Dan Margulis Applied Color Theory

Superadvanced Classes, 2009

Superadvanced classes
Posted by: Dan Margulis
Fri Mar 20, 2009 9:06 am (PDT)

In response to offline inquiries, I do indeed have plans to post here what happened during the superadvanced courses that ran in San Diego during the last two weeks--but I will not do so for another week and a half.

Spending that much time with people who really, really know how to handle images was a once-in-a-lifetime opportunity and, I felt, very productive. We learned much from each other.

I took advantage of the talent level to test some new ideas. Normally, all members of any of my classes work on the same image and we compare results. This time, on about half the exercises (and the attendees did not know which) there were two versions of the same original, and half the class, chosen randomly each time, was in fact working with a handicap. Or, at least, the average observer would *think* that it was a handicap. Then, we compared all the results, regardless of which version was the start point.

This is a test that can't be run in normal classes because there's too much skill variation. The better retoucher would almost always win in spite of the handicap. But here, with everyone at more or less the same talent level, and 16 different versions of each image to compare, one can get a very good idea of what the par result is and how much the handicap affected people's ability to get there.

I had a feeling that something like what did happen was going to happen, but it came as a shock to some people. Right now I am spending a lot of time analyzing the results further, to make sure that what seemed to everyone to be happening is in fact what did happen.

I need to finish up with this because I leave on Monday for Photoshop World, and the findings may affect my sessions. For those interested, I'm doing a four-hour LAB session on Tuesday afternoon (separate registration required) and two other two-hour sessions during the course of the show. On Thursday afternoon my sessions cover channel operations and on Friday at 11 comes the update of the picture postcard workflow.

You'll here more about all of the above week after next.

Dan Margulis
___________________________________________________________________________

Re: Superadvanced classes
Posted by: "Jeremy Schultz"
Tue Mar 24, 2009 4:56 am (PDT)

As one who was invited and opted not to attend the superadvanced class, I?m really looking forward to hearing the play-by-play. :)

Jeremy Schultz
___________________________________________________________________________

Re: Superadvanced classes
Posted by: Dan Margulis
Tue Mar 24, 2009 7:55 am (PDT)

It shall be done, so many things happened that I am preparing a series of posts rather than just one, in case threads develop out of any of them. But it's not going to happen until next week when I get back from Photoshop World. Besides, events here in Boston may impact what I have to say about what took place in San Diego.

Dan Margulis
___________________________________________________________________________

Superadvanced/PSW 1--economic observations
Posted by: Dan Margulis
Sun Mar 29, 2009 5:27 am (PDT)

Now that Photoshop World has ended, I will post a series of messages discussing some of the things that happened during the recent superadvanced courses in San Diego, as well as in Photoshop World itself.

The superadvanced courses brought together a formidable array of talent. We drew certain important conclusions on the spot about various techniques. Because of the complexity of the issues, though, I want to review the classes' work carefully before issuing any manifesti. I couldn't do that last week while preparing for Photoshop World, so will try to do so this week in between filling out the mountain of tax forms that U.S. residents have two more weeks to complete.

Meanwhile, therefore, I'll post on some topics that require little or no image research, starting off with the 600-pound gorilla lurking behind all of us, the economy. There were certain hopeful signs as well as a lot of foreboding ones.

The good sign in the superadvanced courses was that around a third of the attendees said their businesses were not affected by the economy at all and that they would not know that we were in a crisis except by reading newspapers. The rest were feeling it to some extent. The professional retouchers seemed to be the ones doing the best, on the whole. This is in keeping with the history of the last 15 years--those who can manipulate color well are rare enough that they make their own opportunities and are always better off than the rest of the graphics field.

As against that, three different people whose companies I would expect to pay for these courses volunteered to me (I would not ask such a question, obviously) that their companies had originally agreed to do so but had reneged, and that they were footing the bill themselves. My feeling is that of the 14 people who attended the superadvanced courses, a maximum of two were being paid for by companies--and it could very well be zero.

