Big curve moves in 16 bits

[GSimian]

When working in 8 bits, it's easy to make a curve adjustment layer with a luminance blending mode, and then seriously change the shadows or the highlights in an image, without changing the color. In 16 bits, such a curve will oversaturate the color. There is no luminance blend in the 16 bit history brush, so I don't see how I can make such an adjustment, withour the strong color changes. Any ideas?

[Greg Vander Houwen]

Welcome GSimian,

Would a conversion to Lab mode then a Curve adjust on the Lightness channel work for you?

Greg

[GSimian]

Never thought of that! Thank you!! Would the RGB>LAB>RGB conversions lose any data (as Bruce Fraser claims)? Or would it be invisible?

[Greg Vander Houwen]

Well, I bow to Bruce on Color conversions and single malts ;)

I believe that if you convert from your working RGB into Lab and back the only loss you would experience is if you made an adjust in Lab that went beyond your RGB gamut and therefore on conversion would have to get translated back in. But let me see if I can find out.

Greg

[Rick Miller]

GSimian.

Why would you want to work in 16 bits per channel? Besides giving you a real pretty histogram, that's about the only benefit you'll get. I know Bruce Fraser is a Photoshop God, but I'll have to disagree with him on this one. No one has yet to show a picture that demonstrates the REAL WORLD superiority of working in 16 bits per channel.

Theoretically 16 BPC sounds awesome: 65536 levels > 256 levels - this should allow for smoother transitions and no combing of the histogram. Although in the real world, pretty histograms don't mean better pictures. Additionally your file size balloons to twice the size so you're left saying "yes, I do want a larger file size with no noticeable different... although I've got a real pretty histogram".
:d

I'll follow up with 8 BPC vs. 16 BPC later.

Additionally, you wrote:
Would the RGB>LAB>RGB conversions lose any data (as Bruce Fraser claims)? Or would it be invisible?

Well, it causes spikes in your histogram which don't look so pretty.
While it is lossy - as long as you only do it once, you're not going to see visual degradation on screen and I dare say: ain't gonna see it on print
either.

I'll follow up on this later as well.

Rick

[Bruce Fraser]

I'll respectfully agree to disagree with Rick over the efficacy of 16-bit/channel workflows. I've proven to my own satisfaction numerous times the benefits, and they have nothing at all to do with pretty histograms.

Yes, you will always lose some data to quantization error going from RGB to Lab and back. You'll also always lose some data making curve adjustments! With a 16-bit/channel file, the loss is unlikely to be readily detectable, and EXTREMELY unlikely to be visually significant.

So go ahead and do it, and don't worry.

Bruce

[Greg Vander Houwen]

Thanks for following up for me Bruce,

Good to see you here.

Junior member simply doesn't apply here so I've swapped it out.

Greg

[GSimian]

Ohmygod, Bruce answered my question! Now I can die happy!
Actually, I have a digital camera, which produces 16 bit digital files, and I want to balance them and correct them, before converting to 8 bits and handing them over to the client. Of course the client could apply their own adjustment layers, but I don't think they'll hire me again if I hand in flat, green images, or flat green images with a correction adjustment layer - just as you wouldn't hand in green chromes, even knowing the client will retouch the images to death, anyway. Thus, I want to hand them perfect 8 bit images. I'm open to any and all suggestions about automating such corrections (shoot a gray card, make an action that grey-balances to the card, applies the correction to all the images....etc.)
Thanks again to all of you for opening my eyes wide,
George Simian

[Rick Miller]

Ah, Bruce Fraser of Clan Fraser. Greetings kinsman, I hail from Clan Baird.
;)

It is excellent to see one of the elite masters of Photoshop join us here on Greg's site. And hey, another Scottish troublemaker, we like him already.

