From: Dan Margulis
Date: Mon, Feb 21, 2000, 8:35 PM
RE: By the numbers, bad shadows
Folks,A problem that continually comes up is people who attempt to correct in CMYK by the numbers but start with files that were separated with different numbers in mind, especially when separated with Photoshop's goofy black defaults. Here's a sample of such a problem; my reply is in a separate message.
Dan Margulis
--------------- Forwarded Message ---------------
I'm trying to correct some scanned photos I have. I've identified the hilites, shadows, etc. and am attempting to use curves to improve the pics. However, when I try to make any photo I have conform to the 5c3m3y hilite / 80c70m70y70k shadow formula you have suggested. I always get a picture which is somehow worse than when I started.
I've tried to carefully calibrate my monitor. I think I know what I'm doing when I adjust the curves, etc. I'm not sure I really know exactly where on the curve to adjust.
For example, in one photo I have a hilite of 3c0m1y0k and a shadow 64c52m51y91k. When I massage the values with curves to that they conform to the specs you suggested I get a result that is really wonky.
I change the 3c to 5c and the 64c to 80c. The pic gets decidely blue. I change the 0m to 3m and the 52m to 70m, eliminating much of the blue cast. I change the 1y to 3y and the 51y to 70y and get a yellow cast in parts of the pic and a blue cast in the other part. I leave the 0k, but change the 91k to 70k. Now I'm left with a pic that has a bit of a blue cast in one part, a yellow cast in another, and a general washed out effect.
Can you help me at all, in a general way? I've really scoured the first several chapters of your book, but I'm still not getting good results.
I think I'm not really sure exactly the best way to change the curve. For example, to change the cyan place, I just found the 3c mark and changed it to 5c. And I found the 64c mark and changed it to 80c.
Perhaps, I'm going about things incorrectly. Anyways, any help you could suggest to me would be really appreciated. I have several students and other teachers I would like to be able to help. However, I've got to figure it out myself first.
From: Dan Margulis
Date: Mon, Feb 21, 2000, 8:35 PM
RE: Re: By the numbers, bad shadows
>>I'm trying to correct some scanned photos I have.>>Without seeing them, etc., etc., but still from what you describe I believe I can give some assistance.
>>For example, in one photo I have a hilite of 3c0m1y0k and a shadow 64c52m51y91k. When I massage the values with curves to that they conform to the specs you suggested I get a result that is really wonky.>>
What you say suggests two major problems. 1) the separation settings that were in use when this picture was brought into CMYK were Photoshop's poor defaults, which result in this eccentric shadow value and make the image much more difficult to correct. 2) The shapes of your curves are probably wrong, and in the black channel they are definitely wrong. Because of the poor separation defaults the moves in the color channels are having a disproportionate effect.
To demonstrate this, I'd suggest making three copies of the original image. With the first, load the curves you had written that gave you the wonky result, and save the image for comparison.
With the second version, load the very same curves, but before clicking OK go into the black curve and eliminate the point you had put in, substituting the following instead. Take the top right (100%) point and drag it downwards to 80%. Now insert a second point at 30% and drag it up to 35%. Compare the result to your first one. Does wonders for eliminating that washed-out look, eh?
With the third version, open your CMYK Setup (PS5) or Separation Setup (PS4) and be sure that the values are set to Light GCR, 85% maximum black, 300% total ink. Now, convert the image to LAB color, and convert it back into CMYK. This will force Photoshop to honor these new settings, and your shadow should become, roughly, 80C70M70Y75K, which will be far easier to work with than what it used to be.
OTOH, for this particular image the values seem to have been fairly close to begin with. So you may or may not be able to improve things. The big gains show up when some of the numbers are seriously off.
Dan Margulis
From: J Walton
Date: Tue, Feb 22, 2000, 10:43 AM
RE: RE: By the numbers, bad shadows
Personally, I would balance the Hilight/Shadows in RGB. Especially when a new black needs to be generated anyway; in RGB you just make the numbers equal (if nuetrality is the goal) and then you can make global contrast adjustments in the Master Channel.I know, I know, Dan doesn't like global corrections. ;~) Anyhoo, RGB is an awfully nice tool when you have to "rescan" a badly scanned image.
