If you take an RGB triad and apply the same non-linear curve to the R channel, G channel and B channel, unless R=G=B, each color will fall in a different part of the curve. Even when all three curves are identical, the result will curves will be to shift the hue of some (but not all) colors. I'll see if I can put together a simple PSD file which demonstrates this effect.
If you take an RGB triad and apply the same non-linear curve to the R channel, G channel and B channel, unless R=G=B, each color will fall in a different part of the curve. Even when all three curves are identical, the result will curves will be to shift the hue of some (but not all) colors. I'll see if I can put together a simple PSD file which demonstrates this effect.
OK I thought that was what you might be refrerring to ala Dan but the term artifact threw me.
You realize that this behavior is done by design and in fact takes more code and work on the part of the Adobe engineers to do?
This is a quote from Mark Hamburg on ACR/LR curves:
Question. Does ACR/LR use a RGB "master curve" (composite curve) when applying brightness/contrast or the curves? Or is it working on luminance data?
Answer: Neither actually. Lightroom and Camera Raw preserve hue but adjust saturation to match with contrast changes because having saturation and contrast get out of sync tends to look rather odd.
Thanks for your patience on this. I have found this thread to be quite illuminating.
If I had an ideal set-up, I would probably adopt many more of your recommendations than I now follow. I am working on a four year old laptop with an OK 17" screen and 2 gigs of memory. I'm using PS CS1 for editing.
With this setup, 16 bit editing is basically out of the question. I would tear my hair out waiting for re-draws of the images. For similar reasons, opening a bunch of files to do skin tone comparisons is also out of the question. My machine cludges along as it is.
Similarly, calibration is only useful to some degree for me. Change the angle of the laptop screen a degree or two, or simply slouch a bit, and alot of the benefit of calibration plunges down the drain. This doesn't even get into the effect that changing light has on a laptop screen.
Someday, I may have the money (and convince my wife to spend it) so that I can have basically unlimited performance and storage. Today isn't that day, so I try to live with the best compromises I can afford.
From what you wrote, one of the main advantages of working with and preserving layers is for when you come back to a file after a long period of time. My attitude about that is that I am still learning so much that the main thing I would learn going back to a file in a year is how stupid I was a year ago. So I'm content to start from scratch.
Think about this. Suppose you had a RAW file that you processed with some early version of ACR and then polished up in PS a couple of years ago. You might now want to go back and do some stuff with the file, but it might be better and quicker just to go back to the RAW, process it with the new ACR or Lightroom, and finish it up with whatever your new processes and knowledge have taught you in the meantime. Since I'm still very much on the steep part of the learning curve, it seems pretty obvious to me that I could do a better job faster on just about any picture I took a year or a year and a half ago.
Thus, preserving my work simply isn't a priority for me. In principle, I'd like to know how, and I intend to learn how to do this easily at some point. But right now, everything I do is just a bunch of compromises.
With this setup, 16 bit editing is basically out of the question. I would tear my hair out waiting for re-draws of the images.
You youngsters <g>. I recall, back in 1990, Photoshop 1.0.7 on a Mac Iici, 16 megs of ram, I had to rotate a 25mb image I shot 1 ½ degrees because the 4x5 wasn't perfectly level. It took 22 minutes to process! Fun.
I actually started my conversations with Adobe (to later become an alpha and beta tester) by recommending Beep when done. Just took too much effort to keep looking up from reading material to see if the progress dialog was dismissed.
Anyway, I understand your point. But trust me, whatever you're doing, it ain't that slow!
Similarly, calibration is only useful to some degree for me. Change the angle of the laptop screen a degree or two, or simply slouch a bit, and alot of the benefit of calibration plunges down the drain. This doesn't even get into the effect that changing light has on a laptop screen.
Don't have that problem with my Sony Artisan's. Yup, I'm not a fan of LCD's for color critical work. They are getting better. Once the Fluorescent lights are gone, they will be a lot better. Actually, viewing angles are getting pretty good these days.
Think about this. Suppose you had a RAW file that you processed with some early version of ACR and then polished up in PS a couple of years ago. You might now want to go back and do some stuff with the file, but it might be better and quicker just to go back to the RAW, process it with the new ACR or Lightroom, and finish it up with whatever your new processes and knowledge have taught you in the meantime.
The beauty of raw is it's just a data source. As the converters get better, its actually useful to revisit the rendering from square one because you can produce vastly superior results. What's really cool about Lightroom is the edit list that remains intact after you quit the application. Or the ability to build variations, multiple rendering instructions.
OK I thought that was what you might be refrerring to ala Dan but the term artifact threw me.
You realize that this behavior is done by design and in fact takes more code and work on the part of the Adobe engineers to do?
This is a quote from Mark Hamburg on ACR/LR curves:
Question. Does ACR/LR use a RGB "master curve" (composite curve) when applying brightness/contrast or the curves? Or is it working on luminance data?
Answer: Neither actually. Lightroom and Camera Raw preserve hue but adjust saturation to match with contrast changes because having saturation and contrast get out of sync tends to look rather odd.
