A Toast to Greece: comments on individual versions

Dan Margulis

901 Chosen for the par version. This person created the most exciting version with a relatively simple procedure. He balanced the colors normally and added a lot of Vibrance. He added shape with two applications of H-K (luminosity only) which he states was the first time he had used it on an image. This accounts for the very realistic shaping of the faces.

The color of the faces is perhaps another story. The younger people at the near table are fine but the warm lighting over the head table leaves a tremendous feeling of orangeness in the fleshtones. An easy solution would be another little-used action in the PPW panel: Skin Desaturation. Since it targets only reds that might be skin, it won’t damage the stripes of the U.S. flag as other methods might. Usually Skin Desaturation would be applied before Color Boost, but there would be no problem in applying it to this version right now, and it would improve things.

902 This one came in at the last minute and I did not check before posting. It is one of four versions where the entrant disregarded the requirement that the file be in sRGB. This one is tagged Adobe RGB which makes it necessary to convert it before comparing it to anything else. Please, everyone, in future double-check that you are entering in the correct colorspace.

As it stands the shape is very nice but the color is way off, the skintone seems very desaturated. The reason can be measured in the tablecloth overhang near us. It shows up as green, which can’t be right. That green kills the red in the fleshtones.

903 Since I’m complaining about the rules I point out that this person erased the two lights as well as some leaves in the upper right corner. That may be what a client would want but it isn’t appropriate in a group study, because if I had wanted to include this in a par version it wouldn’t work, any more than if the person had slightly rotated the image.

Other than that, it’s the opposite of #902: nice color, no shape. The reason? Compare it to the par version, #931. The fleshtones are about the same darkness. I find a typical 52L in the fourth man from the left in both versions. The highlight in his shirt, however, is 10L darker in #903. So, about a quarter of the transition between lights and skin is gone and all the white stuff looks dead. See note to #926, which has the same issue, for a solution.

904 This person’s notes confess that he became fixated on the lighting and on preserving the feeling that this shot was taken at night. That the background tree loses a lot of detail is no big deal IMHO. However, if we were present at the scene, we probably wouldn’t see the faces as bright as they are in #901. OTOH even by candlelight we would see the faces at the head table clearly enough to recognize them. In this version they couldn’t be, so I say that the presentation is considerably too dark (which does not mean he should be shooting for #901 or anything like it.)

905 “My initial approach,” writes this person, “was to start off very badly wrong and to spin my wheels for days/hours, then realize my mistake, and start over." He tried to juggle a lot of prioirities in his final workflow and came up with a reasonable result with faces slightly washed out. As #904 is too obviously a night scene it looks to me like this one is too much taken-in-sunshine. I do, however, endorse the nice touch of vignetting out the top of the center tree, which further directs us to the people. Compare that effect to #915, where the fleshtones are roughly similar.

906 Chosen for the par version. Here’s my entry. Like many others, I was bothered by the crossing light sources. Unlike others, I decided to concentrate on not letting the whites get out of control and let the faces fall where they may. I tried doing this first with curves; not being satisfied, I had the inspiration of starting over using H-K reversed, which lightens the image overall instead of darkening it, while neutralizing the whites even further. Unfortunately, I did not grasp what #901 demonstrated, that normal H-K is very helpful with the shape of the faces. So, my H-K reversed version, when complete, wound up being pretty good in the whites and bad everyplace else.

Nothing daunted, I tried a third version, using Selective Color in RGB to delete color from Neutrals. That, combined with my first version, produced something fairly decent except in the highlights. So I blended it into my H-K reversed version through an inverted mask that only permitted changes in the whites. Other than that the good shape in the faces derives from curves and Hammer actions.

907 Similar in feel to #905 but more agreeable faces. It’s a good rendition if a bit lifeless. It seems that this person was dead set and determined to obliterate the crossing cast entirely so that the orangeness of the head table vanishes altogether and it seems that the scene is uniformly lit. The difficulty with such a doctrinaire position is that it seems to have a bleaching effect on the people at the head table and it becomes difficult to introduce more color.

