Ok this question has been bugging me for a while now so I’d like to hear your opinions. When correcting images for print, especially when working in LAB or a large RGB space, I usually pull in the ends of the curve to prevent clipping when going to CMYK (Input 0 & 255, Output 4 & 250 or whatever).
Now my question: Is there anything wrong with using one action-based adjustment layer that does nothing but adjusting the end points and do the rest of the adjustments on an additional layer below or above the first one?
I know I could just do it all in one but I don’t like to mess around in the corners trying not to accidently change the end points. It’s just a cleaner way to have one channel just for details or correction and one just for output, plus the output one has to affect the whole image, while the other one(s) might be only for a part of the image. Do I harm the image more that way, do I miss something?
The miracle of adjustment layers (and they were indeed a miracle when first introduced) is that individual color and tonal adjustments can be separated (i.e., performed separately) for precision, convenience or both WITHOUT compounding pixel damage.
Well, they are flattened one at a time, so there goes the theory of not compounding pixel damage. It’s rarely an issue though. I often use multiple adjustment layers both in RGB and Lab. In Lab it’s very useful to separate the L, a, and b channels on respective layers, letting you adjust individual opacity to suit – at least until you gain a better understanding of how the channels relate.
I agree with Doug and Peter on the issue of multiple adj layers.
However, I have always been mystified by the standard practice of flattening the endpoints when going to CMYK as stated below:
When correcting images for print, especially when working in LAB or a large RGB space, I usually pull in the ends of the curve to prevent clipping when going to CMYK (Input 0 & 255, Output 4 & 250 or whatever).
The problem with a move like that is that any specular highlights in which you DON’T want a dot will absolutely have a dot. And, any areas of the image where you DO want max black will not have max black.
The key is to have a good RGB (or Lab) file prior to conversion. If the RGB (or Lab) file does not exhibit clipping at either end of the curve, the CMYK profile will insure that the CMYK file will have no clipping as well. Some editing may be necessary once in CMYK, but, more often than not, I am usually pushing highlights up and shadows down, not the other way around.
Two Adjustment Layers can, as You implied with »they are flattened one at a time«, indeed do damage to an image that might not become obvious if the same were achieved with one, especially if one works in 8bit. I noticed this in files with lots of overlapping Adjustment Layers or Gradients where banding simply disappeared on conversion to 16bit. The reason being that the values the higher-up Adjustment Layer takes as basis for its changes have been rounded to fit into the rougher »grid« of only 256 steps, I suppose. (That is pretty easily demonstrated if on makes a black to white-gradient in 8bit and applies two extreme adjustments that cancel each other out, like a Curves-layer that moves Input 100% down to Output 30% and a second one that moves Output 100% left to Input 30%; serious banding should be visible; on changing the file to 16bit it should dissapear.)
" for precision, convenience or both WITHOUT compounding pixel damage. " not true as Christoph has said. Effectively Photoshop flattens each layer down in sequence on top of the base, to repeat what others have told you.
So if you can do it one curve, its generally better than using two. Color correctors out there the worst thing you can do it to pile many adjustments on top of an 8 Bit image as this WILL damage its integrity quite seriously. Although having said this its amazing what you seem to be able to get away with!
Wow devastating! I’m glad I asked. So either order will be harmful. What’s better then, going to 16Bit and then flattening or doing it all in one layer?
Over and over I was told that adjustment layers are interacting and there is no harm. I’ve never read about that in any of the 20+Ps books, 100+mags or dozens of learning videos I ran through. This should be mentioned all over the place, in my opinion.
Does the same in 16 bit, its just that you get a lot more room for flexibilty if you move things back and forth. Its still never a good idea to do too much cumulative adjusting, especially curves.
No, most people don’t understand this. Unfortunately too many book authors, only feel they need to know just enough to write their book on the subject, no more. As this and other repeated errors prove.