Oh interesting. I wasn't aware that the ACR/Lightroom curves worked that way. I know I have ended up with some wacky colors and oversaturated images when I have started with low contrast orginals. However, it is possible that the black point and exposure controls were at fault rather than the curves. I love the idea of a Lightroom only workflow, and I certainly go that route for family snap shots and the like because of the efficiency. However when I really want to make an image look its best I have had a hard time getting there without Photoshop. With the addition of the Clarity control and the improved sharpening, there does seem to be the potential to go that route with more images.
As for increasing contrast without increasing saturation, he is right. It often does look strange. However, I don't always agree with Lightroom on how much to increase the saturation for a particular curve and image. The saturation and vibrance sliders are there to help with this problem but so far I have not found the results to be completely satisfying. I'll have to toy around in Lightroom to see if I can figure out how to duplicate the net effect of my standard LAB workflow.
Don't have that problem with my Sony Artisan's. Yup, I'm not a fan of LCD's for color critical work. They are getting better. Once the Fluorescent lights are gone, they will be a lot better. Actually, viewing angles are getting pretty good these days.
Have you seen the new Lacie 526, not sure if it is even in production? I am wondering how it compares to the Artisan.
For similar reasons, opening a bunch of files to do skin tone comparisons is also out of the question. My machine cludges along as it is.
One thing I have considered doing is creating a small file with a bunch of color patches of reasonable skin tones to use as a reference. Something akin to that might put a smaller load on your machine than opening a full image.
Don't have that problem with my Sony Artisan's. Yup, I'm not a fan of LCD's for color critical work. They are getting better. Once the Fluorescent lights are gone, they will be a lot better.
Indeed. I use a cheap $200 ViewSonic CRT and for critical work I prefer it to any LCD I am willing to pay for. I have found that if I am preparing an image for the web that I do need to view it on my LCD at work so I know what most people are (or more to the point, aren't) seeing.
Oh interesting. I wasn't aware that the ACR/Lightroom curves worked that way. I know I have ended up with some wacky colors and oversaturated images when I have started with low contrast orginals. However, it is possible that the black point and exposure controls were at fault rather than the curves.
That's why generally its best to work with the tools in the order presented (top down, left to right). There's a fixed processing order by design. That's one reason you see two histograms that look different. The one in Curves is based on the data after using the basic controls. You can of course work out of order but you might end up chasing your tail.
I love the idea of a Lightroom only workflow, and I certainly go that route for family snap shots and the like because of the efficiency. However when I really want to make an image look its best I have had a hard time getting there without Photoshop.
Its important when talking about raw processing to look at the vast differences in the two tools and the data being affected. I do all the heavy lifting (tone and color) in LR. its faster, its operating on a very different kind of data than baked, colored pixels in Photoshop. Its all metadata instruction being applied to a data source. There are operations that simply can't be duplicated in Photoshop on gamma corrected images (like WB, highlight recover etc).
There's is still areas that need addressing such that I'd spend even less time in Photoshop such as soft proofing and output sharpening (capture sharpening in 1.1/4.1 is very nice). I still need to move into Photoshop to do output sharpening and soft proofing at the very least and of course precise pixel editing. The healing brush is awesome when you need to remove a dust spot on 100 images but precise cloning? Nope, you need Photoshop which is a true pixel editor.
Look I have a pretty good road map for what's coming in LR (and no, I'm not talking <g>). We'll all be using Photoshop but a lot of work can and I would submit should be done from raw data in LR, ACR or another processor (if you're on the Mac, Raw Developer is also a very good package although it doesn't handle a fraction of the other functionally of LR and the other modules). Printing out of Photoshop is a drag, a great experience in LR.
With the addition of the Clarity control and the improved sharpening, there does seem to be the potential to go that route with more images.
Indeed. Clarity was something I had to do in Photoshop on a routine basis and don't anymore. That's originally a wonderful technique that came from Mac Holbert at Nash editions for doing Midtone contrast tweaking. You can do it in Photoshop easily (I have an action that does it) but now I just use Clarity in LR.
Because for skin tones CMYK and LAB both provide an understandable relationship that can provide an important guideline as to whether you're on target or not. I've never found an RGB equivalent. If you can provide me with one, I'd be ecstatic. But so far, only CMYK and LAB can do this. I know that Y should be a few point greater than M, and that C should be a fraction of either of them. Or, in LAB, I know that B should be more than A. They're not hard and fast rules, but it's a great aid in making sure that your skin tones are correct. And this cannot be replicated in RGB.
In RGB, you want G >= B with both a fraction of R. It's really just the inverse of the CMYK guideline you are used to. If G >= B, then M <= Y and A <= B in LAB. At this very gross level of approximation (and that's really all this is) CMYK is RGB. That's very different from the fine distinctions Andrew wants to make. These guideline are just to get a rough idea whether things can be improved.
Brand new PDF just appeared on the Adobe Site today by Dr. Karl Lang (he designed the Radius Pressview then the Sony Artisan reference display). He's a very good tech editor too.
In RGB, you want G >= B with both a fraction of R. It's really just the inverse of the CMYK guideline you are used to. If G >= B, then M <= Y and A <= B in LAB. At this very gross level of approximation (and that's really all this is) CMYK is RGB. That's very different from the fine distinctions Andrew wants to make. These guideline are just to get a rough idea whether things can be improved.
