Can't figure out 16 bit per channel RGB Channel in PhotoShop - NEWBIE
Sarah
My question is fairly basic (read dumb). A 16 bit RGB channel results
in a total of 48 pixels which means you can potentially have 2.81
trillion colours.
8 bit gives you 16.7 million and humans can only see about 10 million,
so is there any reason to use 16 bit per channel RGB?
I'm probably missing something blantently obvious but any advice would
be greatly appreciated, i.e. is it for CMYK printing purposes or am I
completely missing the point and there is a different use for it
altogether.
Any help would be greatly appreciated.
Regards,
Sarah.
Bart van der Wolf
"Sarah" <shami...@hotmail.co.uk> wrote in message
news:1129812223.8...@o13g2000cwo.googlegroups.com...
> First of all, thanks to all who replied to my previous question.
>
> My question is fairly basic (read dumb). A 16 bit RGB channel
> results
> in a total of 48 pixels which means you can potentially have 2.81
> trillion colours.
>
> 8 bit gives you 16.7 million and humans can only see about 10
> million,
> so is there any reason to use 16 bit per channel RGB?
8-bit/channel will give 256 colors for each channel, making
256*256*256 colors in total. That is probably fine for a final result.
However, if we want to do several cumulative image adjustments on low
noise images then that can result in posterized gradients. To avoid
that it can help to work in 16-b/ch, where accumulating rounding
errors will have less effect on the final 8-bit/channel that's going
to be used in final output.
It can also help in avoiding posterization when converting between
color space profiles (although the type of profile plays a role as
well).
Bart
Bill Hilton
>
>8 bit gives you 16.7 million and humans can only see about 10
>million, so is there any reason to use 16 bit per channel RGB?
Yes, if you are heavily editing the files using 16 bit helps avoid
posterization and other problems ... here's a very good write-up on the
advantages and disadvantages, from Bruce Fraser ...
http://www.creativepro.com/story/feature/7627.html ... if you find you
don't need to make many edits then it's less useful.
Bill
Mike Russell
news:1129812223.8...@o13g2000cwo.googlegroups.com...
Hi Sarah,
Reasonable people differ on this topic, and it's worth discussing again
every once in a while.
In normal gamma spaces, such as are used on Macintosh or Windows, it is
difficult or impossible to come up with an actual image that shows
differences between being edited in 8 bit versus 16 bit. In fact, no such
image has ever been presented to me in spite of several open invitations,
and reward of $100. For this reason alone, I find the argument that 16 bit
per channel data is an obvious requirement for excellent work is not
justified.
I have serious issues with the Bruce Fraser example that Bill refers to. It
is based on an artificial image, and in the end it is a tautological
argument, demonstrating only that in performing the conversion from 16 bit
to 8 bit, Photoshop introduces 1/2 a bit of noise. I have also see people
use a histogram to show that gaps appear in 8 bit edit that do not show in a
16 bit edit - again this is unsatisfactory because it does not use an actual
image.
Bart van der Wolf has kindly presented an example, however it is, (please
correct me if I'm wrong) a gamma 1.0 image. If you work with such images -
and some people do - hibit images are justified. Years ago I worked at a
place that developed hardware used to make digital movies, and 10 bits per
channel were necessary to avoid banding in the shadows, again at 1.0 gamma.
That said, many people, including some excellent professional photographers,
feel that even an imperceptible difference in quality may eventually be
significant to their images - if not now, then perhaps in the future. As a
tool provider, this is enough reason for me to fully support both 8 and 16
bit manipulations in Curvemeister.
In the end, my feeling is the best way to decide is to use some of your own
images - perform extreme edits on them and see if you can see banding, or
other artifacts. If you are not experimentally inclined, and simply want to
make every effort to assure quality in your images, you may simply decide to
go with 16 bits, even if you cannot prove to yourself that there is an
incremental quality benefit.
--
Mike Russell
www.curvemeister.com
Jim
"Mike Russell" <RE-MO...@Curvemeister.comRE-MOVE> wrote in message
news:r_T5f.2986$dO2....@newssvr29.news.prodigy.net...
> "Sarah" <shami...@hotmail.co.uk> wrote in message
> news:1129812223.8...@o13g2000cwo.googlegroups.com...
>> First of all, thanks to all who replied to my previous question.
>>
>> My question is fairly basic (read dumb). A 16 bit RGB channel results
>> in a total of 48 pixels which means you can potentially have 2.81
>> trillion colours.
>>
>> 8 bit gives you 16.7 million and humans can only see about 10 million,
>> so is there any reason to use 16 bit per channel RGB?
>>
>> I'm probably missing something blantently obvious but any advice would
>> be greatly appreciated, i.e. is it for CMYK printing purposes or am I
>> completely missing the point and there is a different use for it
>> altogether.
>>
>> Any help would be greatly appreciated.
>
> Hi Sarah,
>
> Reasonable people differ on this topic, and it's worth discussing again
> every once in a while.
>
> In normal gamma spaces, such as are used on Macintosh or Windows, it is
> difficult or impossible to come up with an actual image that shows
> differences between being edited in 8 bit versus 16 bit. In fact, no such
> image has ever been presented to me in spite of several open invitations,
> and reward of $100. For this reason alone, I find the argument that 16
> bit per channel data is an obvious requirement for excellent work is not
> justified.
>
> I have serious issues with the Bruce Fraser example that Bill refers to.
> It is based on an artificial image, and in the end it is a tautological
> argument, demonstrating only that in performing the conversion from 16 bit
> to 8 bit, Photoshop introduces 1/2 a bit of noise. I have also see
> people use a histogram to show that gaps appear in 8 bit edit that do not
> show in a 16 bit edit - again this is unsatisfactory because it does not
> use an actual image.
any because I don't want to.) However, as near as I can tell, the effect is
merely cosmetic, because in the end the image gets converted to 8 bits per
channel before printing.
