LAB COLOR

[Mark D Segal]

I was looking at Guillermo's results on that one, Mark. I certainly came away with the understanding that according to Guillermo's tests, hue was not exactly preserved (though it was close) in an CR curve adjustment:

Guillermo:
Sorry, I didn't realize he recapped his results on the final page of the Your Curves thread, but these are his results. Have you found/heard something to the contrary? Or do you figure the changes are too close to count?

Post # 188

Gloria
[a href=\"index.php?act=findpost&pid=155338\"][{POST_SNAPBACK}][/a]

Gloria - ya - what's to the contrary is that the writer of the code says what I reported. As well, most of my test results bear him out. I did have one outlier which no-one who saw it could figure out. But otherwise it behaved as advertised. I never really understood how Guillermo got the hue shifts he reported - would have taken a lot of back-and-forth to get to the bottom of that and I'm not sure I would have been the peson able to decipher exactly how he got what he got, so I didn't push it.

[laughfta]

Gloria - ya - what's to the contrary is that the writer of the code says what I reported. As well, most of my test results bear him out. I did have one outlier which no-one who saw it could figure out. But otherwise it behaved as advertised. I never really understood how Guillermo got the hue shifts he reported - would have taken a lot of back-and-forth to get to the bottom of that and I'm not sure I would have been the peson able to decipher exactly how he got what he got, so I didn't push it.
[a href=\"index.php?act=findpost&pid=155357\"][{POST_SNAPBACK}][/a]

Thanks Mark. This will go back on my sometimes, probably, most of the time, somewhat list of interesting PS effects.

Gloria

[Mark D Segal]

Well, I have some news for you: real life is more complicated. I usually use a combination of several methods. It depends on the actual images and on the personal level of acceptance, how laborous the process is.
You did not understand what I was writing. Of course you can compare them, but not in ACR. The above way may be right with your workflow, but that is not how I am creating panoramas.
This is exactly what I do and what I described above. The difference is only that I often convert the sky in Lab for adjustment.
You too could profit from the experiment I describe in my previous post.
You got at least this one right. I am using selective color adjustments in PS (the adjustments in ACR are of very limited use). However, when adjusting the sky, I prefer Lab. I do agree, that color adjustment in Lab may be a pain.
[a href=\"index.php?act=findpost&pid=155353\"][{POST_SNAPBACK}][/a]

Grerat, we've had news for eachother - keeps discussion lively   .

I agree - no one method of selecting a sky works all the time. I wasn't saying that. Just in my experience, more often than not working the blues in those HSL tabs of CR4.x can really go a long way to getting a very good end result. To that extent, it alleviates the workload and gymnastics in PS.

And I agree - one cannot compare image adjustments between two versions simultaneously in CR; I just don't think that's a big limitation, but as they say YMMV.

For some time I also reverted to Lab for certain kinds of enhancements and found it very effective. But it is a tricky space to work with, and I just find that new tools and techniques have substantially reduced its relative attractiveness - but I'd be the last one to say it's not useful some of the time. There will always be those situations............

In the procedure you recommended to Andrew, I could understand where you were going until item #5, and there you lost me. So I don't get the point of it. Something isn't explained quite right there - at least to my understanding. I'm assuming that whether we are talking panoramas or more usual photographs the issue at question here is not related to that, but rather to the relative effects of making luminosity tweals in RGB Luminosity versus LAB Luminosity - right?

[Mark D Segal]

Thanks Mark. This will go back on my sometimes, probably, most of the time, somewhat list of interesting PS effects.

Gloria
[a href=\"index.php?act=findpost&pid=155361\"][{POST_SNAPBACK}][/a]

Yup - at this point that's where it's at with me too.

One can go on and on experimenting, but you know, with a properly profiled display and a good printer you can see where these curve adjustments are heading and there's umpteen ways at various stages of the creative process for dealing with unwanted side-effects, so it just isn't that big a deal - but that's me. I think once one gets into - say- advertising work, whatever one does the numbers often need to be spot on - but even in those cases based on what we've been told about the behaviour of those curves, it's not at all clear that L*a*b is necessarily the most efficient or effective route to colour accuracy.

[Panopeeper]

I'm assuming that whether we are talking panoramas or more usual photographs the issue at question here is not related to that, but rather to the relative effects of making luminosity tweals in RGB Luminosity versus LAB Luminosity - right?

Exactly. The point is, that color integrity is not maintained (can not be maintained) in RGB. This means, that if you change the luminosity, the color changes as well, except if all non-zero components have the same value. Grey, red, green, blue, yellow, cyan and magenta remain grey, red, etc.