Worse, the class was a skewed sample, because a lot of people had already dropped out, one presumes for economic reasons. These classes had been in the making for more than a year, and there was a lot of interest. In September, we needed to make a final decision as to how many sections to hold. So we asked everyone to look deep in to their souls and state whether they were ready to commit to attending, and that we would, after announcing the schedule, be collecting money within the month.

Owing to my impeccable sense of timing, the stock market crash occurred between the time that people committed and the time that they had to finalize registrations. Consequently, 27 people *said* they would attend, but only 16 actually went ahead with it two weeks later, and we lost two more along the way.

Perhaps ominously, the dropouts were segregated by profession. Two people worked for newspapers, and one can understand why they could not come--newspapers are currently in a financial position that makes General Motors look like the Rock of Gibraltar. But the cut also swept seven photographers with it. Only three people who made the final classes would consider themselves photographers first and foremost--and none of them are what I would call typical in terms of their business model.

I had lunch in Boston with one of the photographers who dropped out. He said things had gotten better since then. His business is mostly providing CMYK files to advertising agencies, and says that right now things are hopping, although pricing has become very competitive.

Meanwhile, Photoshop World itself also did better than might have been expected. It looked like attendance was off in the preconference sessions, which start a day early and require an extra payment. The main 3-day show, however, did not appear to be down in attendance from last year by very much, if at all.

It did seem to me, however, and also to several other instructors, that the attendees were 1) older than usual; 2) more sophisticated than usual. I actually redid one of my lectures on the fly to make it more difficult because I was seeing body language from the audience that I didn't like when I was explaining simple points.

This trend in the audience is being driven by an influx of prosumers--fiftyish folk from other professions, who are deadly serious about their habit and have the money to support it. I polled the audiences before two of my classes with a series of questions that were intended to elicit (without asking it outright) whether the attendees are trying to make a living in the graphic arts. The number who were was lower than in previous years--but the level of Photoshop sophistication was higher. Which I suppose is a good thing. Those footsteps we hear behind us are the prosumers catching up to us.

Dan Margulis
___________________________________________________________________________

Re: Superadvanced/PSW 1--economic observations
Posted by: "Jeremy Schultz"
Sun Mar 29, 2009 3:35 pm (PDT)

Interesting material, and not unexpected‹it follows some trends that have been suggested in Dan's previous findings.

I would be in the category of those who find their business not affected by the economy‹my 1Q this year has turned out better than the last one. I know one or two projects that were coming to me that fell apart partly because of the economy, but it was not the deciding factor. I opted not to attend the superadvanced course because my first child was expected in January and I decided it would be better not to be away from home.

I am neither a newspaper designer nor a photographer, BTW.

Jeremy Schultz
Design and illustration for print and the web
___________________________________________________________________________

Superadvanced/PSW 2--Gaussian blur vs. Surface blur
Posted by: Dan Margulis
Thu Apr 2, 2009 3:06 pm (PDT)

One of the main subtopics at the superadvanced courses was a comparison of various ways of reducing noise. Several threads could be started on what was discussed, but this time I want to cover an area that I myself had misunderstood. Those who are interested in the picture postcard workflow may find some interesting things with which to experiment here. I do not have recommendations yet, but only call people's attention to the potential for an improved workflow for certain images.

Everybody has his own favorite method of noise reduction. It seems, however, that the majority of attendees were making extensive use of the Surface blur filter, which didn't even exist until Photoshop CS2.

I had previously thought, based on some testing I did when CS2 was released, that Surface blur was to all intents and purposes Gaussian Blur, with the addition of a threshold field by which we could protect certain edges from the effect of the blur. The Threshold is expressed in terms of levels, maximum 255. If we put in a threshold of, say, 20, the filter won't touch adjacent pixels that vary by more than 20 levels. Consequently, it's often possible to use the filter to reduce or eliminate noise, while not blurring the edges of objects.