Now let's wrestle! (or a Caber toss or the Scottish Hammer)
:nuts:

While I'd go so far as to say your book Real World Photoshop should be required reading for anyone serious about using Photoshop professionally (it blew me away and kicked my butt), I'll have to join camp with Photoshop Hall of Famer Dan Margulis, and author of Professional Photoshop 6: The classic guide to color correction, on this one.

Dan has had numerous threads on his Colortheory list devoted to the 8 bit vs. 16 bit controversy, the most recent where he tested images submitted by Todd Flashner (16-bit files scanned from film in native mode from a high-quality Leaf scanner). Dan started with two files, one the original 16-bit and one he converted immediately to 8-bit.

Dan saved all modification settings and applied the same ones to both images. At the end of all corrections Dan converted the 16-bit file to 8-bit and compared it to the one in which all corrections were done in 8-bit.

Additionally, he submitted:

1. Description of the image being tested
2. Procedure
3. Result
4. Evaluation

Dan's assessment remained: "there's no benefit to
working in 16-bit (although there's no loss either)" and "
I have experimented a good deal with this and have never been convinced that there's any merit to it.

This issue becomes very mystifying to the average photoshop user (heck, even to most advanced users) when they hear those at the elite level disagreeing on the net advantage/disadvantage of a 16 bpc/8bpc workflow.

I've emailed Dan and asked him if it would be acceptable to post his 16 bpc vs. 8 bpc test and conclusions.

My position is:
One needs to critically examine the assumptions a position is predicated on: a critical thinker evaluates the claim and the warrant. The claim is made about the superiority of a 16 bpc workflow, the warrant (so far as I've read from proponents of 16)being that you have More levels: 65536 vs. 256 - obviously 65536 > 256 - and that the histogram of a 16 bit is smooth, unlike the gaps/banding/quantization that exist in the 8 bit. (I'm not trying to construct a straw argument, so correct me if this is not the case) I'm concerned about the output; if pretty histograms translated to pretty pictures I'd be pretty hip on 16, although I've not seen anyone demonstrate the superiority of the OUTPUT of a 16 bit file.

An argument establishing the validity of a 16 bit environment might be:

Claim: 16 bit is superior to 8 bit.

Warrant: The 65536 levels that allows for more information ("potential for detail"), provides superior output as demonstrated by picture "16 bit" and picture "8 bit".

The pictures and methods demonstrating this position would then be available for evaluation and critical examination by those interested in a 16 bit environment.

Additionally, Dan Margulis writes: I have experimented a good deal with this and have never been convinced that there's any merit to it. So, for several years, I've been asking those who claim that it works to provide me a few sample images that they think edit better in 16-bit.

For me to justify working in a 16 bit environment I'd need to see a demonstratable net advantage. A 16 bit environment doubles my file size and limits my ability to use certain functions of Photoshop (although your book addresses the latter in a very cogent manner).

I'm not so arrogant as to think one way is the only way - usually it's not. As we've seen there are alternative methods/workflows that people take. It's through healthy debate and discussion that we learn from one another and justify our assumptions and/or learn new ones. I must say, I've learned so much from your book, Real World Photoshop, and Dan Margulis' book, Professional Photoshop 6, than anywhere else - that's why these two books are required books for the classes I teach.

Are you any relation to the Scottish fiddler Alasdair Fraser? I saw him at the Scottish Highland Games in Pleasonton.

In honor of the picture of you in your kilt from Real World Photoshop, I humbly submit mine. (posted in the general photoshop discussion version of this)

Rick

[Bruce Fraser]

Rick,

I have never, anywhere, claimed that 16 bits is superior to 8 bits. What I've said is that from my own experience, 16 bits gives a great deal more editing headroom and flexibility than 8 bits. If you don't need it, don't use it. I've always been pretty straightforward about the plusses and minusses, and I'm not trying to sell anyone on 16-bit workflows.

For those people who, like me, have found that 16-bit workflows offer tangible advantages, I offer help and advice in working around many of its seeming limitations. For those who are curious, I offer some things to look for, but all I really acre about is helping people get images looking the way they want them.