Thanks John, I have briefly covered this in another thread here, but I may as well take the time to "flesh out" the point again in this topic (sorry about the pun, it was intended).
I entered digital imaging before Photoshop was king (originally as a compositor and then into general prepress), way back in the time of proprietary prepress colour systems and drum scanning, before the "desktop publishing" revolution. Since I "grew up" in the subractive colour world of ink on paper, I am very comfortable evaluating colour in SWOP TR001 or similar CMY(K) - but like DavidTO and many others, find doing the same thing in the additive RGB mode harder. Perhpas for different reasons, David may not have a prepress background. The subtractive mixing of colour is understood by children when painting, with a little experience. Often photographers who view the world in light find RGB colour mixing and evaluation more intuitive.
As Rutt correctly points out, in "theory" RGB is the inverse of CMY, so it should be no big deal to think "opposite". As printing inks are not pure, they are not simply the exact inverse of RGB values in a synthetic RGB working space in Photoshop...but we are not talking of such absolutes here, just a general ratio. For the task at hand, the general rule still applies with a little variation and adjustment, depending on the RGB working space at hand (just as the CMYK numbers depend on the 4C print process they are targeted to).
Nobody is saying that there is a magical and correct CMYK recipe for "correct" skin, or other "memory colours" such as blue sky or green foliage. What has been found over the decades in print production is that there is a ratio of cyan to magenta and yellow that produces a pleasing result on press. Most often the CMY ratio is presumed to be for a web press or other common offset print condition with a similar gray balance. The reason that such numerical evaluation and adjustment was necessary was due to the technology of the time. Drum scanning used to be performed with no preview on a monitor of the scanned image, but they could sample colour readings from areas of the image to evaluate what was taking place with the scan and the on the fly conversion to the chosen CMYK settings and to make adjustments if needed. These adjustments or "corrections" were not simply chosen at random, a drum scanner operator used to serve a four year apprenticeship learning the craft.
Today with RGB originals and Photoshop, one does not have to evaluate colour in CMYK mode. As has been noted in this thread, there are other colour models that may serve the purpose better. Andrew mentioned his preference for LCH that was made popular in old LinoColor scanning software - while John, David and others have mentioned that LAB mode evaluation in Photoshop via info palette is a similar and viable option that they employ. HSB mode could also be used for skin, but it should not be used for evaluation of neutral shadows.
Obviously, I find the info palette and colour sampler tool in Photoshop to be a critical part of my workflow. Another important palette is the color palette. One can sample an image colour and view the ratio of the various channels in different colour modes via the position of the sliders. The sliders indicate a pleasing ratio for a given colour based on their position or the "shape" that they make.
The basic evaluation method that the prepress industry came up with decades ago is to sample a representative area and find the magenta value. This establishes the basic level in the skin. The yellow value is usually at least equal to and slightly higher than the magenta value, perhaps 10-30%, one now has a hue change from light pink to light red. The cyan value is the critical one and seems to be missed when visual evaluation only is performed, even if on a calibrated/profiled/colour managed monitor and software etc. The basic guideline is that the cyan should be anywhere from 1/8 to 1/3 the magenta value (higher cyan with higher magenta and yellow for a good strong tan, faces and children will vary this guideline). The cyan kills the red and gives the skin hue a realistic appearance and depth. As has been noted by many, skin being too red in print (press, inkjet) is more often due to a poor red/cyan channel than the green/magenta and blue/yellow being the main problem, although they may need tuning as well.
To illustrate this point, I have attached an animated GIF image to this post. The old PDI test image has been used to provide two common samples of skin in "normal" lighting. This image has not been "corrected" or "enhanced". This is the original colour. The posted image is sRGB, but the original colour values illustrated are Adobe RGB and SWOP v2 CMYK (TR001), in addition to LAB. In all cases, the same basic ratio can be found, as indicated in the animation. With such a simple rule/ratio to follow, one can be confident that the reproduction of skin will be pleasing, where as in the past one may not have been sure of this.
As many people have stated - using colour managed monitor and the info palette is the best approach. The info palette can be used for evaluation of memory colours and not just endpoints/neutrals, although some only advise the later and not memory colours.
One can read about memory colours in this short extract/review:
In RGB, you want G >= B with both a fraction of R. It's really just the inverse of the CMYK guideline you are used to. If G >= B, then M <= Y and A <= B in LAB. At this very gross level of approximation (and that's really all this is) CMYK is RGB. That's very different from the fine distinctions Andrew wants to make. These guideline are just to get a rough idea whether things can be improved.
So, by that logic, I can keep doing what I'm doing...
These are the two skin samples in question. The old CMY formula traditionally used by prepress is verified by modern capture and colour conversion. As the previous animation indicates, color palette evaluation in RGB, CMYK or LAB mode will result in similar indicators of pleasing skin tone.
In this iTunes/QuickTime podcast video for Adobe Photoshop Lightroom, there is a visual method of adjusting colour - which I presume is similar to what Andrew Rodney suggests, rather than more extensive use of the info palette that John and others have gained appreciation for from books from Dan Margulis. Granted, this example of editing is more subjective than targeted, which is the point I guess!