When setting a low threshold is advantageous Gaussian blur is of little use. I thought that if the threshold was not needed, there would be no point in using Surface blur at all. That is, I thought that if we set the Surface blur threshold to 250, we would get a very similar result to if we had just used Gaussian blur at the same radius.

It was pointed out in class that this is not correct. The Gblur is more diffuse. The fadeout of blurred areas is larger and slower, given the same radius. The Surface blur therefore gives the impression of holding shape better, even though it is just as effective against noise. As against that, when a halo occurs where two edges meet, the Surface blur halo is more abrupt.

The difference of the two methods is unlikely to be significant when the radius is small. However, we sometimes used blurred overlay layers to enhance highlight and/or shadow detail. Also, the picture postcard workflow routinely multiplies through a blurred layer mask. In both cases the Gblur radius is usually 30 or higher. Using Surface blur with an enormous threshold can make a difference, for better or for worse. (Note: to get something comparable to the Gblur effect, I suggest using 1.5 times the radius for Surface blur, thus a Surface blur radius of 45 puts one in the same general look as a Gblur of 30).

I've experimented with this a little since coming back, with inconclusive results. Most of the time there isn't much difference. When there is, I haven't yet been able to generalize about when one way is better. I've had a couple of bad results with Surface blur putting in extremely clunky halos. OTOH, sometimes it *avoids* the unpleasant halos.

If I were working today on an extremely important job, I'd take the time to make one blur each way and try each. Also I would consider the possibility of blending the two blurred versions 50-50 and loading that. Such a blend sometimes gives better results than either parent.

The real gain for Surface blur comes in highlight/shadow enhancement, when there isn't a whole lot of variation in the highlights and shadows to begin with but there *is* a major edge elsewhere, as often happens with relatively dark skies. I loaded an example image, dark_ocotillo.jpg, into the Files section of yahoogroups in case anyone wants to play, but the concept is simple. Moonlit scene in the desert. The foreground is too dark to show much detail but it's there if we can pull it out. Dark, thin branches of the ocotillo reach into the sky, which itself has some detail in its lighter clouds. Detail is also available in the moon. Therefore, this looks like a case for the Shadow/HIghlight command, followed by placing an inverted copy of the red channel on a layer by itself. The layer is then set to Overlay mode and blurred.

Using Gblur produces major and unpleasant halos around the branches. With Surface blur, however, one can pick a threshold low enough to exclude the branch/sky edge, while high enough to get all the benefit of blurring the clouds and the foreground.

Dan Margulis
___________________________________________________________________________
.
Superadvanced/PSW 3--angle of AB curves
Posted by: Dan Margulis
Sun Apr 5, 2009 4:56 pm (PDT)

For those who, like me, find that one of the most effective ways to boost color is with simple steepening of the AB curves of LAB as illustrated in Chapter 1 of my LAB book, here's a recommendation to change approach somewhat.

As we all recognize, if the image dictates, the angling of each curve can be handled independently, as can the top and bottom halves of each curve. However, if you can't see any reason for special handling, my published recommendation is to steepen the A and B curves by equal amounts.

With the benefit of time, I now understand that this is not correct. Other things being equal, we should steepen the A more than the B.

I speculate that the reasons are three:

1) As pointed out near the end of Canyon Conundrum, the average color value in the world is something like 0a5b. Assuming that to be correct, the world is not biased toward either magenta or green, but it is biased toward yellow as opposed to blue. Steepening the B curve aggravates this "natural cast".

2) We are not nearly as interested in making blues and yellows intense as we are reds and greens. Really brilliant blues are uncommon and in any case we are unlikely to be able to print them. Pale yellows aren't that common either. If there are significant yellow objects in the image chances are they're pretty bright already and don't require a lot of help from us.

3) Reds and natural greens contain a significant yellow component. As we intensify the colors, I believe it is usually more objectionable to go in the direction of a fiery, or more orange, red. Most of the time we prefer something rosier, if it is to be brighter. Similarly, as natural greens get more intense, we don't want them to get too yellow, but we don't mind so much a slight tilt toward blue.