Dan and I have a hard-won mutual respect, and generally enjoy each other, but trolling for controversy is much more up Dan's alley than mine. It's easy to create examples showing that 16-bit editing is unneccessary, or essential, depending on your agenda. I don't find such exercises worthwhile.

FWIW, the move shown on p349 of Real World Photoshop results in posterization that gets emphasized downstream if it's done on a 24-bit version of the raw image. I could probably have obtained similar results working in 24-bit from the start, but it would have been more work.

A large part of the reason I started using 16-bit workflows is the generally abysmal state of most scanner software. I quickly became convinced that the rational way to treat most scanners was to turn off all corrections, grab everything the scanner can capture in a raw 48-bit RGB file, and work from there in Photoshop. If you make the corrections in the scanner software instead, you're working on the high-bit file, you just aren't bringing it into Photoshop.

I've also observed a fairly large number of cases where converting to output space in 24-bit resulted in banding in skies, that disappeared when the conversion was done in 48-bit.

But it's basically like the argument over global warming. If you want absolute proof, it isn't there. I prefer to err on the side of caution. But I don't hold a gun to anyone's head and force them to work in 16-bit. If you feel it's unneccessary for your work, don't use it. It's really that simple.

If you want to argue, we should argue about something important, like why Lagavullin is so much better than Laphroaig...

[Greg Vander Houwen]

I am fond of the... if it works for you, use it theory. But it is valuable to know the when and why...

Now I need to hunt down some Lagavullin so I can be further enlightened.

Thank you gentlemen, good stuff.

Greg

[Rick Miller]

Bruce of Clan Fraser wrote:
we should argue about something important, like why Lagavullin is so much better than Laphroaig...
You gotta love the description they give Lagavulin:

The aristocrat of Islay malt whiskies...
DOMINATING CHARACTER OF DEPTH AND COMPLEXITY
Islay whisky is the most dramatic and complex of all. Whilst many differents elements contribute to its distinctive, robust character, the effect of the peat on the water that flows over it is clearly evident in the whiskys taste.

Lagavulin embodies all these attributes, but stands out from other Islay malts by its extra depth and silky dryness. Probably the most pugent Islay malt, Lagavulin is distinctive and robust with a heavy, powerful aroma. The smoky peaty taste is a delight to the discriminating drinker. Bottled with unusual maturity at 16 years, this is a malt to be sipped and savoured.

Oooh, that has Suavé written all over it! If the taste is in anyway near its description then I submit its nomination as the Official Scotch Malt Whisky (along with Guinness as the Official Beer and Blue Hawaiian as the Official Tiki drink). How say you Greg?

Bruce, it has been both an honor and a privilege sharing the field of academic inquiry with you.

Cheers!

http://www.scotchwhisky.com/images/lagavul2.jpg
Off to go find me some Lagavulin!

Rick

[Jeff Schewe]

Well Rick. . .

I'm a little less flexible than Bruce when it comes to 8 bit vs 16 bit. Personally I strongly advocate starting all color/tone edits in 16 bit before dropping down to 8 bit for creative manipulation. Why? Because since starting to do this, banding has not been a problem for me. Prior to scanning or capturing in 16 bit, banding was a constant problem.

I'm well aware of Dan's "Challange" but feel no inclination to get sucked into it.

Fact is, if you start serious image editing in 8 bit, you only have 256 tones/channel. You tone or color correct, you have lass. The banding may not show up at this point. Then, you do some more editing. . .retouching, cloning, filtering-at each step, you drop data due to rounding errors or just data loss. About the time you've spent a few hours editing an image, you may notice that there are areas where grdations are no longer smooth. Running a filter (G-Blur is a great one to show banding) or doing layer effects may break the gradations further. At this point, the odds are real good that the real risk of banding occurs. Or, maybe it's at the stage where you separate to CMYK. Whenever. . .some time, somewhere you go over the edge and drop enough tones where banding hits the image.