Jim
Mike Russell
[re banding or other differences in 8 vs 16 bit images]
> Well, I have seen this in some real life images of mine. (No, I won't
> post any because I don't want to.)
Well, if you ever change your mind, I'm all ears - or eyes - as the case may
be :-)
> However, as near as I can tell, the effect is merely cosmetic, because in
> the end the image gets converted to 8 bits per channel before printing.
That's a good point. In fact, most people print at somewhat less than 7.5
bits per channel.
--
Mike Russell
www.curvemeister.com
tacit
"Mike Russell" <RE-MO...@Curvemeister.comRE-MOVE> wrote:
> In normal gamma spaces, such as are used on Macintosh or Windows, it is
> difficult or impossible to come up with an actual image that shows
> differences between being edited in 8 bit versus 16 bit. In fact, no such
> image has ever been presented to me in spite of several open invitations,
> and reward of $100. For this reason alone, I find the argument that 16 bit
> per channel data is an obvious requirement for excellent work is not
> justified.
Hmm. I didn't know you'd issued such a challenge. I have such an
image--a digital camera photograph taken without a flash under dim
incandescent lighting. As you might imagine, the image turned out flat,
dark, and very, very yellow. Color correction and tonal correction
produced an acceptable image, but introduced noticeable banding in the
shadow behind the subject.
The shadow was not a critical part of the image, so this banding was not
objectionable, but I suspect that in a 16-bit image it would not have
occurred.
I often encounter digital images that are so underexposed, for whatever
reason, that all of the image information falls in the lower quarter of
the histogram. Creating a normal tonal range by using the Levels command
to stretch the histogram often introduces banding in any smooth
gradation in the image; the same is not true of 16-bit images.
--
Art, photography, shareware, polyamory, literature, kink:
all at http://www.xeromag.com/franklin.html
toby
> "Sarah" <shami...@hotmail.co.uk> wrote in message
> news:1129812223.8...@o13g2000cwo.googlegroups.com...
> > First of all, thanks to all who replied to my previous question.
> >
> > My question is fairly basic (read dumb). A 16 bit RGB channel results
> > in a total of 48 pixels which means you can potentially have 2.81
>
> Hi Sarah,
>
> Reasonable people differ on this topic, and it's worth discussing again
> every once in a while.
>
> In normal gamma spaces, such as are used on Macintosh or Windows, it is
> difficult or impossible to come up with an actual image that shows
> differences between being edited in 8 bit versus 16 bit. In fact, no such
> image has ever been presented to me in spite of several open invitations,
> and reward of $100. For this reason alone, I find the argument that 16 bit
> per channel data is an obvious requirement for excellent work is not
> justified.
Scepticism is good, but examples are not too hard to find; I scanned
and adjusted a number of B&W negatives a while back that would have had
no hope as 8-bit scans. Only the extra depth allowed me to set white
and black points and have a decent quality result (the contrast change
was large). Admittedly this is a rather extreme situation, not an
ordinary decent quality image, but any of those original scans would
fit your criterion.
>
> I have serious issues with the Bruce Fraser example that Bill refers to. It
> is based on an artificial image, and in the end it is a tautological
> argument, demonstrating only that in performing the conversion from 16 bit
> to 8 bit, Photoshop introduces 1/2 a bit of noise.
Which is actually a standard image processing trick, not a bug. Simple
quantisation would produce considerably worse banding in some cases
(imagine a subtle gradient). Adding one (8-bit) unit of noise allows
the appearance of a smooth gradient between single 8-bit levels. An
error diffusion dither would be even smoother, of course.
> ...
> --
> Mike Russell
> www.curvemeister.com
Mike Russell
>> In normal gamma spaces, such as are used on Macintosh or Windows, it is
>> difficult or impossible to come up with an actual image that shows
>> differences between being edited in 8 bit versus 16 bit. In fact, no
>> such
>> image has ever been presented to me in spite of several open invitations,
>> and reward of $100. For this reason alone, I find the argument that 16
>> bit
>> per channel data is an obvious requirement for excellent work is not
>> justified.
>
> Hmm. I didn't know you'd issued such a challenge. I have such an
> image--a digital camera photograph taken without a flash under dim
> incandescent lighting. As you might imagine, the image turned out flat,
> dark, and very, very yellow. Color correction and tonal correction
> produced an acceptable image, but introduced noticeable banding in the
> shadow behind the subject.
Hi Tacit,
At this point, I'm not presenting this as a challenge, but I'd definitely be
interested in seeing any image that you think might be best corrected in 16
bits. I suspect that certain underexposed images would be good candidates
for this.
> The shadow was not a critical part of the image, so this banding was not
> objectionable, but I suspect that in a 16-bit image it would not have
> occurred.
>
> I often encounter digital images that are so underexposed, for whatever
> reason, that all of the image information falls in the lower quarter of
> the histogram. Creating a normal tonal range by using the Levels command
> to stretch the histogram often introduces banding in any smooth
> gradation in the image; the same is not true of 16-bit images.
I'm ready to be convinced, and I think that this discussion can be
meaningful, provided it is illustrated with images.
---
Mike Russell
www.curvemeister.com
Bill Hilton
>
> Reasonable people differ on this topic
I think if you read the 20 best selling Photoshop authors you'll find
that 19 of them come out in favor of 16 bit editing for many
situations, certainly writers like Fraser and Deke McClelland and David
Blatner and John Paul Caponigro and Barry Haynes and Katrina Eismann.