From: Luca Ragogna
Date: Tue, Feb 22, 2000, 9:25 AM
RE: Re: By the numbers, bad shadows
on 2/21/00 8:31 PM, Dan Margulis wrote:> With the third version, open your CMYK Setup (PS5) or Separation Setup
> (PS4) and be sure that the values are set to Light GCR, 85% maximum black,
> 300% total ink. Now, convert the image to LAB color, and convert it back
> into CMYK. This will force Photoshop to honor these new settings, and your
> shadow should become, roughly, 80C70M70Y75K, which will be far easier to
> work with than what it used to be.Don't you have to convert to LAB first, then change the settings for there to be a change?
From: Dan Margulis
Date: Wed, Feb 23, 2000, 12:14 AM
RE: RE: By the numbers, bad shadows
JWalton writes:>>Personally, I would balance the Hilight/Shadows in RGB. Especially when a new black needs to be generated anyway; in RGB you just make the numbers equal (if nuetrality is the goal)...>>
It's pretty easy to set the highlight/shadow in whatever colorspace you like, and probably easiest of all in LAB. No quality loss or gain one way or another.
>>...and then you can make global contrast adjustments in the Master
Channel.>>But this is very bad advice. It's true that using the master curve isn't the guaranteed disaster that it would be in CMYK, but it's almost invariably going to cost contrast as opposed to three individual channel curves. Those serious about color correction should never be using the master curve either in RGB or CMYK.
Dan Margulis
From: Bill & Mary Anne Stewart
Date: Wed, Feb 23, 2000, 8:45 AM
RE: RE: By the numbers, bad shadows
>Those serious about color correction should never be using the
>master curve either in RGB or CMYK.
>Dan MargulisAbsolutely. Dot gain curves in all channels is in my view the single biggest improvement in PS 5 and 5.5,
Bill Stewart
The Bayberry Design Co., LLC
20 Hartford Road, Suite 32
Salem, CT 06420
stewart@bayberrydesign.com
http://www.bayberrydesign.com
Digital Graphic Design and Publishing Solutions for Print and the Internet
From: Bob Smith
Date: Wed, Feb 23, 2000, 8:18 AM
RE: Re: By the numbers, bad shadows
Dan Margulis wrote:> Those serious about color correction should never be using the
> master curve either in RGB or CMYK.I can see why master curve corrections might be a problem for CMYK but why in RGB? Assuming I'm pleased with color balance from highlight to shadow and only want to increase or decrease contrast why would individual channel moves be more appropriate than a global move?
Bob Smith
From: Dan Margulis
Date: Wed, Feb 23, 2000, 12:14 AM
RE: Re: By the numbers, bad shadows
Luca writes:>>> With the third version, open your CMYK Setup (PS5) or Separation Setup
> (PS4) and be sure that the values are set to Light GCR, 85% maximumblack,
> 300% total ink. Now, convert the image to LAB color, and convert it back
> into CMYK. This will force Photoshop to honor these new settings, and your
> shadow should become, roughly, 80C70M70Y75K, which will be far easier to
> work with than what it used to be.Don't you have to convert to lab first, then change the settings for there to be a change?>>
I *did* change the settings. They must have been at 100% before the initial separation was made. I have changed them to 85% before reseparating.
Dan Margulis
From: Andre B. Lopes
Date: Wed, Feb 23, 2000, 7:12 PM
RE: Re: By the numbers, bad shadows
Hi, Dan.I've been reading the last messages posted at Applied Color Theory and I believe you haven't understood the main problem in a question.
You wrote:
>> With the third version, open your CMYK Setup (PS5) or Separation Setup (PS4) and be sure that the values are set to Light GCR, 85% maximum black, 300% total ink. Now, convert the image to LAB color, and convert it back into CMYK. This will force Photoshop to honor these new settings, and your shadow should become, roughly, 80C70M70Y75K, which will be far easier to work with than what it used to be. >>
Luca wrote:
Don't you have to convert to Lab FIRST, then change the settings for there to be a change?>>
(the emphasis is mine)You answer:
>> I *did* change the settings. They must have been at 100% before the initial separation was made. I have changed them to 85% before reseparating. >>
Well, I believe Luca was in doubt about the CORRECT TIME to alter the CMYK Setup, and his question make some sense in many situations (although not in this specific one). I've found the same doubt many times in my classes. Let me explain.
