Seated in Grass: Comments on individual versions

Dan Margulis

Here are comments on the individual submissions in this case study. As with a couple of others, there were so many good versions that nobody should take my quick choices for the par version too seriously, as there are around ten versions that could reasonably have been chosen.

I also suggest that if you are going over this list you download them and compare some versions by toggling back and forth, using Apply Image or similar. Some of the differences are quite subtle and hard to detect otherwise (which doesn’t make them any less useful).

Dan Margulis

1001 This person did a series of blending maneuvers with the objective of adding contrast and detail. That certainly shows up in the face, but also in the grass. The question is whether the model will like the exaggeration of her freckles. Nevertheless the color is excellent, in my view it’s better in that respect than the par. 

Part of the reason the color is so appealing is that it’s correct in the context of the skintone. Compare this one to 1002, where the skintone is portrayed as lighter and also pinker, which would be correct under the circumstances. So the colors of these two would not be interchangeable. However, I would say that blending 1002 into 1001 in Lighter mode, 50% opacity, would give something superior to either.

1002 Here is the three-minute version that I did in 2017.  By coincidence, my direct competition is the next entrant.

1003 The average of the five MIT retouchers. I trust you will agree with my assessment that #1002 is better. The face has more range because the undershirt is lighter while the shadows are at least as dark. There was also a mild channel blend which gave more structure to the face, plus MMM for color variation. A consequence of the darker undershirt in #1003 is that all colors are subdued because they have more of the contaminant. By looking at this version, you wouldn’t know that the woman has red hair, which is obvious in all our good versions.

1004 I have no comment on this one. I fail to see what is so difficult about a requirement that files be submitted in sRGB. I already rejected one file that came in with the wrong profile last night. That’s when this one came in, too, but I didn’t see that it’s sRGB with a 1.4 gamma, IOW a false profile. So, if you want to compare this image to another, you have to convert it, and I once again ask that everyone double-check that they have the right size and the right profile before submitting.

1005 Very pleasant color, could use a little more depth in the face if that can be done without losing too much of the very nice hair color.  Paco Márquez used a simple approach here, he retouched out some of the facial imperfections but did not go overboard. The only unusual thing: he felt that the jeans were too light and so created a second version to darken them.

Many of us believe that it was beneficial to have the grass unrealistically colorful. Paco defends the practice succinctly: “The greens might be a bit on the brightly saturated side but I feel that enhances the separation of tones (G&R) and the overall brightness.” In his post to the main thread, he comments that he tried to move the face toward pink, thinking that this is appropriate for redheads. Given that decision, he was wise, I think, to make the sweatshirt more colorful than most others did. It is an insurance policy against people deciding that he had made the skintone too beet-red, and it helps us distinguish the long hair from the shirt.

1006 And speaking of general approaches, Arthur Margolin also speaks for a lot of us: “I only did a modicum of retouching of her face to smooth over blemishes and other irregularities because it appears that in the original she isn’t wearing make-up and I sought to retain a ‘natural’ look as much as possible.” His approach was to do five separate versions and blend them intelligently.

The result is a face rather similar to #1005, but slightly lighter eyes (good), bluer grass (depends on the mood you’re trying to set, I’d describe #1005 as friendlier), and darker sweatshirt (not so good, we lose the long hair). But the real reason I prefer #1005 is that here we don’t really see that she has red hair, and I suspect she would not be happy about that.

1007 Here’s the best individual version from the five MIT retouchers. They all got about the same result, which is why this one is so similar to #1003, the average of the five. I am happy to see that a healthy majority of us are getting better results than these two.

1008 Much too blue, measurable in the undershirt (sharply cyan) and the very desaturated fleshtone.

1009 Like #1001 and #1005, this version emphasizes yellowness in the background, but this person extended the yellowness to the face as well. He says that all the basics here are simple but more complicated moves are not so easy. He did not use MMM, because he felt it was exaggerating the freckles too much, so he compensated with more Color Boost. The result is something that is not as balanced-looking as #1005, yet more dramatic and some people might prefer it, particularly if the eyes were lightened.

1010 A relatively conservative but effective version, using a simple approach that included H-K and a luminosity blend. He stays away from the gaudy greens of some of the earlier versions; they would not be appropriate here because they would conflict with the relatively gray face.