The 20th is Dan Margulis and even Dan finally came around and admitted
16 bit files could be useful a few years back. So saying this is a
controversial subject doesn't really hold up except for a few hold outs
on the internet.
> I have serious issues with the Bruce Fraser example that Bill refers to.
> It is based on an artificial image
He's using a boundary condition (gradient) to make his point on the web
(which is not exactly set up to allow you to show subtle visual
differences), but you can see similar problems with real life images
that also have subtle shifts in color, especially when printed large.
Here's an abstract image of mine that is similar to a gradient in that
there are many subtle transitions ... I have to be very careful when
printing it or colors block up or wash out, depending on the printer
paper/profile used. Many of the transitions are lost just converting
it to sRGB for web display. This is the kind of image that benefits
from 16 bit/channel edits and careful color management ...
http://members.aol.com/bhilton665/tulipabstract.jpg
Bill
Clyde
This old topic again. <sigh> I guess I've been reading in newsgroups too
much.
I've also tested for my own satisfaction. I have never seen any
difference between 8 bit and 16 bit files. Of course, I have a testing
methodology that requires real photos viewed at real distances. I'm not
interested in viewing anything at 400% or any magnification; that is
completely useless.
However, I do edit many files in 16 bit mode. I guess I'm just being
safe. I do really twist some pictures sometimes and 16 bit may give me a
tad more smoothness. Luckily have have a very fast computer and it isn't
an issue. At the end they all get converted down to 8 bit.
Clyde
hoff...@fho-emden.de
as far as I remember, Dan Margulis says that it is useful to
import files with more than 8-bits-per-channel if the device
delivers this higher resolution (12 bits, camera RAW, scanner),
which leads to 16-bits-per-channel in PhS, because there is
nothing between.
After the few steps of basic corrections (levels and white
balance) everything else can be done with 8-bits-per-channel.
My experience.
Gradients are not images but test patterns. Gradients should be
dithered, anyway.
The rasterizer (halftoning, like Floyd-Steinberg) will mix up
everything: replace up to 16.7 millions of colors by drops of
four inks.
Some years ago, cute writers told us that gaps in a histogram
would indicate a bad image quality. Totally wrong, because
everything is mixed up by the halftoning process.
Best regards --Gernot Hoffmann
toby
hoff...@fho-emden.de wrote:
> Mike,
>
> as far as I remember, Dan Margulis says that it is useful to
> import files with more than 8-bits-per-channel if the device
> delivers this higher resolution (12 bits, camera RAW, scanner),
> which leads to 16-bits-per-channel in PhS, because there is
> nothing between.
>
> After the few steps of basic corrections (levels and white
> balance) everything else can be done with 8-bits-per-channel.
> My experience.
A series of corrections that modifies transfer functions (such as
levels or curves) will almost always produce a more accurate result
when carried out at a higher depth, because every sequential
apply-function-and-requantise will degrade the image (less so if
dithered, of course). The greater the precision of representation, the
less degradation caused by each step.
> Gradients are not images but test patterns. Gradients should be
> dithered, anyway.
>
> The rasterizer (halftoning, like Floyd-Steinberg) will mix up
> everything: replace up to 16.7 millions of colors by drops of
> four inks.
>
> Some years ago, cute writers told us that gaps in a histogram
> would indicate a bad image quality. Totally wrong, because
> everything is mixed up by the halftoning process.
Actually it's not totally wrong; a Floyd-Steinberg (or other error
diffusion) method if properly implemented can be regarded as a
continuous tone approximation; discontinuities and quantisation steps
in input will be faithfully rendered. In other words, every (R,G,B)
combination, whether that is 8, 16 bit or any other precision, will
produce a different halftoned balance. It is true that coarse
halftoning can *conceal* the existence of such artefacts, but fine
halftoning will faithfully reproduce them.
>
> Best regards --Gernot Hoffmann
Mike Russell
>> Mike Russell writes ...
>>
>> Reasonable people differ on this topic
>
> I think if you read the 20 best selling Photoshop authors you'll find
> that 19 of them come out in favor of 16 bit editing for many
> situations, certainly writers like Fraser and Deke McClelland and David
> Blatner and John Paul Caponigro and Barry Haynes and Katrina Eismann.
> The 20th is Dan Margulis and even Dan finally came around and admitted
> 16 bit files could be useful a few years back. So saying this is a
> controversial subject doesn't really hold up except for a few hold outs
> on the internet.
It would be more informative to see a photograph than a list of authorities.
In any case, some of these people reference Bruce Fraser, so this is really
a testimony to his well-deserved popularity.
>> I have serious issues with the Bruce Fraser example that Bill refers to.
>> It is based on an artificial image
>
> He's using a boundary condition (gradient) to make his point on the web
> (which is not exactly set up to allow you to show subtle visual
> differences).
So one must be very subtle indeed, to see these differences, viz:
Quoting from http://www.pitt.edu/~dash/type1620.html#andersen
"One day two [weavers] came to the emperor's city [...] , claiming that they
knew how to make the finest cloth imaginable. Not only were the colors and
the patterns extraordinarily beautiful, but in addition, this material had
the amazing property that it was to be invisible to anyone who was
incompetent or stupid."
> but you can see similar problems with real life images
> that also have subtle shifts in color, especially when printed large.
> Here's an abstract image of mine that is similar to a gradient in that
> there are many subtle transitions ...
At last, an image, and an interesting one at that. But you call it an
"abstract image". It has been modified so extensively that it barely
resembles a photograph at all.
> I have to be very careful when
> printing it or colors block up or wash out, depending on the printer
> paper/profile used. Many of the transitions are lost just converting
> it to sRGB for web display.