As many people have stated - using colour managed monitor and the info palette is the best approach. The info palette can be used for evaluation of memory colours and not just endpoints/neutrals, although some only advise the later and not memory colours.
Well said Stephen. Thanks for taking the time to write this.
And that PDI fleshtone target can take on a tinge of orangey to pinkish yellow then to rust, to maroonish brown just by taking a bathroom break. The blue background doesn't make it any easier as well. Sheer torture.
And that PDI fleshtone target can take on a tinge of orangey to pinkish yellow then to rust, to maroonish brown just by taking a bathroom break. The blue background doesn't make it any easier as well. Sheer torture.
I hate that document and wish people would instead use better examples of possible color scenes. Bill Atkinson has some excellent targets on his site.
In this iTunes/QuickTime podcast video for Adobe Photoshop Lightroom, there is a visual method of adjusting colour - which I presume is similar to what Andrew Rodney suggests, rather than more extensive use of the info palette that John and others have gained appreciation for from books from Dan Margulis. Granted, this example of editing is more subjective than targeted, which is the point I guess!
I watched that video and was amused by the number of times he used the word "correction," sometimes correcting himself about his own "misuse" of the word. One thing that he doesn't seem to understand is that the verb "correct" can mean to make something more accurate, or sometimes it simply means to adjust something. Thats just a small semantic point.
The idea underlying it bothers me some. Some people seem to think that just because a process doesn't lend itself to accuracy that can be measured by a machine, then it must be "totally subjective." I think this is a mistake. There are lots and lots of areas of human activity that are neither scientifically measurable nor totally subjective. To a large extent, color correction is one of these areas. There may be no single right way to correct a photo, but there are many, many wrong ways to do it. Right and wrong here are matters for agreement, not for measurement.
Recently, we also discovered that reading an election ballot was one of these areas that lies in the middle. No machine could say whether a hanging chad was a vote, but that did not make the process of reviewing the ballots "totally subjective" either. My guess is that there are many, many things we do every day that fall in this middle area, and our sometimes sloppy words actually work well in these sloppy domains.
I hate that document and wish people would instead use better examples of possible color scenes. Bill Atkinson has some excellent targets on his site.
One common memory colour that is often critical is pleasing reproduction of skin. I doubt that there will be much argument from the folk here at DG on this point.
The Atkinson composite image test does not include any new skintones (the LAB image, right?) - the only skin featured is a reduced version of the PDI chart that is being maligned!
So, It would appear that the theory is not being dismissed, just the reference image? OK, then...
So, let's try a different reference image for skin tone, the replacement image for the old PhotoDisc, now the Getty Images Test Image:
The Caucasian lady to the left, an average skin reading in CMYK (Adobe SWOP v2 Web Coated TR001) and RGB (Adobe RGB 1998) is:
16c 47m 51y 0k
191r 141g 110b
65L +23a +28b
Once again, the old prepress CMY ratio and the Photoshop Color palette slider positions in RGB and LAB show a helpful relationship as a guide to pleasing colour. The RGB skin is not out of gamut for SWOP v2 CMYK. Again the image was not edited by me to conform to the legacy target values suggested by prepress tradition, the original colour was captured/presented this way by the test target creator as an "accurate" rendition of the photographed scene. That human observers and modern capture/colour conversion agrees with the old prepress general ratio is not surprising to me.
It really puzzles me why some are so resistant to admit the benefit of taking colorimetric readings inside Photoshop using the info palette and various colour reading modes, for all areas of the image and not just endpoints and known/presumed neutrals. These same people often advise similar measurement based methods (using hardware & software not built into Photoshop) for calibrating and profiling monitors and inkjet printers - but seem to reject similar methods of evaluating images and image edits inside Photoshop using the built in tools provided.
The one thing that can throw a monkey wrench into using the numbers for fleshtones is if the subject has that pinky peach foundation makeup. It can imbue an overly vibrant appearance due to its spectral reflectance qualities causing cyan and yellow numbers to get thrown off.
I have never edited any image for output to a commercial press, but I can see how difficult it must be. I flipped through one of Scott Kelby's books illustrating a skintone edit tip in the bookstore one day. It was the one of a close-up on a woman's face wearing sunglasses and noticed the before picture looked better than the corrected shot which made her look jaundice. Maybe it was press drift. Not sure. The tutorial did show the CMYK readouts and I couldn't tell if the amount of yellow was too much or not.
There is one color target where I could actually confirm CMYK readouts as being predictable as they should appear on press and that was viewing a print of the untagged Ole No Moire color target found on PS install CD's printed from Copy Craft's sheetfed Komori waterless press at 300lpi.
This press uses a closed loop color check system using a constantly scanning spectrophotometer that automatically adjusts for press drift, so the CMYK dot percentage sizes in the PS readouts reads exactly that on the print as viewed under a magnifying glass. But the print still showed overall gain in the cyan channel in neutrals, but the fleshtones were spot on.