Then what do you do? You can't go back and put in data that is lost.

So, my approach is to preserve all the data up front so I have more data at the end to spend. I only start real manipulation _AFTER_ I've done all the tone and color correction in 16 bit. I only do my serious manipulation on "Prefect 8 bits/channel" images.

You can take Dan's argument if you like. . .a lot of people do. But, while I agree there is no "image" where tone & color correction is superior in 16 vs 8 bit, and therefore the ultimate "proof", I will tell you that banding isn't a problem for me and others who START in 16 bits and only drop down to 8 bit when the tone and color is already corrected. . .then, the 256 tones go a lot further.

Prove it to you? Don't need to. . .I prove it to myself every day. I'm perfectly happy working my initial corrections in double the file size. It's an easy trade off for the remarkable lack of banding problems I have down the road.

Again, 16 bit in the begining to preserve as much data as you can and then spend that data wisely in 8 bit = no banding at the end.

[Jeff Schewe]

P.S. Greg. . .I get a kick out of this forum. . .you've done a lot of work on it. . .

But. . .when does one advance beyond "Junior Memeber" ? <BG>

[sPECtre]

I think that it's around 30 posts, but I guess that the gurus like you deserve something else than junior...

I'm really happy to see that the "big guns" coming to this forum, thats the good thing about our community : everybody can share ideas, teach, learn, etc...

Mr Schewe, Fraser, Greg, and the others, thank you!

[Hannu Hurme]

Yes, things don't look too good for GSimian though, still only zero posts (somebody fix that :))...

[Rick Miller]

Greetings Jeff.
Greg must do something about that "Junior member" thing for you.
:D

Now for the content:

As a critical thinker I want to base my workflow on something that is manifest and provable.
Correct me if I'm wrong, but I respectfully and humbly submit that no empirical evidence exists that indicates the image output superiority of a 16 bit workflow over an 8 bit workflow. The net results allow one to critically examine the assumptions the 16bit workflow is predicated on; if empirical evidence indicated that a 16bit workflow output superior images over an 8bit workflow then I'd advocate this position.

As a teacher my students demand me to justify my assumptions - if I tell them that a 16 bit environment offers superior output quality, then I better be able to show them real world pictures that warrant this claim. As of now, I can't. Anything else is supposition.

Bruce talks about the added editing headroom, and as I originally stated - this makes sense theoretically (65536 levels > 256 levels). Although my position deals with the output from these environments, not the editing headroom. My original post advocates the "noticeable difference" and "better pictures" standards. Additionally, my original post indicates "Theoretically 16 BPC sounds awesome: 65536 levels > 256 levels - this should allow for smoother transitions and no combing of the histogram. Although in the real world, pretty histograms don't mean better pictures." As you can see, this is quite consistent with the "noticeable difference" and "better pictures" standards I'm talking about.

You indicate:
"But, while I agree there is no "image" where tone & color correction is superior in 16 vs 8 bit, and therefore the ultimate "proof", I will tell you that banding isn't a problem for me and others who START in 16 bits and only drop down to 8 bit when the tone and color is already corrected. . .then, the 256 tones go a lot further."
This has been my position all along, applying the "noticeable difference" and "better pictures" standards - no evidence exists that indicates the superiority of outputting images from a 16bit workflow. I believe we're in congruence on this part.

As I'm unfamiliar with the relationship between you and Dan Margulis I don't know what your comments regarding Dan's "Challange" pertains to. I'm simply interested in a cogent academic inquiry that is applicable in the real world environments my students work. Dan's inquiry/challenge seems to fit this bill.
As I indicated in a previous post - this issue becomes very mystifying to the average photoshop user (heck, even to most advanced users) when they hear those at the elite level disagreeing about the net advantage/disadvantage of a 16 bpc/8bpc workflow.