You can still get away with a surprising quantity of adjustment layers on an 8 bit image. Reference your histogram, and remember – most printing can only with up to 150 levels in an image anyway so ‘damage’ may not be that damaging after all.
This particular tidbit of information regarding flattening, I think, was laid bare by Chris Cox somewhere on this very forum several years ago.
There are, of course, times where two separate adjustments are more effective in improving the image, particularly when you’re narrowly targeting a portion of the tonal range. Usually it’s better to do your overall and then another separate adjustment. Sometimes trying to cram too much into one set of Curves ends up compromising the image too, but Mark is absolutely correct about the fact that you can twist and turn even 8 bit files more than you ever thought possible and never see it in print. And, as I mention often, the entire world of stock is an 8 bit per channel jpeg land that seems to get by just fine.
I am using a process that actually created up to 8 different adjustment layers and half of those are curves. I get much better results than if I single step this and also much better results than if I try to do a lot of this in ACR.
I use CS4 and ACR5.2 and at one point was even going to post the actions I wrote so that it may help others really tweak and edit their photos.
But if the group feels that the answer is not to use adjustment layers in this manner, I will skip posting my techniques and just do what I do and not bring this up. I don’t want to provide something here that will only be challenged and misleading.
I will say this though. I swear by my actions that create my work flow and have gotten results that are outstanding. I will also say that I do not print. All of this is to display in a color managed program. I see no problem with multiple adjustment layers. Maybe because I do not often repeat the part of the curve with two of the adjustment layers is why I get the results I get.
I even use a layer set to 50% gray and use that to burn and dodge and then sometimes I go a step further and adjust levels for that layer to do a global exposure correction. Again, it works for me. That is all I can report. I am happy with my results.
No one has suggested that you shouldn’t use multiple adjustment layers, only that you are aware of what’s happening behind the scenes. I use adjustment layers much in the same manner as you do, and it only helps the image in the end.
I would say that if you’re interested in the utmost quality, there are much better options than ACR out there, especially on the Mac.
Yes, I agree. ACR is better than it was, but I keep saying still far from where it needs to be.
Pfigen, what other options are you referring to?
Take a look at this… (Warning: 1.5M image)
This image is large so I can try to keep intact the image without too much more compression or anything as it is quite busy with lots of pixels for all the small leaves and all…
Part of the problem in this discussion is the idiolectal use of destructive and non-destructive in this context.
What we mean by destructive is nothing more than changing a pixel in an irreversible manner. It doesn’t mean you are actually "destroying" it, nuking, spoiling, desecrating or wasting it.
Obviously, when we manipulate a pixel or a number of pixels, we are actively seeking to change them. Otherwise we wouldn’t be touching them. We do so because we hope to improve or enhance the appearance of a pixel, even if in our parlance that translates into a "destructive" move.
ACR has much worse shadow detail, higher noise and stairstepped diagonals compared to CaptureOne. One images like sunsets, which are particularly taxing to digital sensors, Iridient Digital’s Mac only Raw Developer is unbeatable, giving much smoother near gamut transitions than either C1 or ACR.
I’m starting this year out with a bang, so I don’t know when I’ll be able to post the comparisons, but hopefully in the next couple of weeks.
We hope those comparisons are based on ACR 5.2, which is significantly improved from even 5.1 and light-years ahead of ACR 4.x, let alone earlier versions.
