ask them to redo the scan. This was their third attempt and the best 
of the three.)

The middle histogram is of the entire flattened file (Barracks 2) 
after all adjustments were made to achieve the final image. There are 
lots of gaps and spikes.

The last histogram is of just the sky portion of the image after 
final adjustment. Not much left.

The histograms say this image should not print but it does. The image 
is at the limit. If I try to push it any farther the sky begins to 
posterize. If I was in 16-bit I could push it farther.

My point is that the histogram is a guide to check during image 
adjustment but all that really matters is the print. So make the 
adjustments you need to achieve what you want and see if it will 
print. Don't give up on an adjustment just because the histogram 
looks poor or even terrible.

Martin



--- In DigitalBlackandWhiteThePrint@y..., ruhrfoto@y... wrote:
> Fellows,
> 
> if I open an image file in PS 5.5 or 6.01, which I saved in 8 bit 
> grayscale modus before and proced only a little curve tweaking, 
> for instance elevating the 3/4 tones by 3% and lowering the 1/4 
> tones by 3% in order to get a little more contrast 
> (image/adjust/curves) I always heavily destroy my histogram. 
> That means multiple levels count zero while others count 
> doubled values compared to their neighbours.
> That never happens if I do the same adjustments in a 16 bit 
> greyscale file.
> Switching back from 8 bit to 16 bit doesn´t help.
> Is that normal, or did I miss something.
> Thanks for your advise.
> Bernd
> 
> PS 
> I use to scan in 16 bit greyscale or 48 bit RGB, but I save my 
> "ready to print files" in 8 bit, so if I want to change them later 
I 
> must handle the 8 bit file.

on 8/14/01 12:32 PM, mwesley250@... wrote:

> The first histogram is of the 8-bit grayscale tiff file (Barracks 1)
> made on a drum scanner by a service bureau. Not the greatest, as you
> can see, and leaving me with a lot of adjustment to do. (Yes I did
> ask them to redo the scan. This was their third attempt and the best
> of the three.)

When I was looking at your print of this image last evening I wondered if
it's possible that whatever occurrence you experience with the sky area of
this image might not be attributable to noise, rather than a busted
histogram. I assume when you fatigue the the shy it's is through darkening
it, and/or adding contrast. This always seems to expose defects in film and
scans for me. If you were to apply an equal move in the opposite direction
instead, i.e., lightening it how would it look? Heck, lets really test it.
If you put an equal move of lightening/flattening the area AFTER your
darkening/contrast move how does it look? Such a yo-yo move should really
kill the histogram - how much worse for wear does the sky look after the
fore and aft, compared to no move at all?

Isn't testing fun??????

Todd

Increasing contrast (which I find is what I'm doing most of the time) is the
thing that kills the histogram, because you are stretching a smaller range
of values into a larger range. Decreasing contrast shouldn't really cause
any problems. Here's an easy experiment that illustrates it: open a new
blank image in Photoshop and use the gradient tool to create a gradient from
pure black to pure white across the image. The histogram should be
relatively smooth. Now use the levels tool to reduce the contrast so the
gradient goes from, say, 60% gray to 40% gray. The histogram will be
squashed into the middle, but should still be as smooth as before. Now use
the levels again to reverse it back to a black to white gradient. The
histogram will have huge gaps, and there should be obvious banding in the
image.

Of course, as others have said, all this histogram worship has been a bit
overblown. It all depends on the image. Not all images will have smooth
histograms -- an image of black and white stripes for instance -- and the
only thing that really matters is how the print looks. That being said,
getting a handle on the principles at work has finally helped me understand
why I was getting posterization in some of my images, and switching to a
16bit workflow has helped me solve those problems.

-Jason

-----Original Message-----
From: Todd Flashner [mailto:tflash@...]
Sent: Tuesday, August 14, 2001 1:55 PM
To: DigitalBlackandWhiteThePrint@yahoogroups.com
Subject: Re: [Digital BW] Re: Image Histograms Destroyed

I assume when you fatigue the the shy it's is through darkening
it, and/or adding contrast. This always seems to expose defects in film and
scans for me. If you were to apply an equal move in the opposite direction
instead, i.e., lightening it how would it look? Heck, lets really test
it....

Jason,

I don't disagree with anything you are saying, but I mean that even in in an
analog scenario, contrast and density reveals inherent flaws.

So, assuming my experiment is a flawed one, how else might one test to
determine if what they don't like in their image is a function of a broken
histogram, vs, bad data, which is merely being exposed as such?

Todd

> Increasing contrast (which I find is what I'm doing most of the time) is the
> thing that kills the histogram, because you are stretching a smaller range
> of values into a larger range. Decreasing contrast shouldn't really cause
> any problems. Here's an easy experiment that illustrates it: open a new
> blank image in Photoshop and use the gradient tool to create a gradient from
> pure black to pure white across the image. The histogram should be
> relatively smooth. Now use the levels tool to reduce the contrast so the
> gradient goes from, say, 60% gray to 40% gray. The histogram will be
> squashed into the middle, but should still be as smooth as before. Now use
> the levels again to reverse it back to a black to white gradient. The
> histogram will have huge gaps, and there should be obvious banding in the
> image.
> 
> Of course, as others have said, all this histogram worship has been a bit
> overblown. It all depends on the image. Not all images will have smooth
> histograms -- an image of black and white stripes for instance -- and the
> only thing that really matters is how the print looks. That being said,
> getting a handle on the principles at work has finally helped me understand
> why I was getting posterization in some of my images, and switching to a
> 16bit workflow has helped me solve those problems.
> 
> -Jason
> 
> -----Original Message-----
> From: Todd Flashner [mailto:tflash@...]
> Sent: Tuesday, August 14, 2001 1:55 PM
> To: DigitalBlackandWhiteThePrint@yahoogroups.com
> Subject: Re: [Digital BW] Re: Image Histograms Destroyed
> 
> I assume when you fatigue the the shy it's is through darkening
> it, and/or adding contrast. This always seems to expose defects in film and
> scans for me. If you were to apply an equal move in the opposite direction
> instead, i.e., lightening it how would it look? Heck, lets really test
> it....