1011 Chosen for the par version. This is pretty much the direct opposite of #1010, it’s about as playful a treatment as we might accept, whereas #1010 is about as conservative as we might accept. Therefore, it is a solid bet that blending the two together 50-50 will produce a result better than either. And this is indeed the case. This person also avoided MMM, thinking that it was too damaging to the hair. But as to the grass, he adds his voice to the anti-reality group: “In my final version, the background is very bright with louder color than expected in nature. Although some may argue this might compete with our subject, I concluded (after fiddling with multiple possible adjustments) that the bright color contrasts well with her red hair and perhaps adds a sense of excitement. Looking back, becoming less timid about stronger colors has been one learning area for me while participating in the challenges (sometimes that is good, maybe not always). I will be interested to see how others manage the background.”

1012 If you’d like to continue the debate about realism in the background you might wish to toggle back and forth between this one and #1015. The woman and the sweatshirt are very similar in both. The big variation is in the grass: here we have realism, in #1015 a bright yellow fantasy. If I have to choose between the two I guess I pick #1015, but the real winner would be blending #1015 into #1012 in Color mode. This tells me that having relatively dark grass is fine, but it has to be fairly colorful.

1013 And here comes the disagreement. Rather than jack up the greens, should they be suppressed? Here’s this person’s answer:

Since this is a casual portrait and the lady has freckled skin I don’t think it’s necessary to retouch the skin in any way. However, while acknowledging the freckles we don’t need to enhance them to a point where they become distracting. The eyes are again important and require care. The green background is very distracting and needs to be suppressed. In this photo the face and eyes will make or break the result. Unlike the wedding dress in the veiled bride the clothes here are not so important.

Given these objectives the person achieved them well, as can be seen by toggling this one with #1010. both have conservative flesh color, dull grass, a dark sweatshirt, and similar treatment of the hair. As he says, #1013 is much preferable because of the face and the eyes. But the grass and the sweatshirt seem to me more important than this person thinks.

1014 The skintone seems too light and the freckles too prominent, seemingly the result of blending strong contrast-enhancing curves to the green channel. The color choices are similar to those of #1013, which is a much darker treatment. Since this person presumably doesn’t want to go that far, blending 50% of #1013 into this one would likely give him what he wants.

1015 Chosen for the par version. Simple procedures lead to this attractive result. I prefer this one to #1013 on the grounds of overall feel with some reservations. This person tried to get range into the face by lightening the teeth almost to the point of a highlight. I prefer the treatment in #1013. And that one makes clear how important it would have been to engineer better eyes into this one.

1016 This person did not actually try to suppress the greens, but when he was in Color Boost he excluded them, not wishing them to become unrealistic. He applied a lot of MMM to the face, and treated the eyes well. Granted the relatively dark skin it should probably bit a tiny bit more yellow and less pink. So I would compare it to the lighter #1002, which I also consider somewhat too pink. The two are trying for different goals and a lot of personal taste is involved, however the comparison leaves me with the feeling that the face in this version is in shadow. Many of our versions have grass distinctly darker than the face and others have it as lighter. Both styles can be effective, but having the grass at roughly the same darkness as the face doesn’t seem to work unless both sides are brighter than what we have here.

1017 The color choice is reasonable and conservative. The person chose appropriate highlight and shadow points in the woman’s clothing. But he selected the background and darkened and added contrast to it. That might have worked better if he had done some blending or H-K in the face to give it more shape. As it stands the background seems better defined than the face is.

1018 To read this person’s description of his workflow is to know the meaning of the myth of Sisyphus. Like most of us he found no basic color problem and concentrated on building a decent face. Every time he tried something else he thought it made the face seem out of place in context, and he began to curse out Chevreul. “And so of all the case studies,” he writes,  “I ended up with the least satisfactory entry.  At least up to now I've felt pretty good about my entries (until seeing the others' entries
12 hours later).”

I have no idea why he thinks this. I see no major problems here, it’s a fine version.

1019 Rich and attractive colors. When I first looked at it I underrated it, I thought the person had retouched out most of the freckles. According to his text he did not. One minor point that bothers me: he whitened the teeth. Because they are surrounded by rather yellow flesh, the teeth actually look blue to me (they’re not, according to the info palette, they’re 1a7b, but everybody else has them warmer and more yellow).