There is no such problem in Lab. The other side of this is, that when converted back to RGB, the change of luminosity becomes non-linear. I don't find that a problem, particularly because any contrast enhancement changes the luminance proportions anyway, so the integrity of luminosity is a moot issue. Although this change does effect the contrast to some degree, I have never seen this as an issue, but I can imagine applications, where this might pose a problem. However, color integrity is IMO more often an issue.

[Brian Gilkes]

Brian,

Not having gone to "freegamma.com yet, which I shall, meanwhile could you elaborate a bit what you have in mind by "sufficient levels of accuracy" and relative to what kind of exacting applications? Also, do you understand how Lobster solves whatever this problem may be - I intend to visit their site too - last time I went there, it was Mac only which ruled me out so I lost interest.

Cheers,

Mark

About 3.5 years ago I had a job to reproduce some hundreds of pieces of artwork. These  consisted of oil, acrylic, pencil, crayons and stuck on bits of siver and gold metal. I learned a lot about colour inconstancy as light sources changed (on Epson K2 inkset then) and the need to edit custom profiles. Even after sorting out aforementioned hassles I would adjust saturation and tone changed and VV. Blending modes helped a lot but when it came to getting those colours perceptually accurate , I was still chasing my tail. When Les Walkling showed me Lobster (no , I am not a born again salesman) I could not believe how you could crank curves up and down and only get the shift you wanted. The screen images remained "clean". At that time I tested the numbers and was getting the small (and in some cases large) shifts that Free Gamma claimed. The end of that story is I finally got those colours right with a new workflow that incorpoated the software. I may have told this story a long time ago , but the audience is larger now and the software is dual platform.
When required ,I use the Lobster approach. I don't use it all the time, as you have to flatten layers , which I prefer not to do , and the file size is large.
I was on CS at the time and have recently moved to CS3 .I have not re-done the tests, but will do so when I get a break. The tests, as suggested in the Lobster manuals, take a while and some concentration.
Cheers,
Brian
www.pharoseditions.com.au

[laughfta]

The point is, that color integrity is not maintained (can not be maintained) in RGB. This means, that if you change the luminosity, the color changes as well...

But if hue is maintained in ACR, wouldn't it be easier and just as effective for you to make adjustments there, or with the HSB sliders therein? That is, if your only reason to move to Lab is to maintain color integrity?. (Noise, haloes, and  global application being other considerations in select images)

[Mark D Segal]

Exactly. The point is, that color integrity is not maintained (can not be maintained) in RGB. This means, that if you change the luminosity, the color changes as well, except if all non-zero components have the same value. Grey, red, green, blue, yellow, cyan and magenta remain grey, red, etc.

There is no such problem in Lab. The other side of this is, that when converted back to RGB, the change of luminosity becomes non-linear. I don't find that a problem, particularly because any contrast enhancement changes the luminance proportions anyway, so the integrity of luminosity is a moot issue. Although this change does effect the contrast to some degree, I have never seen this as an issue, but I can imagine applications, where this might pose a problem. However, color integrity is IMO more often an issue.
[a href=\"index.php?act=findpost&pid=155413\"][{POST_SNAPBACK}][/a]

Hi Pano,

You are correct that shifting the RGB composite Curve in Normal Mode does not preserve "colours"'; more accurately it definitely does not preserve saturation - it was designed that way on purpose - and there can be slight hue shifts as well. I have tested this issue of hue shifts at least several times over the past few years when it surfaced in discussions on this website and others in various contexts. My results have always indicated that these hue shifts tend to be very slight, if any. This I found by retaining the RGB working space, but looking at the measurements "before" and "after" in RGB, L*a*b and HSB measurement systems. By the way, I have two articles on this website dealing in part with this subject (Do Curves Throw You a Curve, and the sequel).

Now, as to the merits of L*a*b for preserving "colour" - the FACT is - as I mentioned in a previous post here, "L" has only fair chroma consistency, so if you are after "colour integrity", the "L" channel in theory is not where you will be certain to obtain it. I completely agree with a point made above by Jeff Schewe that shifting to L*a*b for this purpose isn't worthwhile compared with using the RGB Curve in Luminosity Blend mode. On a print of a real-world photograph, you really do need to search long and hard for any significant difference of results between shifting the "L" Curve or shifting the RGB Curve in Luminosity Mode. And with a bit of skill you can often simulate an L*a*b result in Camera Raw so darn close you lose track of which image is which. Of course this would vary from image to image, but the basic point is there.