Dan Margulis, one of the first four Photoshop legends to be inducted into the Photoshop Hall of Fame (http://www.photoshophalloffame.com/winners.html) has been extremely generous with his time in this matter; the tests he conducted covered:

1. Description of the image being tested
2. Procedure
3. Result
4. Evaluation

Dan's assessment remained: "there's no benefit to working in 16-bit (although there's no loss either)"
and "I have experimented a good deal with this and have never been convinced that there's any merit to it. "
Additionally, in my email contacts with Dan he indicated: "Personally, like you I've seen no advantage in any of the tests I've done."

Dan and Bruce are among my favorite Photoshop authors (along with Davis, Evening, Caponigro, Eismann, Haynes and Crumpler, and Deke) I owe a significant amount of my advanced photoshop knowledge to them. The one thing that's constant among them: they have different workflows. You and I have different workflows (although I must say I really dig the Texas Longhorn(?) skull you've got above your workspace...cool).

On another issue:
I think fellow Photoshoppers here would love to get their hands on your book "Jeff's Photoshop" when it comes out. Would you mind posting an outline?

Respectfully...
Rick

[Jeff Schewe]

Rick. . .

Again, it's not the first color & tone edits that cause the banding. It's the accumulated data loss over a period of various editing proceedure that cause banding. That's why Dan's "proof" is not relevant. I don't have an image where a few edits will show conclusively that 16 bit is better than 8 bit. I don't think such an image exists. And, there's no assurance that a particular series of step WILL produce banding. That's why Dan's stance is rather irritating. He's not saying 16 bit isn't better. . .he's just wanting somebody else to prove it to him & the world.

Well, I don't go for that. In my experience, when starting in 8 bit, I have experienced banding due to loss of data in gradients. Since starting to scan & capture in 16 bit and tone/color correcting in 16 bit before dropping down to 8 bit to do manipulations, I've experience little or banding. What does that tell you?

Again, it's all about data conservation-and how one spends the data and what are the results. A 2 or 3 step process won't prove anything one way or other. But, let me ask this, have you experienced banding in images? If so (and I assume you have experienced this) to what do you atribute this to? Then, ask yourself this, what could be done to prevent the banding? What I am telling you is that if in the _VERY BEGINING_ one uses 16bit to tone & color correct (and this assumes that the 16 bit/channel input _IS_ indeed 12, 14 or 16 bit) then one has the full 256 tones to play with and spend in manipulations after dropping down to 8 bit.

Now, which images are most suseptable and therefore more likely to band? Images that have long tone gradations such as a studio lit gradated backgrounds or skies with subtle shading. Images with lots of textural detail probably won't band even with radical tone corrections. Images where tone adjustments need to be made using gradations (graduated lighting or darkening) will greatly benefit form being done in 16 bit vs 8 bit. Doing large area dodging and burning will break up and band real easy in 8 bit.

Again, what your head tells you is that you want some sort of proof that 16 bit is superior to 8 bit, right? Well, that's something you need to prove, one way or the other, to yourself. Dan's approach is that he wants somebody else to prove it to him. . .meaning somebody else has to do the work to prove it to him. My feeling is that I've proved it sufficiently to myself that I don't need to prove it to him <BG>.

Your milage may vary. . .

[Jeff Schewe]

BTW. . .

Amazon seems to think my book will be coming out this month, which is news to me and my publisher. The book isn't done and I'm not sure when it will be. So Amazon has it wrong. But telling them anything is like arguing with the IRS. They never admit they've made a mistake ;~)

[Greg Vander Houwen]

Jeff my friend,

Thanks for coming by. It's great to see you up here.

But. . .when does one advance beyond "Junior Member"

Right now for you ;).

All,

Jeff is an old friend and holds my deepest respect in his artistic/photoshop skills. He is one of those rare people that is both a skilled artist and true master of Photoshop.

Please give him your welcome and check out his work at...

http://www.schewephoto.com

Greg

[Bruce Fraser]

I wanted to pick up on a key point that Jeff made.