Note how the handling of the warm cast on the right varies from that of #901, where the person didn’t care at all and just let the right side become bright orange, and from #906, where I tried to reduce the effect while not killing it completely.

Another issue: many of the techniques people used had the unfortunate side effect of reducing the redness of the stripes of the U.S. flag. No big deal: at the end we can just set the sponge tool to Saturate and paint the strong color back in. As this version is somewhat lacking in color interest, that should have been done here.

908 Another agreeable effort with good color and dubious luminosity. Rex Butcher concentrated on getting the fleshtones “right”, which he did, on the assumption that this is a normally lit picture. In context, however, they’re too light, because the background indicates that this is a dark picture. So, for example, in the woman third from left, the hair is too dark for the fleshtone. So, all the whites are in a compressed range, resulting in a flat appearance.

909 To some extent this one has believable features, at the cost of quite a bit of noise in the faces of the older people. Using this much sharpening, including ALCE, is dangerous in this type of picture. The man sixth from left seems to have a huge light source reflecting from his cheek, making him look like something out of a monster movie.

Roberto Tartaglione did an excellent job of neutralizing the cast as far as the people at the head table are concerned, it is hard to tell from them that anything has happened. But look at the tablecloth overhang beneath them. He eliminated the orange cast by adding blue, but forgot that the blue would turn the tablecloth into a color that doesn’t match the left half of the picture. So, this should have been corrected.

Also, this version offers a lesson about equalizing tonality. It isn’t a classic sun/shade image, of course, but Roberto correctly noted that there’s too much range between the lights and the darker half of the picture. So, he wheeled out two techniques that are designed to bring the two sides closer together: the Bigger Hammer and a false profile/multiply. But these methods presuppose that both sides are important. Here, I don’t think that it’s true. Compare this one to #919, which has somewhat similar people, but to my mind is preferable because it downplays the background trees, which we shouldn’t be paying so much attention to. Better, perhaps, to just darken the midtone originally rather than do something that enhances shadow detail.

910 A contestant from one of my classes who did not worry about the orange cast on the right. Fair enough; neither did #901, but this one would need to get lighter before it could be considered in the same league.

911 A careful workflow produced this excellent version. The person did three sets of corrections and blended them. One was aimed at enhancing color generally, the second was done in Camera Raw for a different result, and the third used a false profile, but unlike #909, care was taken not to allow the background trees to pick up too much detail. He decided not to attack the crossing casts directly. Instead, he painted neutrality into the tablecloths on each side of the picture, and he painted coolness into the faces of the older people at the head table to kill the orangeness. This appears to have been quite effective.

912 Arthur Margolin decided that this picture is not so much a family shot as a tribute to the older generation. He responded creatively by shining a spotlight on the head table, leaving the younger folk on the left in relative darkness. When I first opened the image I thought it was overall too dark but the more I study it the better I like it. Arthur writes, 

One significant challenge was to reconcile the various color casts such that the image made visual sense as a scene warmly lit, reflecting the emotional warmth of the gathering. The biggest hurdle was getting the color of the faces to be as close to not unbelievable as possible. Because each person's face interacted with the color casts in what turned out to be a frustratingly different way, I'm not convinced I have surmounted that hurdle (e.g.,I have multiple different and unsatisfactory versions of the woman standing at the head table at the far right).

My method was ad hoc and improvisatory.  In general I made much use of the dodge and burn tools as well as numerous curves targeting selected portions of the image, masking out others. The increase in contrast, which the original seemed to require, introduced or exaggerated odd shadows that I tried to dodge and burn and paint my way out of, not entirely successfully I'm afraid; for example, on the shirts of the three gentlemen and the sweater of the woman on the left.  
(Most perplexing to me throughout was what seemed to be a persistently higher than usually recommended cyan reading in the faces that could be reduced only by creating a strong red cast, which was objectionable and so dispensed with.)

That last should not have been worrisome. Normal fleshtone guidance presumes normal lighting. This is not; it’s a night shot. It’s to be expected that the fleshtones are more neutral than if in sunlight, thus, more cyan.

But, meanwhile, if you’d like to try this spotlighting effect on your own image, #920 gives you a golden opportunity to do so. See comments on it.