The quirks of the random-number generator placed five similar images very close together. This version and four of the next five (excluding only #1021)  are the only ones of the entire set that use a background that is dark, saturated, and bluer than most others. I haven’t previously compared these five to each other but probably I should, because this color choice for the background is reasonable but it imposes certain obligations on the rest of the image. And all five of these are excellent versions, without any obvious error.

So I put the five up side by side and the big differences are that #1020 and #1023 are lighter than the others, and that #1020 and #1022 made little if any effort to make the background interesting whereas the other three put in accents of some kind. #1024 has the best detail in the face; #1022 is the most conservative of the three darker ones and therefore fewer people will actively dislike it than any of the other four. #1020 and #1024 have the most interesting eyes.

I’d say that the lighter faces, #1020 and #1023, would work better with some of the versions with much lighter backgrounds, while the other three would look out of place. As for the background I can understand why the instinct would be to downplay it and maybe in some of the lighter versions that would be the right call. Here, though, I think I prefer the added pop in the grass of #1020 and #1024 rather than the blurrier #1022.

1020 Another person already commented in the main thread “With a mild vignette, blurred background, and smoothed skin, entry 1020 could easily be from a professional photo studio and many clients would like it well.” This is probably true, and the person took a relatively simple way to get there. The only questionable move: he selected the background grass and made it cooler, bluer. This makes the skin look yellower by comparison, which is not good when it’s this light. Better IMHO to leave the background yellower, which would drive the face toward pink. The question of whether this ought to be treated a studio shot, which tend to be softer and more conservative, or as a candid of a healthy woman in the outdoors remains open.

1021 Chosen for the par version. A combination of good moves created something quite akin to the par version itself. He used the Velvet Hammer and MMM + CB, did some retouching (including, I’d say, the eyes, but I don’t find that in the description). He got to the point where the skin was slightly dark and slightly blue, whereupon he came up with two clever moves. First, he used Channel Mixer to blend the red into the blue, Luminosity mode. This both lightened the skin and reduced the freckling, which was very pronounced in the blue channel. Then he applied a yellowing curve, but limited it to darker areas. This made the eyes and hair more interesting, as well as presumably giving a slightly sunnier appearance to the skin.

1022 Chosen for the par version. Done outside of Photoshop, I commented on this one in my notes to #1019. The two have similar darkness and shape in the face. #1019 is more vivid and therefore more exciting, but this one, being more conservative, seems to portray more personality. Originally I liked it better than #1019, now I have a tougher time making up my mind. If this one had been given lighter eyes I’d prefer it for sure.

1023 Like most of us, this person found little fault with the original color. He then selected the foreground and background for separate work. The effect might have been better had he lightened the lightest part of the grass so as to add contrast to it and make it seem lighter overall. As matters stand we might find the face too light for the overall scene.

1024 Chosen for the par version. This is my current try, not terribly different from the #1002 that I did three years ago. I wanted to add depth to the face, but my preferred method, blending the green into the red, would have put me in Frecklesville. So I substituted Velvet Hammer and a couple of hits of the luminosity half of H-K. The color was fine, as it was from the outset, but I was losing the redness of the hair, the eyes were too dark, and I was still exaggerating the freckles too much.

So, on a Luminosity layer, I applied Dust & Scratches at a high threshold to the green channel only. This reduced the mottling without losing the basic texture of the skin. I also blended the red into the green in Lighten mode. This was both to lighten the skin and to reduce further the freckling. I assumed I was going to have to exclude the sweatshirt, which would get lighter as a result of the blend, but when I saw the result I decided to keep it that way. The lighter shirt makes the long hair stand out better. In many versions it’s hard to tell that this woman has hair draped over her shoulder.

Finally, I made a grossly light and colorful version, which I used to paint in the eyes and to add highlighting to the hair

1025 Here is what #1024 would look like in a sunnier setting. Both versions have excellent shape in the face. But since the grass is much lighter in this one it follows that the face should be as well. It’s nice and realistic and sure to please the client for the same reason that #1022 is: quibble about the background if you like, but there’s nothing to argue about in the foreground, nothing that the client can complain about,

1026 As with #1004, no comment on this one because it was sent with the wrong profile.

1027 No facial detail here. It looks like the woman was born without a nose.

1028 The par version.

Join to automatically receive all group messages.