I would expect so, and I think you altered the image to achieve just such an
"abstract" effect. Although your image is derived from a photograph, it is
an artificial image, like Bruce Fraser's gradient. Your image is not
relevant to the way we should edit ordinary photographs.
> This is the kind of image that benefits
> from 16 bit/channel edits and careful color management ...
> http://members.aol.com/bhilton665/tulipabstract.jpg
Let's accept that this is "the kind of image that benefits from 16
bit/channel ". In one second anyone, child or adult, will see that your
example is very different from the vast majority of photographs. What can
we conclude from this?
Simply that you have never encountered a photograph that clearly benefits
from editing in hibit. If you had, being one of the more reasonable and
intelligent people in this news group, you would have made your point by
using that photograph, instead of an "abstract image".
If we can bear to controvert a handful of authorities, and if our images do
not look like highly blurred and saturated abstractions, those of us who do
most of our editing in 8 bit color, using the popular working spaces
supported by Photoshop, can rest easily.
Unless others express interest in this somewhat tired topic, I'll bow out
and leave Bill to have the last word.
---
Mike Russell
www.curvemeister.com
Clyde
Yes, but can you see it?
I don't argue that the mathematics would show that there are advantages
to 16 bit in almost every phase. My argument is that I can't SEE any
difference at 100% magnification down to normal viewing.
If it doesn't make a visible difference, it isn't important.
Clyde
toby
This thread can be boiled down to: "In some cases, yes."
DD
<RE-MO...@Curvemeister.comRE-MOVE> wrote:
>
>Unless others express interest in this somewhat tired topic, I'll bow out
>and leave Bill to have the last word.
>---
>Mike Russell
>www.curvemeister.com
>
I'm sure I am not the only one following this tread
word for word. But, enough is said for anyone to
come to a conclusion - in fact a relieve;
why bother about higher bit.
Dave
KatWoman
"DD" <d...@d.co.za> wrote in message
news:sgskl1ps85tjd5kik...@4ax.com...
scientific methods, remember silver halide printing, the zone system and
other methodologies for making "fine art" photos. Others of us need quick
commercial stuff that can be printed anywhere, or have clients who want
faster and cheaper. Some of us prefer working intuitively by 'eye" and
always have done so. Some of us farmed out darkroom work while some of us
slaved over our prints on archival paper with hand etching and coloring
techniques. Some guys take their film to Walgreen's. Some of you sold and
8x10 for $10.00 some of you charge considerably more
What are you trying to produce as your final result? of course different
techniques take you to different results. If you find good results in 8 bit
why bother with 16? especially if you can't see the difference. If you are
quite picky and find problems printing your files by all means use the 16
bit and see if you get better results. There is room for a wide variety of
techniques, everyone must find hi/her own niche and comfort level.
I have seen pro photog's like Ellen Von Unwerth actually brag about her lack
of technical expertise, her talent is in capturing a moment of feelings.
Some of her stuff is out of focus, mostly shot on automatic 35mm with normal
lenses. People pay her a lot of money for these poorly crafted but still
commercially viable work. Then you have the "Ansel Adams" crowd, so
carefully zoning by the numbers, using precise techniques, maybe making 100
prints to get one they consider worthy. Some of us value creativity,
freedom and ease of use over slaving techniques. Each method can produce
good work.
You have to work the way your muse directs.
hoff...@fho-emden.de
and Nan Goldin's photos are all 'out of focus'
but nevertheless more interesting (IMHO) than much
of this common pizza design Photoshop stuff.
'Pizza design' means: as many layers as possible.
The image processing aspect (now my engineering view):
No source image is 'true'.
How can we expect that the reproduction quality depends
on valid decimals (in limits) ? Especially the halftoning
process is always an approximation how to replace color
channel depth by a spatial distribution of ink drops.
Much progress in the last years, but no eternal scientific
truth.
There are many other improvements (like super cells) or
newer diffusion algorithms which can improve the printed
quality for 8-bit-per-channel CMYK (which is anyway the
standard for RIPs).
Best regards --Gernot Hoffmann
Bill Hilton
>
>It would be more informative to see a photograph than a list of
>authorities.
You personally held a "challenge" a couple years back, offering a cash
reward for anyone with an image that showed that 16 bits was better
than 8 bits for some cases, and while most images did not exhibit this
at least one did since you ended up paying off, right? So you should
already have one such photo in hand. I think Dan Margulis made a
similar request minus the reward, and got enough images showing the
benefits of high bit edits that he changed his mind on the topic.
>> Here's an abstract image of mine that is similar to a gradient
>> in that there are many subtle transitions
>> http://members.aol.com/bhilton665/tulipabstract.jpg
>At last, an image, and an interesting one at that. But you call it an
>"abstract image". It has been modified so extensively that it barely
>resembles a photograph at all ... I think you altered the image to
>achieve just such an "abstract" effect. Although your image is
>derived from a photograph, it is an artificial image, like Bruce Fraser's
>gradient.
No, it's not an "artificial image" with digital manipulations (which I
think is what you mean), it's an unmanipulated photo. I shoot stuff
like this a lot of the time, as do many other people I know and
respect. The only "manipulation" here is in-camera, adjusting the
plane and focus (or out-of-focus to be precise) to emphasize the shapes
and colors at the expense of detail.
>Let's accept that this is "the kind of image that benefits from 16
>bit/channel ".
OK ... and here's a screen dump showing the original RAW file in a RAW
converter window (and a few others from the same memory card), with no
digital modifications ...
http://members.aol.com/bhilton665/screen_dump.jpg ... I think this kind
of makes my point :)
>In one second anyone, child or adult, will see that your
>example is very different from the vast majority of photographs.
To me that's a good thing.