Also this untagged CMYK target looks way too light when assigning PS's standard coated SWOP v2 so I assigned the sheetfed v2 and the image looked much more closer to the Komori print. See attached scan of the print next to the assigned sheetfed profile on the right.
The problem I've always had viewing dot sizes off press runs is you can't see the yellow because it's too light. Doh!
The one thing that can throw a monkey wrench into using the numbers for fleshtones is if the subject has that pinky peach foundation makeup.
Agreed Tim, as mentioned earlier, children and faces change things a lot. Ideally one can sample many areas and take it all with a grain of salt, remembering that these are just guidelines or they can become more specific and targeted to output if this is a "correction" to a "known" problem rather than just a subjective general edit in an intermediate working space.
I just checked the famous "musicians" scene of the Caucasian, Asian and Negro women with their instruments and ethnic dress, that was originally presented as untagged CMYK. I averaged 10 common press spaces of this image into an ECI-RGB file and an average non facial area of the Caucasian lady on the left is:
Again, a similar pleasing ratio is reflected in the color palette via the position of the CMYK/RGB/LAB sliders, as shown in the animation on page 8 of this thread. Note the difference in colour and tone from the Getty image, both ladies are very different and so is the lighting - but both are acceptable (one could argue that the Getty woman is too dark and magenta or the musician too yellow and or not magenta enough etc, this is the subjective nature of such evaluations).
Did you get my email requesting the link to that Blue Jacket CMYK gamut edit tutorial I used to get off your site but now have lost since you redesigned your site?
I really thought that was such an informative editing tip.
Did you get my email requesting the link to that Blue Jacket CMYK gamut edit tutorial I used to get off your site but now have lost since you redesigned your site?
I really thought that was such an informative editing tip.
Hi Tim and thanks for the interest, yes I did get your email about the page that is offline with the new site redesign.
I will email you offlist tonight when I have access to email again and not just the web.
Hey, Andrew, we hear a lot about your theories of image improvement, but we never see any of your images. You once did post a link to an image for me on a different forum, and I loved it. Why not show us some of your work on dgrin? Maybe even with before/after versions? After all, the proof of the pudding...
Comments
If you take an RGB triad and apply the same non-linear curve to the R channel, G channel and B channel, unless R=G=B, each color will fall in a different part of the curve. Even when all three curves are identical, the result will curves will be to shift the hue of some (but not all) colors. I'll see if I can put together a simple PSD file which demonstrates this effect.
OK I thought that was what you might be refrerring to ala Dan but the term artifact threw me.
You realize that this behavior is done by design and in fact takes more code and work on the part of the Adobe engineers to do?
This is a quote from Mark Hamburg on ACR/LR curves:
Question. Does ACR/LR use a RGB "master curve" (composite curve) when applying brightness/contrast or the curves? Or is it working on luminance data?
Answer: Neither actually. Lightroom and Camera Raw preserve hue but adjust saturation to match with contrast changes because having saturation and contrast get out of sync tends to look rather odd.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Thanks for your patience on this. I have found this thread to be quite illuminating.
If I had an ideal set-up, I would probably adopt many more of your recommendations than I now follow. I am working on a four year old laptop with an OK 17" screen and 2 gigs of memory. I'm using PS CS1 for editing.
With this setup, 16 bit editing is basically out of the question. I would tear my hair out waiting for re-draws of the images. For similar reasons, opening a bunch of files to do skin tone comparisons is also out of the question. My machine cludges along as it is.
Similarly, calibration is only useful to some degree for me. Change the angle of the laptop screen a degree or two, or simply slouch a bit, and alot of the benefit of calibration plunges down the drain. This doesn't even get into the effect that changing light has on a laptop screen.
Someday, I may have the money (and convince my wife to spend it) so that I can have basically unlimited performance and storage. Today isn't that day, so I try to live with the best compromises I can afford.
From what you wrote, one of the main advantages of working with and preserving layers is for when you come back to a file after a long period of time. My attitude about that is that I am still learning so much that the main thing I would learn going back to a file in a year is how stupid I was a year ago. So I'm content to start from scratch.
Think about this. Suppose you had a RAW file that you processed with some early version of ACR and then polished up in PS a couple of years ago. You might now want to go back and do some stuff with the file, but it might be better and quicker just to go back to the RAW, process it with the new ACR or Lightroom, and finish it up with whatever your new processes and knowledge have taught you in the meantime. Since I'm still very much on the steep part of the learning curve, it seems pretty obvious to me that I could do a better job faster on just about any picture I took a year or a year and a half ago.
Thus, preserving my work simply isn't a priority for me. In principle, I'd like to know how, and I intend to learn how to do this easily at some point. But right now, everything I do is just a bunch of compromises.
Duffy
You youngsters <g>. I recall, back in 1990, Photoshop 1.0.7 on a Mac Iici, 16 megs of ram, I had to rotate a 25mb image I shot 1 ½ degrees because the 4x5 wasn't perfectly level. It took 22 minutes to process! Fun.
I actually started my conversations with Adobe (to later become an alpha and beta tester) by recommending Beep when done. Just took too much effort to keep looking up from reading material to see if the progress dialog was dismissed.
Anyway, I understand your point. But trust me, whatever you're doing, it ain't that slow!