on 8/14/01 12:32 PM, mwesley250@... wrote:

> To give you some idea of how bad a histogram can look and still
> provide a good print I have uploaded an image of three histograms to
> the "Files" section of the Group Homepage in the "Message Related
> Files" folder entitled "Barracks Photo Histograms"

Thanks Martin, very informative!

Todd

Oh, I wasn't trying to dispute your methodology regarding defects, just to
illustrate the more basic concept of histogram movement. Obviously,
increasing contrast generally also increases the contrast of the noise,
which is going to accentuate any defects. That's going to be a limiting
factor regardless of what bit-depth you work in, because the noise is part
of your data. All the more reason to concentrate on good exposures and good
scans, to get the data right in the first place.

I'm not sure how you could isolate the effects of noise from those of
histogram movement, or what you would do about it if you could.

-Jason

-----Original Message-----
From: Todd Flashner [mailto:tflash@...]
Sent: Tuesday, August 14, 2001 2:44 PM
To: DigitalBlackandWhiteThePrint@yahoogroups.com
Subject: Re: [Digital BW] Re: Image Histograms Destroyed


Jason,

I don't disagree with anything you are saying, but I mean that even in in an
analog scenario, contrast and density reveals inherent flaws.

So, assuming my experiment is a flawed one, how else might one test to
determine if what they don't like in their image is a function of a broken
histogram, vs, bad data, which is merely being exposed as such?

Todd

> Increasing contrast (which I find is what I'm doing most of the time) is
the
> thing that kills the histogram, because you are stretching a smaller range
> of values into a larger range. Decreasing contrast shouldn't really cause
> any problems. Here's an easy experiment that illustrates it: open a new
> blank image in Photoshop and use the gradient tool to create a gradient
from
> pure black to pure white across the image. The histogram should be
> relatively smooth. Now use the levels tool to reduce the contrast so the
> gradient goes from, say, 60% gray to 40% gray. The histogram will be
> squashed into the middle, but should still be as smooth as before. Now use
> the levels again to reverse it back to a black to white gradient. The
> histogram will have huge gaps, and there should be obvious banding in the
> image.
>
> Of course, as others have said, all this histogram worship has been a bit
> overblown. It all depends on the image. Not all images will have smooth
> histograms -- an image of black and white stripes for instance -- and the
> only thing that really matters is how the print looks. That being said,
> getting a handle on the principles at work has finally helped me
understand
> why I was getting posterization in some of my images, and switching to a
> 16bit workflow has helped me solve those problems.
>
> -Jason
>
> -----Original Message-----
> From: Todd Flashner [mailto:tflash@...]
> Sent: Tuesday, August 14, 2001 1:55 PM
> To: DigitalBlackandWhiteThePrint@yahoogroups.com
> Subject: Re: [Digital BW] Re: Image Histograms Destroyed
>
> I assume when you fatigue the the shy it's is through darkening
> it, and/or adding contrast. This always seems to expose defects in film
and
> scans for me. If you were to apply an equal move in the opposite direction
> instead, i.e., lightening it how would it look? Heck, lets really test
> it....


If you do not wish to belong to Digital B&W, The Print, you may
unsubscribe by sending an email to:
DigitalBlackandWhiteThePrint-unsubscribe@yahoogroups.com



Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

--- In DigitalBlackandWhiteThePrint@y..., Todd Flashner <tflash@e...> 
(snip)
> 
> When I was looking at your print of this image last evening I 
wondered if
> it's possible that whatever occurrence you experience with the sky 
area of
> this image might not be attributable to noise, rather than a busted
> histogram. I assume when you fatigue the the shy it's is through 
darkening
> it, and/or adding contrast. This always seems to expose defects in 
film and
> scans for me. If you were to apply an equal move in the opposite 
direction
> instead, i.e., lightening it how would it look? Heck, lets really 
test it.
> If you put an equal move of lightening/flattening the area AFTER 
your
> darkening/contrast move how does it look? Such a yo-yo move should 
really
> kill the histogram - how much worse for wear does the sky look 
after the
> fore and aft, compared to no move at all?

You actually have a print of a picture I took immediately before the 
one the histograms are from but a very similar situation, although I 
believe the scan of the print you have was PhotoCD and easier to work 
with.

There is a limit of how much adjustment (contrast, brightness, 
levels) you can do before the image degrades. This seems to depend on 
bit depth, the quality of the initial scan (which may be limited by 
the quality of the neg) and the nature of the image. Given the same 
scan quality, you are able to tweak a textured image farther than one 
with large smooth gradients. Not because you aren't getting the same 
amount of posterizing, it's just is harder to see in the texture. 
Also, as Paul pointed out earlier, you also may not see this on 
screen but only when it is printed out. But in most cases, the point 
in 8-bit gray scale where objectionable print quality occurs is way 
past the point where the histogram "dies."

> 
> Isn't testing fun??????

No.

Martin