Wife and I have an exhibit scheduled sometime next year called "Photo
Impressionism" and all 16 images accepted for exhibition are similar to
this (various zooms, blurs, panning, multiple exposures, etc) and none
had any digital manipulation at all, so it's not like this is a one-of
case. Here are a couple of links to the guy doing this type of work at
the highest level, Freeman Patterson (I'm proud to have studied with
him for a week in New Brunswick, he's the most creative photographer
I've ever met) ... his book "Photo Impressionism and the Subjective
Image" is a good source for these techniques and he is strictly a film
guy with no digital work at all ...
http://www.freemanpatterson.com/prints/impressionism1.htm
http://www.freemanpatterson.com/prints/impressionism2.htm
Bill
Mike Russell
[re 16 bit mode]
> You personally held a "challenge" a couple years back, offering a cash
> reward for anyone with an image that showed that 16 bits was better
> than 8 bits for some cases, and while most images did not exhibit this
> at least one did since you ended up paying off, right? So you should
> already have one such photo in hand.
The first part is true. Three years ago, I did issue a 16 bit challenge.
Designing such a challenge was more difficult than I imagined. Although I
gave the final prize to Michael Schaffer, I did not end up with the
hoped-for image that does better when edited in 16 bits per channel than 8
bits.
I'll refrain from responding in detail to the rest of your post. I was
mistaken in thinking that your images were manipulated in an editor. This
does not change the gist of my argument, and raises the question of where
the benefit of 16 bits is, if you did not edit the image in question.
Your images are attractive works of art, and I expect that they will be well
received. Best of luck to both of you with the show!
---
Mike Russell
www.curvemeister.com
Ken Palmateer
suggests the (as yet) uncertain value of 16 bit over 8 bit. Given the
"givens" , at the very least, working in 16 bit simply guarantee
maximum (potential) quality. Ken
Bill Hilton
8 bit edits below but first ...
> Mike Russell writes ...
>
>I was mistaken in thinking that your images were manipulated in an editor.
No problem.
>This does not change the gist of my argument, and raises the question
>of where the benefit of 16 bits is, if you did not edit the image in question.
I showed that particular image because you were challenging Fraser's
use of gradients to make his point ... the image I showed has natural
gradients similar to what Bruce used, albeit over a much narrower color
range. My point was that such gradients do occur in photos at times.
Since my image was exposed properly it requires little editing, but
because of the saturated reds it looks much better converted from the
RAW into a wider gamut space like Ektaspace and then converted to a
tighter gamut profile with perceptual rendering, so for that reason I
would want to keep it in 16 bit mode. Had it been over or underexposed
you would probably see what Fraser was showing with his example.
Anyway, here's another image you might find of interest ... I spent the
first 18 days of Oct in New England hoping to photograph fall foliage
but it was a terrible year, the first week had record heat (mid 80's
instead of 50's where I was) which killed the colors, then we had
record rains and flooding, which didn't help the photography.
I got this shot of a loon in a heavy fog one morning, I wanted a dreamy
effect with just a hint of a shape to catch a mood ...
http://members.aol.com/bhilton665/A2921_start.jpg ... this is unedited
from the RAW, shot at + 1 2/3's off meter reading, and is still a bit
dark. You can see from the histogram that only about 10% of the tonal
range is used, with the bulk of the data between 214 and 235 or so.
The light was so low contrast that an f/2.8 lens would AF but a f/4
lens would not.
So for this thread I converted twice, once at 8 bit/channel and once at
16 bits, and then set the black point on the loon's neck and the white
point on a bit of foam floating near the bird, basically spreading 30
data points across 245 points (I had the white point set for 245).
Here is the 8 bit result ...
http://members.aol.com/bhilton665/A2921_8bit_setblackwhite.jpg
Here is the 16 bit result ...
http://members.aol.com/bhilton665/A2921_16bit_setblackwhite.jpg
Two concessions before you say anything ... 1) yes, this is an extreme
tonal change and 2) both look like crap at 100% ... no argument from me
on those points, but to me the 16 bit version looks less crappy with
less noise. And from the histogram you'd infer that further edits
(perish the thought) would cause less damage with the 16 bit version.
Now while you (or at least I) wouldn't make such a radical tonal shift
with this image (I like it the way I shot it) I did get asked to darken
the bird when I posted this image on a web site ... here's the image I
posted, created from an 8 bit conversion btw ... (click 'back' to see
others from the trip that are not so foggy :)
http://members.aol.com/bhilton665/fc2005/loon_A2921_b.htm ... the first
person to view this told me they could hardly see the bird and I needed
to adjust it, so if I were to do this I'd make a less drastic move of
the black point than in the example, but still a fairly major movement.
At some point you start to see the difference between making an edit
like this in 8 bit vs making it in 16 bit. Exactly how far you have to
move the black point before it becomes noticeable is something I'm not
going to check, but at some point ... if I'm lucky enough to sell some
large prints of this image I'll certainly do the prep work on a 16 bit
version.
Another poster in this thread wrote that some people are willing to go
to more trouble for small quality improvements and I'd plead guilty to
that ... if I could get a 5% improvement for a 100% increase in effort
I'd call it a bargain. I know others don't feel that way but I do.
One last image, this time without the before/after ... my wife shot
this hummingbird earlier this summer and the image was about 1.5 stops
underexposed because she had been shooting in a different direction
with minus exposure compensation set (needed + for this light) and
because the fill flash didn't re-cycle quickly enough for the flash to
be fully powered (something about shooting at 8 frames/sec when the
flash needs up to 1.5 sec to recharge) ...
http://members.aol.com/bhilton665/rufous_U8507.jpg ... the tonal range
wasn't as limited as the foggy loon shot but it was still compressed,
so when someone wanted an 11x14" print from a cropped area it looked
smoother (in the print) going back to a 16 bit file than it did using
the 8 bit version I had originally converted on the laptop for the
jpegs.