Don't have that problem with my Sony Artisan's. Yup, I'm not a fan of LCD's for color critical work. They are getting better. Once the Fluorescent lights are gone, they will be a lot better. Actually, viewing angles are getting pretty good these days.
The beauty of raw is it's just a data source. As the converters get better, its actually useful to revisit the rendering from square one because you can produce vastly superior results. What's really cool about Lightroom is the edit list that remains intact after you quit the application. Or the ability to build variations, multiple rendering instructions.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Oh interesting. I wasn't aware that the ACR/Lightroom curves worked that way. I know I have ended up with some wacky colors and oversaturated images when I have started with low contrast orginals. However, it is possible that the black point and exposure controls were at fault rather than the curves. I love the idea of a Lightroom only workflow, and I certainly go that route for family snap shots and the like because of the efficiency. However when I really want to make an image look its best I have had a hard time getting there without Photoshop. With the addition of the Clarity control and the improved sharpening, there does seem to be the potential to go that route with more images.
As for increasing contrast without increasing saturation, he is right. It often does look strange. However, I don't always agree with Lightroom on how much to increase the saturation for a particular curve and image. The saturation and vibrance sliders are there to help with this problem but so far I have not found the results to be completely satisfying. I'll have to toy around in Lightroom to see if I can figure out how to duplicate the net effect of my standard LAB workflow.
Have you seen the new Lacie 526, not sure if it is even in production? I am wondering how it compares to the Artisan.
One thing I have considered doing is creating a small file with a bunch of color patches of reasonable skin tones to use as a reference. Something akin to that might put a smaller load on your machine than opening a full image.
Indeed. I use a cheap $200 ViewSonic CRT and for critical work I prefer it to any LCD I am willing to pay for. I have found that if I am preparing an image for the web that I do need to view it on my LCD at work so I know what most people are (or more to the point, aren't) seeing.
That's why generally its best to work with the tools in the order presented (top down, left to right). There's a fixed processing order by design. That's one reason you see two histograms that look different. The one in Curves is based on the data after using the basic controls. You can of course work out of order but you might end up chasing your tail.
Its important when talking about raw processing to look at the vast differences in the two tools and the data being affected. I do all the heavy lifting (tone and color) in LR. its faster, its operating on a very different kind of data than baked, colored pixels in Photoshop. Its all metadata instruction being applied to a data source. There are operations that simply can't be duplicated in Photoshop on gamma corrected images (like WB, highlight recover etc).
There's is still areas that need addressing such that I'd spend even less time in Photoshop such as soft proofing and output sharpening (capture sharpening in 1.1/4.1 is very nice). I still need to move into Photoshop to do output sharpening and soft proofing at the very least and of course precise pixel editing. The healing brush is awesome when you need to remove a dust spot on 100 images but precise cloning? Nope, you need Photoshop which is a true pixel editor.
Look I have a pretty good road map for what's coming in LR (and no, I'm not talking <g>). We'll all be using Photoshop but a lot of work can and I would submit should be done from raw data in LR, ACR or another processor (if you're on the Mac, Raw Developer is also a very good package although it doesn't handle a fraction of the other functionally of LR and the other modules). Printing out of Photoshop is a drag, a great experience in LR.
Indeed. Clarity was something I had to do in Photoshop on a routine basis and don't anymore. That's originally a wonderful technique that came from Mac Holbert at Nash editions for doing Midtone contrast tweaking. You can do it in Photoshop easily (I have an action that does it) but now I just use Clarity in LR.
Author "Color Management for Photographers"
http://www.digitaldog.net/
One is on it's way for review. It's actually an NEC product that LaCie OEM's.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Will NEC sell it also or only via LaCie?
Hope you receive it soon, looking to hear your opinion of it.
I expect NEC will sell it too. I'll ask.
Author "Color Management for Photographers"
http://www.digitaldog.net/
In RGB, you want G >= B with both a fraction of R. It's really just the inverse of the CMYK guideline you are used to. If G >= B, then M <= Y and A <= B in LAB. At this very gross level of approximation (and that's really all this is) CMYK is RGB. That's very different from the fine distinctions Andrew wants to make. These guideline are just to get a rough idea whether things can be improved.
http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf
Read it, then consider what I've suggested about scene versus output referred, accurate color, and so forth.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Thanks John, I have briefly covered this in another thread here, but I may as well take the time to "flesh out" the point again in this topic (sorry about the pun, it was intended).
I entered digital imaging before Photoshop was king (originally as a compositor and then into general prepress), way back in the time of proprietary prepress colour systems and drum scanning, before the "desktop publishing" revolution. Since I "grew up" in the subractive colour world of ink on paper, I am very comfortable evaluating colour in SWOP TR001 or similar CMY(K) - but like DavidTO and many others, find doing the same thing in the additive RGB mode harder. Perhpas for different reasons, David may not have a prepress background. The subtractive mixing of colour is understood by children when painting, with a little experience. Often photographers who view the world in light find RGB colour mixing and evaluation more intuitive.