I don't know whether these are the correct reasons but I *do* know that a steeper A curve works better most of the time. In the superadvanced classes, Photoshop World, and the two ACT classes that preceded it, I did most LAB color boosts using an Action that makes the A curve decidedly, although not grossly, steeper than the B.

I'll be posting that Action in a few days, at which time I'll discuss another new recommendation that more of you may be familiar with: instead of the traditional way of perfecting the AB curves starting with an image that isn't colorful enough, I now suggest that in almost all images we should set up a layer that is much too colorful, and then back off.

Meanwhile, I reiterate the suggestion that when applying simple AB curves, you should customarily make the A a bit steeper than the B.

Dan Margulis
___________________________________________________________________________

Re: Superadvanced/PSW 3--angle of AB curves
Posted by: "Jim Bean"
Sun Apr 5, 2009 5:41 pm (PDT)

dan wrote: Meanwhile, I reiterate the suggestion that when applying simple AB curves, you should customarily make the A a bit steeper than the B.

I think this is an accurate and much better approach than previously published... including the variations created by different skin tones... and the word choice, 'rosiery' is right on target... I see a lot of faces and I just could not make those previous 'heavier yellows' work for me...

jim bean
___________________________________________________________________________

Superadvanced/PSW 5--The Color Boost Action
Posted by: Dan Margulis
Thu Apr 23, 2009 1:43 pm (PDT)

In this economy, almost all of us are encountering friends or acquaintances who have fallen on hard times. Highly skilled retouchers, as I noted in an earlier post, are not exempt from the bad economy but are not doing all that badly either.

When one of them starts to divest himself of the usual retoucher lament, that he is so busy that he has no life, that he is getting little sleep and his love life is suffering, that he has a feeling that he could make a good living if it wasn't for having to deal with all these unreasonable clients, I have a standard reply. I ask, "Would you rather have the opposite problem?"

When working with color, that question can also arise. When trying to boost color in LAB, many people already deliberately go too far, intending to reduce opacity to taste afterward. I've concluded that a modified version of this is the correct approach. In other words, given the choice, the problem of being too colorful and having to decide how back off is preferable to not being colorful enough and having to force color in somehow.

This is particularly the case when using the picture-postcard workflow. The first step of that workflow merely provides acceptable but not vivid color; the second enhances detail with luminosity moves. After that second step, the color is usually worse than in traditional workflows, so it tends to need a big boost, and LAB is the space of choice to do it.

Accordingly, I now plan that in most cases I will have a super-colorful version that then gets toned down in an agreeable way. I've got a quick Action that I've given to the last few ACT classes, including the superadvanced, and the reviews are very positive. I've uploaded it to our Files section (Margulis_LAB_Color_Boost.atn).

The Action requires an LAB file. There are two curves adjustment layers. The top one is empty, just default curves. It's there in case we want to make some kind of move in the L channel. Since the workflow doesn't necessarily set highlight and shadow early in the process, this L move may be needed, or we may decide to lighten or darken the picture. But by default, this top layer does nothing.

The real pizazz is on the middle, Color Boost, layer. It consists of an extremely steep B curve and an even steeper A, in keeping with my recent post recommending that the A be enhanced more than the B. The default opacity is 75%, which is way too colorful for any but the blandest images.

The no-brainer way to make the color more reasonable is to reduce this color-only layer's opacity to taste. However, there are often better options. We are usually more willing to accept vivid colors in lighter areas as opposed to darker ones. Therefore, it's worth a try to, rather than reduce opacity, load the L channel as a layer mask, thus allowing more of the impact to hit light areas than dark ones. Then, with the mask loaded, the layer opacity can still be adjusted--either up or down.

If the base file is already fairly colorful and the opacity of the Color Boost layer looked best at around 15 or 20% opacity, then it probably won't make any difference whether you mask the boost or not. But the more bland the base layer, the more likely you are to hit one where the mask will have an advantage.