He does all his serious manipulations starting with "perfect" 8-bit files. The key point is that all the big moves are made on high-bit data. It doesn't matter whether you edit the high-bit data in the scanner driver or in photoshop, but most scaner drivers have pretty crappy controls and expect you to make decisions based on a postage stamp. That's one reason I prefer to bring 16 bits into Photoshop.

That said, there's plenty of work that doesn't require a high-bit workflow -- you need to maintain a sense of proportion. If the job at hand is a plumbing parts catalog, you probably don't need to start out with super-hi-res high-bit scans of three-inch elbows and the like.

But if you're working with imagery in which you're emotionally invested, or if you're attempting the kind of photorealistic compositing for which Jeff is justifiably well-known and well-compensated, you will have a much easier time if you do all your big moves in 16-bit, particularly when you're trying to match components that were shot under different lighting conditions.

In some ways, the difference between 16-bit and 8-bit is like the difference between chromes and negs. Chromes are pretty unforgiving: either the exposure is right, or it isn't, and once an image is rendered on transparency film there's relativly little that you can do in terms of manipulating contrast and color balance before things start to fall apart. The same is true for 8-bit images.

Negs, on the other hand, always contain much more information than you can print. Negs are essentially unrendered images -- the rendering happens in the printing process, and you can make many different renderings from the same neg. Look at the way Ansel Adams' Moonrise over Hernandez, NM evolved over his career for a good example. 16-bit images behave much more like negs than chromes in this respect.

A 16-bit workflow probably isn't for everyone. It really depends on what you're trying to achieve. 16 bits gives you much more flexibility and much more headroom in deciding how you want to shape tone and color for a given output process. (Parenthetical note -- you generally need to make different optimizations and different compromises for different outputs -- there's a big difference between a Lambda print and a newspaper.)

If you don't need high-bit imagery, don't use it. But if you run into occasional problems with posterization or color-banding, I suggest that you at least try a high-bit workflow and see if it ameliorates or even eliminates the problem. Jeff refers to people who start out in 8-bit as "recreational Photoshop users." I tend to think that's a bit harsh, but for the kind of work he does, it's probably essential (those of you unfamiliar with his work might want to visit www.schewephoto.com).

The other reason I always scan 16-bit is simply that exposed film is fragile and irreplacable. I only want to subject it to the rigors of scanning once, and I want to make sure that I capture everything the scanner can pull from the film when I do so. That's just my personal agenda. If it rings a bell, you may be a candidate for a high-bit workflow. But there are many other equally valid agendas, and many of them don't need high-bit scans. Do whatever works for you. I offer high-bit techniques to those people who run into situations where 8-bit doesn't work. If you aren't one of them, that's fine.

[Jeff Schewe]

Greg. . .

Thanks for the kind words and warm welcome and for getting rid of the junior member status ;~)

This seems like a pretty cool and hip place to hang out. Glad I finally found it! And, thanks for spending the time and effort to create the community.

Many thanks. . .

[Hannu Hurme]

Jeff your portfolio is truly amazing, but I have to say that the pig is definetly my favorite :D.

[Jeff Schewe]

Bruce brought up the term "recreational user" in reference to working in 8 bit vs 16 bit. Yes. . .I've said that and yes to a certain degree, I believe that, but. . .I say it tounge in cheek ;}

I also say it to remind myself that for many years, while working on imaging, I worked in 8 bit. It really wasn't until I got an Imacon FlexTight II scanner and discovered the true benefits to getting a reasonably good basic scanner setting and then working in Photoshop. Doing all the color & tone adjustments in Photoshop-where as Bruce indicates, one can rely on the screen view and not have to twiddle with a tiny preview image made a major change in my workflow.