As Rutt correctly points out, in "theory" RGB is the inverse of CMY, so it should be no big deal to think "opposite". As printing inks are not pure, they are not simply the exact inverse of RGB values in a synthetic RGB working space in Photoshop...but we are not talking of such absolutes here, just a general ratio. For the task at hand, the general rule still applies with a little variation and adjustment, depending on the RGB working space at hand (just as the CMYK numbers depend on the 4C print process they are targeted to).
Nobody is saying that there is a magical and correct CMYK recipe for "correct" skin, or other "memory colours" such as blue sky or green foliage. What has been found over the decades in print production is that there is a ratio of cyan to magenta and yellow that produces a pleasing result on press. Most often the CMY ratio is presumed to be for a web press or other common offset print condition with a similar gray balance. The reason that such numerical evaluation and adjustment was necessary was due to the technology of the time. Drum scanning used to be performed with no preview on a monitor of the scanned image, but they could sample colour readings from areas of the image to evaluate what was taking place with the scan and the on the fly conversion to the chosen CMYK settings and to make adjustments if needed. These adjustments or "corrections" were not simply chosen at random, a drum scanner operator used to serve a four year apprenticeship learning the craft.
Today with RGB originals and Photoshop, one does not have to evaluate colour in CMYK mode. As has been noted in this thread, there are other colour models that may serve the purpose better. Andrew mentioned his preference for LCH that was made popular in old LinoColor scanning software - while John, David and others have mentioned that LAB mode evaluation in Photoshop via info palette is a similar and viable option that they employ. HSB mode could also be used for skin, but it should not be used for evaluation of neutral shadows.
Obviously, I find the info palette and colour sampler tool in Photoshop to be a critical part of my workflow. Another important palette is the color palette. One can sample an image colour and view the ratio of the various channels in different colour modes via the position of the sliders. The sliders indicate a pleasing ratio for a given colour based on their position or the "shape" that they make.
The basic evaluation method that the prepress industry came up with decades ago is to sample a representative area and find the magenta value. This establishes the basic level in the skin. The yellow value is usually at least equal to and slightly higher than the magenta value, perhaps 10-30%, one now has a hue change from light pink to light red. The cyan value is the critical one and seems to be missed when visual evaluation only is performed, even if on a calibrated/profiled/colour managed monitor and software etc. The basic guideline is that the cyan should be anywhere from 1/8 to 1/3 the magenta value (higher cyan with higher magenta and yellow for a good strong tan, faces and children will vary this guideline). The cyan kills the red and gives the skin hue a realistic appearance and depth. As has been noted by many, skin being too red in print (press, inkjet) is more often due to a poor red/cyan channel than the green/magenta and blue/yellow being the main problem, although they may need tuning as well.
To illustrate this point, I have attached an animated GIF image to this post. The old PDI test image has been used to provide two common samples of skin in "normal" lighting. This image has not been "corrected" or "enhanced". This is the original colour. The posted image is sRGB, but the original colour values illustrated are Adobe RGB and SWOP v2 CMYK (TR001), in addition to LAB. In all cases, the same basic ratio can be found, as indicated in the animation. With such a simple rule/ratio to follow, one can be confident that the reproduction of skin will be pleasing, where as in the past one may not have been sure of this.
As many people have stated - using colour managed monitor and the info palette is the best approach. The info palette can be used for evaluation of memory colours and not just endpoints/neutrals, although some only advise the later and not memory colours.
One can read about memory colours in this short extract/review:
http://www.panix.com/%7Erbean/color/color4.txt
Sincerely,
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
So, by that logic, I can keep doing what I'm doing...
Dgrin FAQ | Me | Workshops
All colour spaces are one, as somebody once said.
Regards,
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
http://av.adobe.com/russellbrown/20070614_Lightroom_Tutorial_Podcast.mp4
Regards,
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
Well said Stephen. Thanks for taking the time to write this.
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
And that PDI fleshtone target can take on a tinge of orangey to pinkish yellow then to rust, to maroonish brown just by taking a bathroom break. The blue background doesn't make it any easier as well. Sheer torture.
I hate that document and wish people would instead use better examples of possible color scenes. Bill Atkinson has some excellent targets on his site.
Author "Color Management for Photographers"
http://www.digitaldog.net/
I watched that video and was amused by the number of times he used the word "correction," sometimes correcting himself about his own "misuse" of the word. One thing that he doesn't seem to understand is that the verb "correct" can mean to make something more accurate, or sometimes it simply means to adjust something. Thats just a small semantic point.
The idea underlying it bothers me some. Some people seem to think that just because a process doesn't lend itself to accuracy that can be measured by a machine, then it must be "totally subjective." I think this is a mistake. There are lots and lots of areas of human activity that are neither scientifically measurable nor totally subjective. To a large extent, color correction is one of these areas. There may be no single right way to correct a photo, but there are many, many wrong ways to do it. Right and wrong here are matters for agreement, not for measurement.
Recently, we also discovered that reading an election ballot was one of these areas that lies in the middle. No machine could say whether a hanging chad was a vote, but that did not make the process of reviewing the ballots "totally subjective" either. My guess is that there are many, many things we do every day that fall in this middle area, and our sometimes sloppy words actually work well in these sloppy domains.