So going back to Tad's original question in Post #1 above, I think Dan Margulis demonstrated much interesting and useful stuff in the L*a*b book, but I personally find that for the overwhelming majority of images I work-up (roughly 200 per month) there's been no compelling need or advantage of converting to L*a*b, especially since Lightroom and CS3. But if incorporating L*a*b in a regular workflow suits your needs best, then of course that's fine too - what's so nice about Photoshop - multiple paths to similar destinations.

[Mark D Segal]

Brian,

Not having gone to "freegamma.com yet, which I shall, meanwhile could you elaborate a bit what you have in mind by "sufficient levels of accuracy" and relative to what kind of exacting applications? Also, do you understand how Lobster solves whatever this problem may be - I intend to visit their site too - last time I went there, it was Mac only which ruled me out so I lost interest.

Cheers,

Mark

About 3.5 years ago I had a job to reproduce some hundreds of pieces of artwork. These  consisted of oil, acrylic, pencil, crayons and stuck on bits of siver and gold metal. I learned a lot about colour inconstancy as light sources changed (on Epson K2 inkset then) and the need to edit custom profiles. Even after sorting out aforementioned hassles I would adjust saturation and tone changed and VV. Blending modes helped a lot but when it came to getting those colours perceptually accurate , I was still chasing my tail. When Les Walkling showed me Lobster (no , I am not a born again salesman) I could not believe how you could crank curves up and down and only get the shift you wanted. The screen images remained "clean". At that time I tested the numbers and was getting the small (and in some cases large) shifts that Free Gamma claimed. The end of that story is I finally got those colours right with a new workflow that incorpoated the software. I may have told this story a long time ago , but the audience is larger now and the software is dual platform.
When required ,I use the Lobster approach. I don't use it all the time, as you have to flatten layers , which I prefer not to do , and the file size is large.
I was on CS at the time and have recently moved to CS3 .I have not re-done the tests, but will do so when I get a break. The tests, as suggested in the Lobster manuals, take a while and some concentration.
Cheers,
Brian
www.pharoseditions.com.au
[a href=\"index.php?act=findpost&pid=155416\"][{POST_SNAPBACK}][/a]

Thanks for that explanation Brian. I see what kind of images you are talking about - that is indeed very challenging in two respects - getting the artwork to look like it should, and reproducing metallic colours without metallic inks. I did now visit the Lobster website again (I had actually seen it some time ago before there was a WINXP edition). It looks enticing in some respects, but the flattening and file size issues I must say dampen my enthusiasm. It looks like the kind of application one would use when nothing else works properly - and he is careful to keep repeating one should use it on a duplicate of the image. Talking about nothing else working properly, periodically Curvemeister is mentioned as being a useful application for making Curves adjustments in multiple modes. Have you ever worked with that application (I haven't, yet)?

Now that you are on CS3, an interesting thing to try may be to revert to one of those challenging artwork images - a raw file - duplicate it, reset all the conversion adjustments to zero and linear retaining "As Shot" white balance, and reprocess the image as far as you can using all the available tools in CR 4.1+. Then render into PS and do the remainder there, without Lobster, as best you can, and compare the outcome with your Lobster results. If you have the time and interest to do this, I'd be very interested in your observations about the comparative outcomes.

[laughfta]

periodically Curvemeister is mentioned as being a useful application for making Curves adjustments in multiple modes. Have you ever worked with that application (I haven't, yet)?

I use Curvemeister a lot, and find it very effective in making adjustments to skies (an example we have been using). One can create a mask derived from any channel to isolate the sky if needed (usually making further blending unnecessary), and then use a curve from within any color mode to make adjustments. I am finding great benefit from using curves from the HSB color mode on the skies (Lab curves are also available), and these curves all work independently from each other, and info can be read on specific targets while adjustments are being made.

[Chris_T]

Lobster seems like a great tool for those who want to edit luminosity and color separately. I would think that it is used at the front end of a workflow before other layers are generated. So why is flattening an issue?

The site provides no real image examples before and after applying Lobster, making it difficult to see what the effects are. Google and I find no user reviews. Perhaps it is only appreciated by a few, but not very popular.

I did now visit the Lobster website again (I had actually seen it some time ago before there was a WINXP edition). It looks enticing in some respects, but the flattening and file size issues I must say dampen my enthusiasm. It looks like the kind of application one would use when nothing else works properly - and he is careful to keep repeating one should use it on a duplicate of the image.
[a href=\"index.php?act=findpost&pid=155455\"][{POST_SNAPBACK}][/a]

[Mark D Segal]

Lobster seems like a great tool for those who want to edit luminosity and color separately. I would think that it is used at the front end of a workflow before other layers are generated. So why is flattening an issue?