Of course I too work with loads of Adjustment Layers in 8bit-images much of the time. But if You take the time to try the test I described earlier You might understand what I meant and that, even starting with an 8bit-image, changing the layered file to 16bit can improve the quality in problematic files because the rounding effects dont accumulate as steeply. With most corrections on most images it does not make a significant difference, but I was baffled when I first came across an 8bit-file that produced seemingly irremovable banding due to overlapping masked Curves-layers; I tried blurring the masks, adding noise, when the simplest solution was converting to 16bit. Naturally the file-size balloons when going 16bit and I dont want to make anyone change a successful workflow anyhow, but the principle still holds that twice adjusted can differ from once adjusted (if the two adjustments can be achieved in one go that is). Maybe another example will help illustrate my point: Make a vertical black to white gradient, then make two Layer Sets masked to affect the left and right half respectively, in one create a Curves Layer that sets Output 100% left to Input 20%, in the other set achieve the same with two or to make it clearer three Curves Layers. Now the start and end-values are the same but the colors in between, while the same number of steps, occasionally have different values as should be visible at the border between the two masks (or if the gradient uses dither it may appear homogenous with one, but slightly banded with three adjustments). Of course this example is hardly plausible in everyday-work-experience, but the more extreme adjustments are the more prominent these differences can be. If its worth to take this into consideration depends on the image and the adjustments, but I consider it of no disadvantage to know about it.
Thanks for bringing up the subject Greg Apodaca’s Adjustment Layer Analyzer. I downloaded the "demo version" and despaired when it turned out to be an untagged PSD file. Do you have any insight into that?
Also, is there such a thing as a final, non-beta version that you know of? The site offered no easily findable solution.
Crikey Ramon, dont you even know that!? change your color settings. If you still cant open it, then download it again.
Yes its an amazing way to analyze the EFFECTS of adjustments. What Pierre means I think, is that you could us this data to produce composite and potentially less destructive adjustments.
HERE’S a great idea for a feature request Analyse and simplify adjustment layers – allowing you to highlight lets say 20 non-masked adjustment layers and choose a command allowing photoshop to produce a composite adjustment: Maybe a Curve and a Channel Mix that would do its best to reproduce the composite effect of the layers you highlighted.
Much more immediately useful for professionals, and probably a great deal less sophisticated in terms of research, than all the image recognition stuff they have been working on.
Ramón, feel free to build one such wonderful tool yourself, before criticizing him.
I will contact Greg Apodaca, and ask him the reason of the untagging. My two first wild guesses are: to avoid a profile issue, or to warrant the beta status.
He does everything toned down and sized down, probably so everybody can view it without problem. His Photoshop puzzles are super compressed, super annoying but great stuff he’s got. His videos are 300×300 pixels and run in iTunes (album cover panel).
Anyway, so as long as the two channels don’t work in direct opposition, it’s not a big deal to work with several channels on an 8Bit image? Remember, I wanted to know specifically if the workflow with a layer, just to pull in the ends, was ok.
It was created back in the days of PS6 Ramon, when people were still complaining about "take my color management back to Photoshop 5" – and Yes its an a amazingly clever tool, to even work through how that was created is far beyond most people. And Like Pierre says, if you want to CREATE profiled versions of that file, you are welcome.
Certainly for compressing bad files made of many cumulative curves into one, its unbeatable. I remember using the CMYK version back in the days when I worked in a studio full of ‘dated layer set’ color correctors. Does the job.
T., my earlier remarks focused to some extent on the principle of the matter. But judging from my work-experience in most instances 8bit-photographic images will show no significant deterioration when corrections of not too extreme a nature are applied as multiple Adjustment Layers, so generally no big deal. (But deterioration does occur, even if imperceptibly, and the more of the corrections can be achieved either in the RAW-file, if available, or utilizing as few Adjustment Layers as possible the better.)
In an earlier post, I was not suggesting multiple adjustment layers acted as a single adjustment. As someone else said, those layers are "applied"(when the file is flattened or printed) individually… sequentially. So indeed they ultimately are no different than applying a curve directly to pixels, then applying a second curve in a second, separate move, directly to pixels. Successive deterioration in both cases.
But wouldn’t we agree (or wouldn’t we!) that T. Schmidt’s original scenario is by no means extreme or excessive. The adjustments he describes would certainly not damage unacceptably (and probably not even perceptibly) an already reasonably balanced, reasonably pixel rich, high resolution image? That the trouble would begin to be noticeable only when the moves are considerably more radical than he described and/or when they’re applied to a crappy image?