The common thread in both of these examples (one high key, the other
underexposed) is that you have a limited tonal range that gets expanded
with heavy edits. To me that's when 16 bits is better than 8 bits. To
each his own.
Bill
Mike Russell
news:84anl1966kcdfnun7...@4ax.com...
Photoshop CS now offers a 32 bit per channel format. Not all editing
functions are available, as was the case initially with 16 bit per channel.
Would you suggest changing over to 32 bit when possible, using the same
reasoning?
--
Mike Russell
www.curvemeister.com
Ken Palmateer
Well, somewhere along the way, the point of diminishing returns will
have to dictate where to draw the line. Can we see the diff between 8
and 16? Or is this a case of were we've got to climb the mountain
just because it's there? Ken
Bart van der Wolf
"Mike Russell" <RE-MO...@Curvemeister.comRE-MOVE> wrote in message
SNIP
> Photoshop CS now offers a 32 bit per channel format. Not all
> editing functions are available, as was the case initially with 16
> bit per channel. Would you suggest changing over to 32 bit when
> possible, using the same reasoning?
HDRI is not about more precise tonality, its 32-bits/channel are
needed to encode huge dynamic range. The numbers needed dont fit in a
16-bit integer number. The resulting tonemapping challenge is how to
fit that dynamic range in the limitations of 8-b/ch output modalities.
Bart
Mike Russell
news:1130088010....@g49g2000cwa.googlegroups.com...
> I'll try to show an image that does illustrate the advantages of 16 vs
> 8 bit edits below but first ...
[re Fraser's use of gradients to illustrate the hibit advantage]
http://www.creativepro.com/story/feature/7627.html
> I showed that particular image because you were challenging Fraser's
> use of gradients to make his point ... the image I showed has natural
> gradients similar to what Bruce used, albeit over a much narrower color
> range. My point was that such gradients do occur in photos at times.
> Since my image was exposed properly it requires little editing, but
> because of the saturated reds it looks much better converted from the
> RAW into a wider gamut space like Ektaspace and then converted to a
> tighter gamut profile with perceptual rendering, so for that reason I
> would want to keep it in 16 bit mode. Had it been over or underexposed
> you would probably see what Fraser was showing with his example.
I'll stick with my contention that Fraser's article is flawed. In the
interest of not repeating myself, let's accept, for now, Fraser's use of
artificial images, as part of an experiment to illustrate, in an abstract
way, the concept of 8 versus 16 bit manipulations.
Since one of the most basic requirements for an experiment is
reproducability, I think it is reasonable to ask that we be able to
recreate the result ourselves.
The Fraser article does not provide the original images, or any description,
that I could find, on how to recreate those images. So I created an image
that resembles the one Fraser used in his article, using a horizontal
spectrum gradient combined with a vertical black/white gradient in
luminosity mode.
Here's the surprise: I applied the same levels manipulation described in the
article. The result: there is no visible difference in the appearance of
the 16 and 8 bit results.
This took me all of 10 minutes to do, and I invite anyone reading this
article, to do so themselves. You will need a version of Photoshop that
supports 16 bit gradients.
> Anyway, here's another image you might find of interest ... I spent the
> first 18 days of Oct in New England hoping to photograph fall foliage
> but it was a terrible year, the first week had record heat (mid 80's
> instead of 50's where I was) which killed the colors, then we had
> record rains and flooding, which didn't help the photography.
>
> I got this shot of a loon in a heavy fog one morning, I wanted a dreamy
> effect with just a hint of a shape to catch a mood ...
> http://members.aol.com/bhilton665/A2921_start.jpg ... this is unedited
> from the RAW, shot at + 1 2/3's off meter reading, and is still a bit
> dark. You can see from the histogram that only about 10% of the tonal
> range is used, with the bulk of the data between 214 and 235 or so.
> The light was so low contrast that an f/2.8 lens would AF but a f/4
> lens would not.
Again, an interesting image.
> So for this thread I converted twice, once at 8 bit/channel and once at
> 16 bits, and then set the black point on the loon's neck and the white
> point on a bit of foam floating near the bird, basically spreading 30
> data points across 245 points (I had the white point set for 245).
>
> Here is the 8 bit result ...
> http://members.aol.com/bhilton665/A2921_8bit_setblackwhite.jpg
> Here is the 16 bit result ...
> http://members.aol.com/bhilton665/A2921_16bit_setblackwhite.jpg
Both links point to what appears to be the 16 bit image. Earlier this
afternoon both of them pointed to the same 8 bit image. While you are
fixing this, I'd like to know if you did your editiing in a wide gamut space
such as Ektspace?
> Two concessions before you say anything ... 1) yes, this is an extreme
> tonal change and 2) both look like crap at 100% ... no argument from me
> on those points, but to me the 16 bit version looks less crappy with
> less noise. And from the histogram you'd infer that further edits
> (perish the thought) would cause less damage with the 16 bit version.
Re the histogram, I reject it utterly as a sole measure of image quality.
If there are problems it should be visible in the image as well as in the
histogram. Furthermore, Photoshop injects fine noise when converting from
16 bit to 8 bit, and this added noise will "fatten" the histogram, and make
subtle transitions smoother.
> Now while you (or at least I) wouldn't make such a radical tonal shift
> with this image (I like it the way I shot it) I did get asked to darken
> the bird when I posted this image on a web site ... here's the image I
> posted, created from an 8 bit conversion btw ... (click 'back' to see
> others from the trip that are not so foggy :)
> http://members.aol.com/bhilton665/fc2005/loon_A2921_b.htm ... the first
> person to view this told me they could hardly see the bird and I needed
> to adjust it, so if I were to do this I'd make a less drastic move of
> the black point than in the example, but still a fairly major movement.