If I have a CMYK Photo color adjusted to a "Premium Printing CMYK Setup" (Eurostandart/coated/12% dot gain/light GCR/black limit 95% / total ink 320%) and I want to convert it to some kind of "Newspaper CMYK Setup" (Eurostandart/newsprint/35% dot gain/light GCR/black limit 85%/total ink 250%) I'm used to follow this procedure.
1) Open the CMYK file
2) Adjust the CMYK Setup to "Premium Printing"
3) Convert it to Lab (or to a wide gammut RGB)
4) Alter the CMYK Setup to "Newspaper"
5) Reconvert to CMYK
6) Fine tune the colors by numbersI've found some of my students ruining this technique by altering the CMYK Setup to "Newspaper" BEFORE CONVERTING THE IMAGE TO LAB, with the results you can predict. And I believe this is the central point in Luca's question.
In the situation you've related, you didn't have to worry about the correct time of altering the CMYK Setup because you only wanted to change the black generation process and the ink limits (mantaining the same ink colors, dot gain and paper settings). This part of CMYK Setup has no influence in "from CMYK to Lab" conversions, only in "from Lab to CMYK" ones. So you didn't have to adjust your CMYK Setup with the Adobe's "goofy black defaults" prior to converting the file to Lab.
On the contrary, the other part of CMYK Setup (ink colors, dot gain and paper settings) does have influence in both directions of the conversion. So, if you wanna change one of its settings, you must pay attention to following the correct order.
I believe it's a good idea to make this difference clear to the guys of the list, do you agree?
Best regards,
Andre Lopes
From: Dan Margulis
Date: Wed, Feb 23, 2000, 10:25 PM
RE: Re: By the numbers, bad shadows
Bob writes:>>I can see why master curve corrections might be a problem for CMYK but why in RGB? Assuming I'm pleased with color balance from highlight to shadow and only want to increase or decrease contrast why would individual channel moves be more appropriate than a global move?>>
A move in the master curve only holds color balance if the colors in question are gray, or highly saturated reds, greens, and blues. Everything else will change. For example, if you attempt to lighten an image overall by attacking the midpoint of the master curve, it will turn any natural greenery more yellow, and any fleshtones more gray. That can be corrected by reverting to Luminosity afterwards, but there isn't much that can be done about the loss of contrast. In the case of the greenery the green channel, and in the case of the fleshtones the red channel, would get hammered by the move. As these are critical channels in their respective objects, this is not a happy situation.
Dan Margulis
From: INTERNET:abcolor@planet.eon.net
Date: Thu, Feb 24, 2000, 12:10 AM
RE: Re: By the numbers, bad shadows
Dan Margulis wrote:> >>But this is very bad advice. It's true that using the master curve isn't
> the guaranteed disaster that it would be in CMYK, but it's almost
> invariably going to cost contrast as opposed to three individual channel
> curves. Those serious about color correction should never be using the
> master curve either in RGB or CMYK.>>Why is it so bad to use the master curve? How does it cost contrast? Isn't it just easier to correct a flat original this way, and still keep the same color balance?
Ron Kelly
From: INTERNET:shane@archangel.net
Date: Thu, Feb 24, 2000, 1:00 AM
RE: Re: By the numbers, bad shadows
Ron Kelly wrote:
> Why is it so bad to use the master curve? How does it cost contrast? Isn't it
> just easier to correct a flat original this way, and still keep the same
> color balance?Hi Ron,
It's not an indictable offense to use the master curve, but correcting in this fashion (especially in CMYK) can cause unexpected side effects, like colour shifts and contrast loss - or gain.
This has to do with the distribution of colour in an image - across three or more separate components; each of which is independent of the others. There's a tendency to get fooled when looking at a monitor display, where we see these channels merged into one apparently holistic image. However, we have to remember that any colour we see is, in fact, three or four layers deep; each component of which acts in different capacities in different parts of the image.