Duffy
I went hunting on his site for those and coudn't find them. Would you mind posting a link?
One common memory colour that is often critical is pleasing reproduction of skin. I doubt that there will be much argument from the folk here at DG on this point.
The Atkinson composite image test does not include any new skintones (the LAB image, right?) - the only skin featured is a reduced version of the PDI chart that is being maligned!
So, It would appear that the theory is not being dismissed, just the reference image? OK, then...
So, let's try a different reference image for skin tone, the replacement image for the old PhotoDisc, now the Getty Images Test Image:
http://creative.gettyimages.com/source/services/ColorResources.aspx
http://creative.gettyimages.com/en-us/marketing/services/Getty_Images_Test_Image.jpg
http://creative.gettyimages.com/en-us/marketing/services/Getty_Images_Test_Image.tif
The Caucasian lady to the left, an average skin reading in CMYK (Adobe SWOP v2 Web Coated TR001) and RGB (Adobe RGB 1998) is:
16c 47m 51y 0k
191r 141g 110b
65L +23a +28b
Once again, the old prepress CMY ratio and the Photoshop Color palette slider positions in RGB and LAB show a helpful relationship as a guide to pleasing colour. The RGB skin is not out of gamut for SWOP v2 CMYK. Again the image was not edited by me to conform to the legacy target values suggested by prepress tradition, the original colour was captured/presented this way by the test target creator as an "accurate" rendition of the photographed scene. That human observers and modern capture/colour conversion agrees with the old prepress general ratio is not surprising to me.
It really puzzles me why some are so resistant to admit the benefit of taking colorimetric readings inside Photoshop using the info palette and various colour reading modes, for all areas of the image and not just endpoints and known/presumed neutrals. These same people often advise similar measurement based methods (using hardware & software not built into Photoshop) for calibrating and profiling monitors and inkjet printers - but seem to reject similar methods of evaluating images and image edits inside Photoshop using the built in tools provided.
To each, their own - I guess!
Regards,
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
I have never edited any image for output to a commercial press, but I can see how difficult it must be. I flipped through one of Scott Kelby's books illustrating a skintone edit tip in the bookstore one day. It was the one of a close-up on a woman's face wearing sunglasses and noticed the before picture looked better than the corrected shot which made her look jaundice. Maybe it was press drift. Not sure. The tutorial did show the CMYK readouts and I couldn't tell if the amount of yellow was too much or not.
There is one color target where I could actually confirm CMYK readouts as being predictable as they should appear on press and that was viewing a print of the untagged Ole No Moire color target found on PS install CD's printed from Copy Craft's sheetfed Komori waterless press at 300lpi.
This press uses a closed loop color check system using a constantly scanning spectrophotometer that automatically adjusts for press drift, so the CMYK dot percentage sizes in the PS readouts reads exactly that on the print as viewed under a magnifying glass. But the print still showed overall gain in the cyan channel in neutrals, but the fleshtones were spot on.
Also this untagged CMYK target looks way too light when assigning PS's standard coated SWOP v2 so I assigned the sheetfed v2 and the image looked much more closer to the Komori print. See attached scan of the print next to the assigned sheetfed profile on the right.
The problem I've always had viewing dot sizes off press runs is you can't see the yellow because it's too light. Doh!
Agreed Tim, as mentioned earlier, children and faces change things a lot. Ideally one can sample many areas and take it all with a grain of salt, remembering that these are just guidelines or they can become more specific and targeted to output if this is a "correction" to a "known" problem rather than just a subjective general edit in an intermediate working space.
I just checked the famous "musicians" scene of the Caucasian, Asian and Negro women with their instruments and ethnic dress, that was originally presented as untagged CMYK. I averaged 10 common press spaces of this image into an ECI-RGB file and an average non facial area of the Caucasian lady on the left is:
18c 36m 49y 0k (SWOP v2)
197r 164g 133b (Adobe RGB)
72L 14A 24B
Again, a similar pleasing ratio is reflected in the color palette via the position of the CMYK/RGB/LAB sliders, as shown in the animation on page 8 of this thread. Note the difference in colour and tone from the Getty image, both ladies are very different and so is the lighting - but both are acceptable (one could argue that the Getty woman is too dark and magenta or the musician too yellow and or not magenta enough etc, this is the subjective nature of such evaluations).
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
Did you get my email requesting the link to that Blue Jacket CMYK gamut edit tutorial I used to get off your site but now have lost since you redesigned your site?
I really thought that was such an informative editing tip.
Hi Tim and thanks for the interest, yes I did get your email about the page that is offline with the new site redesign.
I will email you offlist tonight when I have access to email again and not just the web.
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/
http://homepage.mac.com/billatkinson/FileSharing2.html
http://homepage.mac.com/WebObjects/FileSharing.woa/wa/Profile_Test_Images.zip.zip?a=downloadFile&user=billatkinson&path=/Public/Profile%20Test%20Images.zip
More links to other targets and charts here:
http://members.ozemail.com.au/~binaryfx/links.html#C (Scroll down to "Characterization & Calibration Targets")
Regards,
Stephen Marsh.
http://members.ozemail.com.au/~binaryfx
http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/