Also, as Bruce suggests, a 16 bit workflow isn't really required for many applications. A simple tone or color correction isn't going to be radically better in 16 bit. And yes, the file sizes DO get pretty big, particularly when one scans at full optical resolution as I do. However, I rarely archive the 16 bit versions (I know Bruce usually does), just the 8 bit corrected version after corrections. So, I only have the initial pain of larger file sizes for the duration of the initial corrections.

[sPECtre]

Sirs (it's an habit over here)...

Excuse me to disrupt your delicious argument, but me, humble scarab, got a few questions that you may answer...

Don't you think that the output device got quite an importance over there? I know you speak about high-end requirements, you use 48bits scanners, or 12bits/channel cameras but...

Heck, the monitors/graphic cards combo doesn't show enough colors, afaik, they work in millions of colors... isn't it a first bottleneck?

Second, the printer, I'm wondering what is the number of shades it can render, even on a dye-sublimation printer You Photoshop Gurus should use...

I'll humbly go back to meditate and learn about your instructive posts...

[Jeff Schewe]

sPECtre asked about output which is a worthy subject. . .

However, the issue really isn't directly complicated. To my knowledge, there is no output device (inkjet, dye-sub or image setter) that can use 16 bit/channel data. And even though it would be useful, monitors will never really be able to use 16 bit either.

But, it's not the monitor or output that makes 16 bit data important. It's the fact that when doing initial tone/color correction in 16 bit and then dropping down to 8 bit for further editing, you are _STARTING_ your serious imaging with a perfect 8bit/channel image (full 256 levels of gray). That way you are maxing out the data you start with. Yes, the output will only be able to render 3 or 4 channels of 8bit (256 levels of gray) but understand that Photoshop is pure math. Every operation, brush stroke or layer blending mode is math. It's not practical to maintain and infinate level of mathmatical presision so at some point mathmatical results are rounded and others are averaged. Thus, over time, data is just lost. So, where to may start with 256 levels of gray, you may loose levels and end up with say only 200 levels. Depending on the image and it's gradations, the levels lost may be enough to introduce banding.

Understand that initial tone/color corrections can often result in some pretty heavy lost of levels. Brining in white and black points in the Levels adjustment basically chops off the clipped data. Once clipped, it's gone forever. Loose enough levels and you'll get banding. Starting the tone/color corrections in 16 bit (65,000+ levels of gray) allows, as Bruce calls it more Head and Tail room.

So, yes, 16 bit won't do any good for monitors or print output, but it's important to preserve as much data (descrete levels of tone) during editing so that at the end, you have as close to 256 levels as possible.

That help?

[Bruce Fraser]

Basically, what Jeff said.

The object of the exercise in high-bit editing is to get the best possible 8-bit image.

When you start out with an 8-bit/channel image, pretty much everything you do to it results in an image that contains fewer than 256 levels per channel. When you use Levels to set blacj and white points, you not only clip everything above your white and below your black point, you also stretch the remaining levels across the 256 possibilities, so where you fomerly had smooth transitions from one level to the next, you now have gaps where some levels are unpopulated. The gaps won't be visually significant, and may not be visible at all, at the time you make the move, but subsequent edits exacerbate them, and eventually they may reveal themselves as obvious posterization or color-banding.

Various people over the years have criticized me for being hung up on pretty histograms. They miss the point. I teach people about histograms simply because they show what happens when you edit images. A pretty histogram doesn't make a good image, and an ugly histogram doesn't necessarily make a bad one. The histograms are simply tools that let you see what's happening, and while you're learning how to manipulate pixels, checking the histograms is a good way to get a handle on what's going on. A lot of people are quite surprised when they see how much data a simple curves or levels move, or a color space conversion, simply throws away.

A newsprint press can probably print at best about 5000 discrete colors, but what really matters is the way the colors in your image map onto the colors in the output. A crappy output process may mask discontinuities in the image, but it's equally likely to emphasize them.