913 Recently, having some time to kill, I purchased a copy of Affinity Photo. Since this Greek image isn’t particularly in need of the PPW panel, I decided to leap into the deep end and do an Affinity version. I can say that Affinity Photo is a very serious application with a lot of useful tools that don’t necessarily have Photoshop equivalents, but there may be some PPW-crucial things that it doesn’t have, I don’t know for sure. Anyway, I felt like I was working wearing handcuffs. What came out was something more similar to #901 than to my own #906.

914 A nice conservative treatment done as a combination of three methods. First, a pretty standard PPW starting by blending the green into the RGB, and including Velvet Hammer and MMM+CB; second an auto-color like routine from a different app; third a PPW version that included H-K, mostly intended for the flooring.

915 Attempts to control the crossing casts with curves were successful in doing that, but unsuccessful in that it left the faces rather flat and without color variation. All the people at left have distinctly orange skin whereas at the head table the skin is pink. Also, it’s useful to compare this to #905, where the skintone isn’t all that different, but the person had vignetted out the top of the center tree, which was an effective move in directing attention back to the people.

916 My “official” entry into this case study is #906. As a test of my proficiency or lack thereof with Affinity Photo, I also produced #913 with that app rather than Photoshop. After doing so, it occurred to me that there was a third alternative. I did this one quickly in CMYK, which has no disadvantage because no colors are outside of the CMYK gamut and which does have the considerable advantage of a black channel that can be sharpened. Also, I did a second CMYK version using Heavy GCR, which I used to hold neutrality in the whites, thus fighting off the crossing cast. After merging the two CMYK versions I traveled to LAB to add color.

When I first compared the two I felt this one had the skin too dark and I preferred #906, but I’ve since changed my mind and now like this one better. The faces are more detailed, and all the people stand out better against the background, this being a result of curving the black channel in the three-quartertone.

917 Here are two examples of identifying a problem and then overcompensating for it. First, like so many of us, this person objected to the strong orangeness of the right side of the image. Second, we all admit that this photo, like many night shots, is grainy and is likely to have unacceptable noise unless we’re quite careful in sharpening. But it’s one thing to say we should minimize the problems and another to say that absolutely any trace of them must vanish. Here, the cast has been reversed: the head table is actually cooler than the table at left. And a noise reduction filter was applied too drastically, resulting in a blurry look. For example, the woman at the center of the table, with the wine held high, wears glasses—but you can’t detect that from this image.

918 A simple procedure led to a basically pleasing result. The person was concerned that the faces were getting too dark (as, perhaps, in #916) so he blended the red channel into the others, Lighten mode, on a luminosity layer. My view is that this washed out some of the faces at the head table, but moreover the U.S. flag should have been excluded from the blend, because its stripes came out bizarrely light. See #919 for a different strategy. 

919 Chosen for the par version. For anyone feeling that some versions are making the fleshtones too bright, this version should do the trick. This person decided that the color issues were complex enough that he would prepare one version to deal with the various casts and then take it from there. He felt that his faces were too dark, but worried that lightening them would cost too much detail. This is, of course, the same problem faced by the person who did #918. Both concluded that they needed some channel blending to lighten the faces since curves would not do what they wanted. In #918 the person did this by blending the red into the RGB, Lighten mode, Luminosity layer. Here, instead, the person in LAB, blended in A into the L in Overlay mode. This lightened the flesh, darkened the trees, and did nothing to the whites, just what was wanted. I like this version better than the par.

920 It is dangerous to fall in love with technique because it so rarely loves you back. It is reasonable to try to direct attention to the head table, where the wartime survivors are. It is not reasonable to try to add such an enormous light source that anybody who actually attended this dinner would know instantly that this version has been Photoshopped to death.  The idea of spotlighting the head table was carried out subtly and successfully in #912 and to a lesser extent in #926. Not here.

There are, however, two consolations for the rest of us. First, that this spotlighting is wildly exaggerated doesn’t mean it isn’t useful. Think back to our Veiled Bride case study, and recall a “ringer” version, #319, which by itself was quite ugly and exaggerated. But I threw it out as a teaser, saying that almost every other version would be improved with a blend of 25% #319, and other list members verified that this was true.