> At some point you start to see the difference between making an edit
> like this in 8 bit vs making it in 16 bit. Exactly how far you have to
> move the black point before it becomes noticeable is something I'm not
> going to check, but at some point ... if I'm lucky enough to sell some
> large prints of this image I'll certainly do the prep work on a 16 bit
> version.
This is your rightful decision about your own technique. Given your obvious
excellent results, I respect your decision, even though it appears, so far,
not to be based on the appearance of the image, but on your own estimation
of the underlying quality. I do not trivialize this. I believe it is the
right of every artist and craftsman to make many such decisions in the
course of their work.
But, if you are going to recommend that others use hibit images for their
editing, I think you should justify it by showing an actual image where you
believe it makes a difference. Perhaps the loon image does this - I haven't
seen it yet.
> Another poster in this thread wrote that some people are willing to go
> to more trouble for small quality improvements and I'd plead guilty to
> that ... if I could get a 5% improvement for a 100% increase in effort
> I'd call it a bargain. I know others don't feel that way but I do.
>
> One last image, this time without the before/after ... my wife shot
> this hummingbird earlier this summer and the image was about 1.5 stops
> underexposed because she had been shooting in a different direction
> with minus exposure compensation set (needed + for this light) and
> because the fill flash didn't re-cycle quickly enough for the flash to
> be fully powered (something about shooting at 8 frames/sec when the
> flash needs up to 1.5 sec to recharge) ...
> http://members.aol.com/bhilton665/rufous_U8507.jpg ... the tonal range
> wasn't as limited as the foggy loon shot but it was still compressed,
> so when someone wanted an 11x14" print from a cropped area it looked
> smoother (in the print) going back to a 16 bit file than it did using
> the 8 bit version I had originally converted on the laptop for the
> jpegs.
Again, a beautiful image, but without the ability to compare the 8 bit and
16 bit manipulations it does not advance the contention that 16 bit edits
give superior results.
> The common thread in both of these examples (one high key, the other
> underexposed) is that you have a limited tonal range that gets expanded
> with heavy edits. To me that's when 16 bits is better than 8 bits.
This may be the case, but without comparison images, we are reverting to
talk. You did provide comparison images, and for the life of me, I cannot
tell the difference between your 8 bit and 16 bit versions. Once again, I
would like to compliment you on the quality of your and your wife's images.
The hummingbird image in particular is incredible!
> To each his own.
We end in agreement. I quote from my first post to this thread:
"... many people, including some excellent professional photographers,
feel that even an imperceptible difference in quality may eventually be
significant to their images - if not now, then perhaps in the future. As a
tool provider, this is enough reason for me to fully support both 8 and 16
bit manipulations in Curvemeister."
I would add to this that the work of many of these individuals who choose to
work in 16 bit, as is the case with yours, Bill, is of excellent quality.
As the evidence stands, I feel justified in asserting that there is no
advantage to editing in 16 bits versus 8 bits.
Once again, I'm speaking of color images in the color spaces commonly used
in Photoshop. Monochrome images, linear gamma, and some of the more exotic
color spaces, such as ProPhoto RGB, are a different issue.
--
Mike Russell
www.curvemeister.com
Bill Hilton
>> http://members.aol.com/bhilton665/A2921_8bit_setblackwhite.jpg
>> Here is the 16 bit result ...
>> http://members.aol.com/bhilton665/A2921_16bit_setblackwhite.jpg
> Mike Russell writes ...
>
>Both links point to what appears to be the 16 bit image. Earlier this
>afternoon both of them pointed to the same 8 bit image.
I only posted them once and clicking the links shows two different
images to me, so I don't know what you are seeing or why you are having
problems with your browser ...
>I'd like to know if you did your editiing in a wide gamut space
>such as Ektspace?
This one was edited in AdobeRGB and converted to sRGB before 'save for
the web', which is my normal flow for most images... I typically use
Ektaspace when the colors are highly saturated and I'm going to print
it (instead of going to the web), like the abstract tulip, but with a
low contrast, low saturation image like this one it is pointless to use
a wider gamut space. I know many people would now use ProPhoto RGB
instead of Ektaspace since it's an even wider gamut space but I'm used
to Ektaspace from my Velvia film scans (the gamut was created by Joe
Holmes to match E-6 transparency film) so I still rely on it for
certain images.
As for the rest of your comments I don't have anything else to add that
hasn't already been said several times :)
Bill
Mike Russell
>>> Here is the 8 bit result ...
>>> http://members.aol.com/bhilton665/A2921_8bit_setblackwhite.jpg
>>> Here is the 16 bit result ...
>>> http://members.aol.com/bhilton665/A2921_16bit_setblackwhite.jpg
>> Mike Russell writes ...
>>Both links point to what appears to be the 16 bit image. Earlier this
>>afternoon both of them pointed to the same 8 bit image.
> I only posted them once and clicking the links shows two different
> images to me, so I don't know what you are seeing or why you are having
> problems with your browser ...
The 8 and 16 bit versions of the images are now showing up correctly.
The 8 bit version also has a strong yellow cast compared to the 16 bit
version that I cannot explain in terms of an 8 vs 16 bit difference. For
the time being I have to conclude that there is something else going on in
addition to any 8 versus 16 bit . Possibly the raw converter uses a
different algorithm to convert 8 and 16 bit images.
If you can provide me with a cropped version of the 16 bit image, I would
like to try my own experiments. Otherwise, I'll simply say this is an
intriguing image that may indicate a difference in editing capability
between 8 and 16 bits.