I haven't done dye-sub for a long time, but printers like the Fuji Pictrography, the Durst Lambda, or the Cymbolics LightJet can come very close to reproducing the theoretical 16.7 million values in RGB. Whether the eye can distinguish anything like that many is an entirely other question...

...and of course, you'd need a very big monitor indeed to display 16.7 million pixels!

If you never have a problem with posterization or banding in your output, you probably shouldn't bother considering a high-bit workflow. But I'd like to know your secret!

[sPECtre]

Of course, you are right! :D

I was not trying to contradict you, it was just something that popped up while reading this thread, and you came, as always with instructive and "common sensed" info...

(now the graphic card makers and printers builders, listen, we DO need 48bit+ output devices ;) )

(about the monitor, Big Blue made one with 9 millin pixels, but I think everybody is waithing for a graphic card to use it! :) )

[markzebra]

Surely - its all about output

The best output you are likely to working to is around 133 lpi or if your lucky 150 lpi. Most of the data that you are trying to preserve is lost in the printing process? has anyone ever seen a print which demonstrates an improved quality with a 16 bit original?

Histograms either look beautiful and smooth or have a nasty habit of looking jagged or messy - What is the REAL actual loss of data that those white stripes represent? probably less than 5% - and with most printing systems it surely has no impact. Yes, sure when you see it on your screen (with millions of colours), you can maybe see a difference.

Maybe when the power of our systems makes it possible for Photoshop to work comprehensively in 16 bit mode the whole argument will become redundant .. or will it. Something tells me that someone, somewhere will be trying for 32 bit editing. Where is the argument to increase printing resolution to take full advantage the huge amount of data that our digital files hold?

[Jeff Schewe]

it was written:
"The best output you are likely to working to is around 133 lpi or if your lucky 150 lpi. Most of the data that you are trying to preserve is lost in the printing process? has anyone ever seen a print which demonstrates an improved quality with a 16 bit original?"

You are mixing apples and oranges. . .we're not talking overall color gamut loss as in converting from RGB>CMYK, we're talking about "data loass" that will end up giving less than 256 levels of tone. And, in point of fact, the worse the reproduction output is, the more critical it is to preserve a smooth gradations.

No, there is no output that can handle 16 bits/channel, but yes, almost any output can suffer banding on conversion and output and therefore benefit from initial 16 bits/channel corrections.

You ever see banding in your output? If you have, that comes from _NOT_ having enough data points in the gradations. . .doesn't have anything directly to do with the small gamut of color.

[markzebra]

Are you sure that the banding on your output wasn't due to either the inadequacy of the RIP or some other crudeness in the printing?

I can understand the logical argument for increasing the sheer quantity of data at the earliest stages (especially in the scan which is notoriously crude on quality) but have yet to see how this can be represented on a reduced quality print.

Don't think i mentioned anything about conversions. I was talking as you were, about data loss, as "illustrated" with banding in a histogram, and was asking what percentage this data loss actually represents.

[Jeff Schewe]

One of the stages where previous data loss due to prior editing will really show up is doing RGB>CMYK conversions. . .if you have _ANY_ near banding data loss in an RGB file, doing the CMYK separation will often cause it to break and the banding to come into full blossem. That's why it's so critical to preserve the data in the early stages of editing. . .data lost can never be recovered. . .

[sPECtre]

for the ones intersested by this thread, some of the information discussed was about Dan Margulis statements. You can find some articles he wrote, excerpt from his book, and excerpts of the colortheory lists here (http://www.ledet.com/margulis/articles.html)

I think it is a must read for EVERY photoshopper!

[sPECtre]

And about the "Margulis Challenge" I found this page :
http://www.brucelindbloom.com/index.html?DanMargulis.html
Made by a VERY knowledgeable member of the "color theory" lists

[docilebob]

This is great stuff. Thank you gentlemen. %)

It`s rare to find this level of competency and maturity in forums on the web. I`m glad you`re all here.( and that I found this place. )

Thanks again for your time and sharing your expertise.