In a CMYK image, for example, Cyan works as a hue contributor in saturated blue tones, but as a hue limiter in saturated red tones. Each of the CMYK channels operates in a sort of inter-dependent relationship with the other three channels. When using a master curve to adjust for contrast, you are affecting every hue differently depending upon the relative strength or weakness of its hue contributors and hue limiters. When using the master curve to add contrast, a saturated red will lose Cyan inversely to it's gain of Yellow and Red; making the reds cleaner and lighter. In a skin tone, where the relationship between hues and contaminants is much more delicate, this could easily lead to reddish "flares" in the skin colour.
This is to say nothing of the Black channel (a special case when we talk about channels) where attempts at contrast gain may lead to super-density in the darkest shadow areas of the image - and detail loss in the areas where Black is beginning to ramp up in its replacement of the other three channels.
I find it a little easier to manage these changes through the use of Levels rather than curves, where it's a little easier to implement general changes across the tonal distribution of any channel. For very specific tweaks, I use Curves. Sometimes, after a series of these adjustments, I'll use the Hue/Saturation controls with feathered selections to fine-tune my control of particular key colours. This allows me to work a little faster with the Curves and/or Levels without agonizing about certain outcomes, although I wouldn't advocate this approach for someone who's just starting out in this field. You definitely have to know when to back-off on certain colour moves!
Well, I've probably said enough to get body slammed, so I'll just sit back and wait for the hate mail to roll in. :)
Best,
--
Shane Steinman
dMACS Project Head
Digital Magazine Advertising Canadian Specifications
c/o ArchAngel Media Inc.
15 Kirkland Court, Richmond Hill, ON L4C 9H4 Canada
Tel/Fax: (416) 410-3355 mailto:shane@dMACS.org
From: Dan Margulis, INTERNET:76270.1033@compuserve.com
Date: Fri, Feb 25, 2000, 7:31 AM
RE: Re: By the numbers, bad shadows
Andre writes:>>1) Open the CMYK file
2) Adjust the CMYK Setup to "Premium Printing"
3) Convert it to Lab (or to a wide gammut RGB)
4) Alter the CMYK Setup to "Newspaper"
5) Reconvert to CMYK
6) Fine tune the colors by numbers>>Oh, it can't be *that* easy! I keep reading that this sort of conversion is impossible in the absence of an ICC workflow. You mean people can *do* this stuff? ;-)
>>I've found some of my students ruining this technique by altering the CMYK Setup to "Newspaper" BEFORE CONVERTING THE IMAGE TO LAB, with the results you can predict. And I believe this is the central point in Luca's question.>>
I've seen the same thing, so you may be right.
>>On the contrary, the other part of CMYK Setup (ink colors, dot gain and paper settings) does have influence in both directions of the conversion. So, if you wanna change one of its settings, you must pay attention to following the correct order.>>
Right.
>>I believe it's a good idea to make this difference clear to the guys of the list, do you agree?>>
Okey doke. In Luca's particular case, it doesn't matter when in the process the maximum black setting gets changed. However, in many other cases involving other components of the CMYK Setup, it would be necessary to change the CMYK Setup *after* going out of CMYK. So, maybe there is an argument that Luca's job should be done that way too, just to develop the right workflow habits.
Dan Margulis
From: Andre B. Lopes
Date: Fri, Feb 25, 2000, 2:31 PM
RE: Re: By the numbers, bad shadows
Dan wrote:>> Oh, it can't be *that* easy! I keep reading that this sort of conversion is impossible in the absence of an ICC workflow. You mean people can *do* this stuff? >>
You can also try to do this using profiles, but doing realiable ICC profiles to the presses here in Brazil requires a lot of sophisticated and expensive devices: spectrophotometres, densitometers and "bafometers" (the brazilian name to brethalyzers). And it's very hard to predict the actual "spirit" of the pressmen (not to say the pressroom temperature, the weather, the side of the paper, etc etc) at the time I'm preparing the files to decide which in a dozen of different ICC profiles is the best one.