>>I'd like to know if you did your editiing in a wide gamut space
>>such as Ektspace?
>
> This one was edited in AdobeRGB and converted to sRGB before 'save for
> the web', which is my normal flow for most images.
No problems with those color spaces.
...
> As for the rest of your comments I don't have anything else to add that
> hasn't already been said several times :)
Perhaps someone will be intrigued enough duplicate my efforts to try to
recreate Bruce Fraser's results in his article:
http://www.creativepro.com/story/feature/7627.html
Take care. You've made some valuable contributions to the discussion.
--
Mike Russell
www.curvemeister.com
JFV...@att.net
Working in 16 bits will also guarantee:
- increasing the file size by several folds. After adding a few layers,
a high resolution image file can quickly be gigabytes in size.
- PS operations will dramatically slow down as the file size increases.
- Large files occupy lots of hd space.
As you questioned in a subsequent post, not all images or operations can
benefit from working with 16 bit. Even if the benefits can be expressed
in math or histograms, they may not be observable to the average (or
even expert) viewers. Images intended for web, or sent to a printer are
all converted to 8 bit. Those who choose to work with 16 bits typically
are working with high resolution files, and will pay the price mentioned
above.
Bill Hilton
>
>If you can provide me with a cropped version of the 16 bit image, I
>would like to try my own experiments.
I cropped out 600 x 300 pixels from each version (8 and 16) and can
email both to you if you want to play around. About 1 MB for the 16
bit one, half for the 8 bit one so not a problem to email them. This
has the loon (where I set the black point (0/0/0), on the neck) and the
bit of foam I used to set the white point (to 245/245/245) so you can
duplicate what I did.
I don't think I saved your email address from the last time we
exchanged files so if you could email me I'll send these in a reply ...
>The 8 bit version also has a strong yellow cast compared to the 16 bit
>version that I cannot explain in terms of an 8 vs 16 bit difference.
To me this just looks like added noise in the 8 bit version ...
> For the time being I have to conclude that there is something else
>going on in addition to any 8 versus 16 bit . Possibly the raw
>converter uses a different algorithm to convert 8 and 16 bit images.
I used Capture One LE, the files look pretty much identical before
editing, though I didn't do any numerical analysis on them.
Bill
Bill Hilton
>
>Working in 16 bits will also guarantee:
>
>- increasing the file size by several folds
The files double in size, which is not "several folds" ...
> PS operations will dramatically slow down as the file size increases.
Speaking from experience, this was a real problem with medium format
sized film scans but is not a problem with typical digital camera sized
files. My 6x4.5 cm scans were 340 and 170 MB (16 and 8 bit), which was
doable, but my 6x7 cm film 16 bit files were 550 MB and these did
indeed choke my computer, even with 2 GB of RAM and a high allocation
to Photoshop.
In comparison, now I'm working with a lot of 8 and 11 Mpixel files,
which are 48 and 66 MB in high bit mode (smaller than even 35 mm film
scans). These hardly slow the computer down at all since I'm always
working out of RAM.
> Large files occupy lots of hd space
With digital the RAW files are archived and they are a fraction the
size of high bit tiffs or psd files, so this is not that big a problem.
Also, internal hard drives are costing less than 50 cents/GB and
externals are often on sale for a buck a gig so disk space is not the
problem, getting excellent images worthy of keeping on the disk is the
problem :) I have two 200 GB internal drives and five external HDs for
multiple backups and don't really see a problem with keeping a small
subset of my files in 16 bit mode that are going to be converted and
printed.
Basically what I'm saying is it's much easier to store and work on
digital files in high bit mode than it is to work on high rez film
scans due to the relative size of the files.
Bill
toby
Bill Hilton wrote:
> >JFV4 writes ...
> >
> >Working in 16 bits will also guarantee:
> >
> >- increasing the file size by several folds
>
> The files double in size, which is not "several folds" ...
Compression is typically less effective on 16 bit images (such as the
RLE Photoshop uses for PSD). This could account for a >2-fold increase
relative to a (compressed) 8 bit file.
--T
>
> ...
>
> Bill
Bart van der Wolf
"toby" <to...@telegraphics.com.au> wrote in message
news:1130169315.6...@o13g2000cwo.googlegroups.com...
SNIP
> Compression is typically less effective on 16 bit images (such as
> the RLE Photoshop uses for PSD). This could account for a 2-fold
> increase relative to a (compressed) 8 bit file.
Which indicates there is a lot of "detail" at the Least Significant
Bit(s), even at 16-b/ch ...
It may be noise, it may be grain, it may be (e.g. color space
conversion) dithering, but it is part of the image. Accumulating that
"detail" in the LSBs of an 8-b/ch image only makes sense after all
corrections have been done, not before.
The remaining question is; will it show in print/on screen (whichever
the purpose)?
That is hard to predict, because there are many factors that can hide
it, but one thing is clear. If it shows, it is too late to remedy,
unless one starts all over again.
Bart
johns...@gmail.com
This article by Mike Chaney (Quimage author) very clearly addresses
this question, and strongly suggest that 16 bits can be very useful,
especially, as cited below, in instances of underexposure. He also
stongly suggests using "raw capture mode" if your camera supports it.
Here is the url :
http://www.steves-digicams.com/techcorner/April_2005.html
Mike Russell
news:1130254064....@g47g2000cwa.googlegroups.com...
Good point. I've experimented with this myself, and raw files can be very
useful for rescuing underexposed images, and for simulating a higher ISO
rating.
Whether editing in 16 bits is superior is a separate, and IMHO still open,
question.
--
Mike Russell
www.curvemeister.com