.......................

[a href=\"index.php?act=findpost&pid=155736\"][{POST_SNAPBACK}][/a]

You can always flatten the luminosity and colour adjustments into the Background layer, but then you lose the principal advantages of working with Adjustment Layers for amending these edits should you wish to re-purpose the file or make changes that you later decide are necessary. The only changes I bake into the file after leaving Camera Raw are crops, perspective adjustments, noise reduction and capture sharpening. The rest is on layers, and as much as allowed on Adjustment Layers.

[Panopeeper]

You are correct that shifting the RGB composite Curve in Normal Mode does not preserve "colours"'; more accurately it definitely does not preserve saturation - it was designed that way on purpose

Only a small technical remark (otherwise I think the subject has been discussed more than enough): not maintaining the color integrity is not per design, it is the nature of RGB, it could not be otherwise.

[digitaldog]

Only a small technical remark (otherwise I think the subject has been discussed more than enough): not maintaining the color integrity is not per design, it is the nature of RGB, it could not be otherwise.
[a href=\"index.php?act=findpost&pid=155862\"][{POST_SNAPBACK}][/a]

Actually no, if you read Mark's article here, you'll that it is indeed by design.

[fennario]

I had been using LAB quite extensively until I purchased CS3.  I've found that I am able to utilize the tools in ACR now to achieve similar effects (tone curve, vibrance, clarity, and HSL tabs) and no longer make the trip to LAB as often.  There are still a few areas where I still find it useful (creating layer masks and new L channels from a single RGB channel) but it is no longer the integral part of my workflow that it once was.

The Canyon Conundrum still occupies a place of honor in my bathroom, but I have just purchased Real World ACR and am eagerly awaiting its arrival.

P.S.  Thank you to the forum members who have participated in this and other discussions on LAB... you have helped me gain a much better understanding of its proponents, merits and issues.

[papa v2.0]

again
what you  have to understand is that when you  view an image its in sRGB ( or if you have a CRT or an LCD or PLASMA) maybe in ABOBE1972

your picture is en coded in a space you cant see!

not the editing space!

don't do visual editing on as RGB monitor

the monitor is and is struggling to be to "negative'

its no where near.

[Tim Lookingbill]

I've tried out editing in Lab space white balancing the blue cast out of sunset lit limestone rock while not affecting the blue in other elements in the scene such as flowers. Far more intuitive and smoother than using selective color in RGB.

However, one issue I found from my own experience working in this space is that it tended to skew my color perception through the adaptive process when I took too long editing in this space. With each edit along the a/b curves tended to ever so slightly change the hue of nearby colors on the color wheel within the image without me noticing.

When I took a break to allow my eyes to adjust a bit, something just looked off about the image looking at it with a fresh eye.

I've seen results similar to this effect on the web in the past of images edited in this space most seeming to have a very neutral carbon black undercolor sheen starting from shadows tapering off into the mids. It's kind of like the desaturated effect that comes about applying a blur on color blend layer to get rid of purple fringing. An overall inconsistancy in hue and saturation levels permiates through different areas of the image giving it an odd look.

And the Dan inspired tutorials I've come across on the web tend to be quite long winded and too much work for what little results that were gained.

[Jonathan Wienke]

I've tried out editing in Lab space white balancing the blue cast out of sunset lit limestone rock while not affecting the blue in other elements in the scene such as flowers. Far more intuitive and smoother than using selective color in RGB.

For something like that, I'd do two developments of the RAW file, the first white balanced for the limestone, the second white balanced for the rest of the scene, layer one over the other, and paint the layer mask to blend to taste. Much easier than selective color tweaks that generally have unexpected consequences in unrelated areas of the image.

[Tim Lookingbill]

Jonathan,

This was a scan from a Kodak UC 400 negative of as scene that seemed to have a much wider gamut than my Epson 4870 and negative could deliver.

The fine tweaks on the 16bit tiffs I could do in Lab were unbelievable. It was like editing the color lookup table of the image itself which can get really hairy if you don't know what you're doing. And apparantly I didn't.

[Tim Lookingbill]

Here's the image. The more I go back to it the more I have say to myself what the hell was I thinking?

You'll see from the top Epson Auto Exposure version what I had to deal with compared to the results of Lab at the bottom which looks more to what I imagined it from memory. The flowers were that exact color and limestone was warmish because of how low the sun was set. It still looks off. It doesn't look how I remember the scene.

The flowers were intense and eyepopping and I couldn't get that look editing in RGB, though I tried, from what the Epson and the Kodak negative delivered.[attachment=4135:attachment]