The same is true with this monstrosity. As I posted to the main thread, blending 15% of #920 into each other version resulted in improvement in all but four.

The second useful hint: People has trouble finding an adequate start point for their correction because the original was so dark yet had light areas. This person tried several things and in pure frustration used the ancient Equalize command, presumably on a layer at about 50% opacity. I, of course, turn up my nose at this but must admit that it only takes a second or two to find out whether it works, and if it does, why not use it?

921 Here’s an intelligent way to produce a nighttime look without making the people look foreboding. This person created a reasonably colorful and correct group and flags, but then the image was overall too light. So he artificially darkened both the tablecloth and the trees.

922 Here careful attention was paid to maintaining neutrality. This causes the appearance of a neutral scene without a variation in lighting from one side to the oehter, yet the faces of the two men at left are somewhat green while the fleshtones of the older men at the head table are beet-red. Also, I say that this is too much of a daytime scene, too light in context.

In answer to a query about this image the person states he added stars to the sky.

923 Too contrasty.  Hair has become black. So have areas of the flags that are known to be blue. Faces at the front table appear to be posterized.

924 Too light overall. Whites lack detail, faces too. You can tell how tonal range is emphasizing the wrong things by looking at the wine bottles and the camera on the table. These things have excellent detail in this rendition, but unfortunately nobody is interested in it.

925 Chosen for the par version. A delightful combination of relatively dark people with warmth in the fleshtone by Robert Wheeler, who has described what he did in the main thread.

926 The basic problem with this near miss is that if the faces are this dark (certainly possible) then they can’t be this red. One nice touch: the faces at the head table have been lightened separately, creating somewhat of the same spotlight effect found in #912.

A friend submitted a somewhat similar version but I didn’t show it here because he had rotated the image. Since he had the exact same problem, I’ll duplicate here what I wrote him offline:

Your treatment is excellent in all respects except the big one—range in the important objects. The lightest parts of the image are of course the floodlights and the candles. But you are not supposed to take them into account when setting range because they are light sources themselves. Instead, somewhere in the white shirts should be a very light point, and you have not got that, so the whole effect is too flat.

To fix:
1) Open a curve and click the white eyedropper wherever you find the whitest fabric, like in the tall guy standing at center. 

2) Use a Blend If so that this doesn’t damage the candles or the floodlights.

3) Darken the midtone to your taste, needed because the first step lightens things too much. (Probably not needed in #926)

927 This is more like what #926 should have been, with a full range in the whites yet the candles have been preserved. This was done in Photoline rather than Photoshop and, according to the individual, took 30-45 minutes to produce.

928 I find this a little gloomy for such a joyful occasion because the faces are so gray. There is also objectionable noise in the faces at the head table.

929 This version came from an ACT class and I find it somewhat successful. The person spotlighted the head table, as was done in #912 and #926. I don’t think he cared that the left side is rather blue because that further emphasizes the older people at the head table. If this were a tad lighter overall it’d be pretty competitive IMHO.

930 Chosen for the par version. The overall effect is quite pleasing. My only quibble: John Lund, like many of us, thought the trees were distracting, so he darkened and vignetted them. Quite sensible, as we’ve seen elsewhere, but I think the move implies that the midtone of the image overall should get a tad darker, otherwise the people are a bit too light in context. The move would help fleshtone detail as well. John did most of the overall adjustment in Lightroom and then used a modified PPW, including H-K, with a lot of individual tweaks. He writes:

I found this image to be less challenging than it looked at first (or maybe I just did a poor job? - we shall see)…

[After completing my submitted version,] for fun I did another, starting with the original jPeg un-modified, and used only conventional Photoshop layers - CRVs, etc. - to get to a pretty good version quickly. Both this one and the re-processed in Lightroom versions are about "85% as good" as my final image described above. And they took maybe 15-minutes each, while I put in at least an hour & a half for my submission. But I like the latter much better with its cleaner color with more subtlety, better contrast & a more natural look. And it was more fun playing with various techniques & options, always a good learning experience.

931 The par.

Join colortheory@groups.io to automatically receive all group messages.