These two methods are enough to make the Action worthwhile, but there's a lot more available for those skilled at editing layer masks. By clicking into the mask curve, we can isolate the tonal range we want to affect, lightening or darkening the mask as necessary. Another powerful tool is to use the A and B channels to augment the mask. Rarely should the A or B be used as a mask itself, but by loading the L first and then applying the A or B in Overlay mode (sometimes inverted) we can emphasize certain colors over others.

Anyhow, this simple Action is a pretty useful gadget--once you get used to the idea that the weird look it imparts when first applied is a better problem to have than the alternative.

Dan Margulis
___________________________________________________________________________

Re: Superadvanced/PSW 5--The Color Boost Action
Posted by: "John Bongiovanni"
Mon Apr 27, 2009 7:58 pm (PDT)

Dan Margulis wrote:

The real pizazz is on the middle, Color Boost, layer. It consists of an extremely steep B curve and an even steeper A, in keeping with my recent post recommending that the A be enhanced more than the B.

So how does this affect your Multiply move in the Picture Postcard Workflow? Here the Multiply is doing an Overlay of A and B channels onto each other. The Overlay is the equivalent of a curve on the A and B channels. What that curve looks like, I don't know, but I know that it is the same curve for A and B channels. What you're saying here (and in the earlier post) is that you need a stronger move in the A than the B channels.

Are you suggesting replacing the Multiply move with some kind of curve? Or something else?

John Bongiovanni
___________________________________________________________________________

Re: Superadvanced/PSW 5--The Color Boost Action
Posted by: Stephen Marsh
Tue Apr 28, 2009 1:08 am (PDT)

John Bongiovanni wrote:

The Overlay is the equivalent of a curve on the A and B channels. What that curve looks like, I don't know...

I have uploaded an "overlay" blend mode curve screen capture to the files section titled: overlay_curve.gif

photoshopnews.com/2007/09/05/how-to-express-blend-modes-as-curves
photoshopnews.com/stories/downloads/blend_curves.zip

Regards,

Stephen Marsh
___________________________________________________________________________

Re: Superadvanced/PSW 5--The Color Boost Action
Posted by: Dan Margulis
Tue Apr 28, 2009 7:49 pm (PDT)

John Bongiovanni writes,

So how does this affect your Multiply move in the Picture Postcard Workflow? Here the Multiply is doing an Overlay of A and B channels onto each other. The Overlay is the equivalent of a curve on the A and B channels. What that curve looks like, I don't know, but I know that it is the same curve for A and B channels. What you're saying here (and in the earlier post) is that you need a stronger move in the A than the B channels.

Are you suggesting replacing the Multiply move with some kind of curve? Or something else?

This is a good question but I'm not sure I have the correct answer. There are several ways of modifying the multiplication to make more use of what I posted. At the moment, what I'm doing is:

1) Multiply as usual, except I am more prone to use the top layer's L channel as the layer mask, rather than the bottom's. Since the top layer has been multiplied its L channel is darker, and the resulting image is lighter. This gives a little more flexibility because the Action has a layer that is reserved for adjusting the L. Also, the lighter image doesn't boost the colors quite as much, allowing the Action's color boost layer a bit more room to maneuver.

2) Apply the Action as usual, but consider the possibility of reducing the opacity of the multiply layer. The Action's color boost layer and the multiply layer both try for the same objective but use different ways to get there and one may be preferable to the other.

This idea is similar to what I am now using for images with tepid color. Before moving from RGB into LAB, on a duplicate layer I apply a master curve that lightens the midtone and darkens the shadow. This layer is set to Color mode. The result is a color boost, but of a different flavor than what's about to happen in LAB.

I keep that layer intact upon moving into LAB, and then apply the Action. In some cases having that extra color boost gives something that the LAB curves don't and in others it's a waste of energy that gets deleted. But I don't mind looking at the choice.

Dan Margulis
___________________________________________________________________________