More than this, I'm really devasted because I just can't finnish my long term academic work trying to choose which combination of four engines currently installed in my Macintosh (Photoshop's Built-in / Apple ColorSync / Kodak CMM / Heildelberg CMM) plus four intents (Perceptual / Relative / Saturation / Absolute Colorimetric) is the best for doing this stuff. But I'm sure that, if I convince brazilian government to give me a two years scholarship at RIT (to get some kind of PhD in Computer Aided Color Managemment) I would find the answers (at least I will improve my English...). Then I will start a new research to decide when to use or not the black point compensation. ;-)
For the time being, I'm still using this old fashioned and jurassic method.
Tropical regards;
Andre Borges Lopes
Publish Magazine - Brazil
http://www.publish.com.br
From: Andrew Rodney
Date: Fri, Feb 25, 2000, 11:55 PM
RE: Re: Dan-Re: By the numbers, bad shadows
on 2/25/00 11:29 AM, Andre B. Lopes at andrelopes@originet.com.br wrote:> More than this, I'm really devasted because I just can't finnish my long
> term academic work trying to choose which combination of four engines
> currently installed in my Macintosh (Photoshop's Built-in / Apple Color
> Sync / Kodak CMM / Heildelberg CMM) plus four intents (Perceptual /
> Relative / Saturation / Absolute Colorimetric) is the best for doing
> this stuff.When you don't hit a big in one of the CMM's the differences between them are so slight it's not really worth worrying about. I use the Adobe CMM whenever possible (meaning always inside of PS where it's only available) because as yet I don't know anyone who's reported any bugs or problems with it unlike some of the others. And with the others the bugs are rare (like doing RGB to RGB conversions which many users never do).
As for the intents, if you're doing conversions for most output with images, Perceptual is the way to go. Absolute Colorimetric is great for proofing where you need to deal with the paper issues for cross rendering.
Andrew Rodney
From: INTERNET:phaedrusd@aol.com, INTERNET:phaedrusd@aol.com
Date: Fri, Feb 25, 2000, 6:52 PM
Re: By the numbers, bad shadows
>>More than this, I'm really devasted because I just can't finnish my long term academic work trying to choose which combination of four engines currently installed in my Macintosh (Photoshop's Built-in / Apple ColorSync / Kodak CMM / Heildelberg CMM) plus four intents (Perceptual /Relative / Saturation / Absolute Colorimetric) is the best for doing this stuff. But I'm sure that, if I convince brazilian government to give me a two years scholarship at RIT (to get some kind of PhD in Computer Aided Color Managemment) I would find the answers (at least I will improve my English...). Then I will start a new research to decidewhen to use or not the black point compensation. ;-)>>Yes, ICC based Color Management can seem daunting to learn. But then so is doing CMYK by the numbers!
I have clients who I've set up with simple systems that give them really good previews and they don't need to be a CMYK expert, or CM expert at all. It can be easy, or really complex, either way you go. It all depends on your approach and your experience.
If you can't finish your schooling because you can't figure out ICC based Color Management you may need to reexamine your calling, because the "by the numbers" way can be just as confusing.(g)
Dennis Dunbar
Phaedrus Productions
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Fri, Feb 25, 2000, 4:30 PM
RE: Re: By the numbers, bad shadows
>Oh, it can't be *that* easy! I keep reading that this sort of conversion is
>impossible in the absence of an ICC workflow. You mean people can *do* this
>stuff? ;-)You will get more quantization errors with this method since it necessitates two conversions. The converstion to 8-bit/channel Lab, or a wide gamut RGB space larger than Adobe RGB will cause more quantization.
A better way to do this, is do a CMYK to CMYK conversion without involving an intermediate color space. This can be done using the Profile to Profile command where you can specify the From: space (what space the image is in now), and the To: space (what space you want the image in), engine=built-in, intent= perceptual (for going from a larger CMYK space to a smaller CMYK space as in this example going to newsprint; otherwise it would be relative colorimetric), black point compensation ON; hit convert.
You should still go into CMYK Setup and change it to your newsprint setting so the on-screen preview is more accurate.
Chris Murphy
Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.