PS LAB Color, Chapter 16 -- Recipe for portraits

13

Comments

  • warszawskiwarszawski Registered Users Posts: 33 Big grins
    edited November 22, 2005
    my subjectif edit - better late than never
    the original is here http://www.dgrin.com/showpost.php?p=200832&postcount=51
    and my result is shown here
    all edits are in LAb, i did not succeed to enhance this picture with the workflow of Chapter16, so i did my best and wanted to share it here
    (in next message i'll post my workflow)
    ... better late than never ...
  • warszawskiwarszawski Registered Users Posts: 33 Big grins
    edited November 22, 2005
    workflow
    .... did not succeed to enhance this picture with the workflow of Chapter16, so i did my best and wanted to share it here
    in this message i'll post my workflow

    1° to my eye the most "important edit is curve layer with multiply (no edit in curve) , it brings the "pop" to colors to the man, i had to mask the face a little bit

    2° secondly i had to desaturate background because of step 1

    3° i did a dodge on face, YES, I HAD TO, but with 50%Grey auxilary layer, which is not destructive and so manageable with opacity

    4° high pas filter to bring a local contrast & ? sharpness
    ... better late than never ...
  • aporiaaporia Registered Users Posts: 145 Major grins
    edited November 22, 2005
    rutt wrote:
    [size=+1]Date for practice set #1 comparisons?[/size]
    . . . What times are good/bad for people who want to do this? This evening could work, but maybe too soon? Tomorrow evening is out for me. Thanksgiving is probably bad for everyone. Friday late afternoon or evening can work for me.
    Sick today. Friday works for me. Afternoon is best but with a little bit of notice, I also could do later in the evening (EST, GMT -5).
    Tom in Niagara (CAN/US)
    Real Body Integrated Arts
    GMT -5
  • pathfinderpathfinder Super Moderators Posts: 14,694 moderator
    edited November 22, 2005
    Friday evening will probably work for me too. Holidays make for tentative on-line plans ne_nau.gif

    Your edit of the little blonde fella looks very nice. :):
    Pathfinder - www.pathfinder.smugmug.com

    Moderator of the Technique Forum and Finishing School on Dgrin
  • bugziebugzie Registered Users Posts: 30 Big grins
    edited November 24, 2005
    a little over done.
    i've been dropping and lurking on these discussions. :D

    many years ago i heard a very important piece of advice. which is: with photoshop, do what you feel needs doing and then *halve* the effect.

    this image looks just plain oversaturated. the sharpening is too harsh. it's lost something of the original. i think the lab techniques being discussed are worthwhile -- i've been trying some. but this is an example of where it's overdone.
  • warszawskiwarszawski Registered Users Posts: 33 Big grins
    edited November 24, 2005
    so OVERDONE ?
    bugzie wrote:
    ... but this is an example of where it's overdone.

    my edit is a workaround of LAB technics that should work but none of LAB moves were satisfying "in my hands" for THIS PICTURE, so i did in Lab only "multiply" trick for deeper colors ...

    i like a tonal separation between foreground and backround, this "gives" more importance for the men (?)

    Somwhere I had to halve my edit , i did not know where and how.

    Do you really think that my edit is so OVERDONE, what others think about it ?
    ... better late than never ...
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 24, 2005
    A suggestion from Dan: keep the a dupliicate of the original before you go off and do whatever you do. Then layer it on top of your edit and play with the opacity slider. This is a really easy and direct way of doing half.

    I'm not even sure which image bugzie things is too much.

    Color preferences are very personal and cultural. Too much for some is too little for others. Print requires more than monitors.
    If not now, when?
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 24, 2005
    Here's my approach to one of the early images posted by Rutt.

    I confess to a certain ambivalence regarding this recipe. It reminds me of the old "Mission Impossible" TV show, where the final result depended on a precise chain of events (scams actually), each one of which required the complete success of all the preceding events, clearly a violation of Murphy's Law.

    I've worked on this image four times, each time taking a different route, and I've gotten four different results. I think each technique here is a valuable tool. Whether the combination and sequence of all of them together constitutes a "tool" in itself, I don't know. I'm as suspicious of the possibility as I am of any agenda that answers all questions at once. For instance, on first examination, this face would seem to be a good candidate for some Man From Mars curves, given the overall lifelessness of the colors. And the A and B channel overlay moves which are essential to this recipe pose some severe problems for the cyan levels. And I must confess to not being at all won over by the HIRALOAM step at the end. But here's what I came up with.

    Before_After.jpg

    Before starting out, I wanted to see what was happening with the skin tones. A reasonable sampling of a well lit, unfreckled patch to the left of her nose reads around 15C 35M 35Y. Not bad. A little cool, but since the later moves are guaranteed to ransack the cyan values, I thought I might get away with it.

    This is the layer palette for my work in LAB:

    lab_layers.gif

    1. I started with the Green channel as a luminosity overlay and that trick was as good as any, better than most. I simply copied the green channel into a new layer and set the layer to luminosity mode.

    2. It definitely helps the lips and gives overall shape the to the face. But it also darkens everything a little, so I added a couple of extra contrast curves after taking the image into LAB. If this were in CMYK, now would be a good time to strengthen the black channel. In LAB there is no black channel, and the Luminosity channel is, as Dan points out, like a sledge hammer when subtle moves are called for. So I merged to new layer and duplicated this layer to a new image which I converted to CMYK. Now I had a real black plate which I copied back to my active image and tuned into a layer mask (inverting first). This restricted my luminosity moves to the actual "blacks", allowing a much stronger move.

    Black channel (inverted) as layer mask:

    BlackMask.jpg

    For my second curve, I CMD-Clicked on the composite icon in the Channels palette to turn the values into a selection, which became my second layer mask, for a lightening move. (This mask is essentially a greyscale version of the image). Since I'd already dealt with the shadows, I locked them down in this curve.

    Here are the contrast curves I used, and the result:

    ContrastCurves.gif
    step3.jpg

    3. This isn'tt bad, as is. There is nice crisp contrast that doesn't look overdone. But the colors are still nowhere. Just for kicks, I tried a MFM curve with the A and B channels, leaving Luminosity alone. I wanted to see some variety, and after playing with the midpoints and anchors I ended up with these curves, producing this result. Both anchors and midpoints were tweaked to bring out the greens and blues as much as possible, without turning her into a Vulcan or something.

    mfm_curves.gif

    Reducing the curve layer to 30% gives the result on the right. If this were Chapter 12 I might play like the Marines, declare victory and pull out. But MFM curves aren't part of this chapter, even though they're probably part of any realistic work flow.

    MFMCurves.jpg


    4. So I turned off the MFM curves and turned to the A and B channel overlay operations. I merged everything else to a new layer, which I then duplicated. I wholly endorse Rutt's approach to this step, placing each overlay operation in its own layer, and so these two new layers became "A" and "B" and the respective channels were overlaid into themselves with Image>Apply Image in overlay mode at 100%, with the B channel on top. I played with different settings but decided that 62% for "B" gave the best result, favoring the yellow tones somewhat.

    Overlay_Full.jpg

    Except that I don't like it. I'm suspicious of this operation anyway. It feels a bit kludgy, a heavy-handed way to accomplish steps that are available in other forms, with greater flexibility. The handy feature here is that it can be placed in an action without giving it much thought.

    The biggest problem that I see is the fact that, since the luminosity stays the same, and by definition the magenta and yellow have to increase, there's no place for cyan to go but down. I'm getting readings of 0, 1 and 2 all over the face which feels way too low. I realize that we're not actually printing on a press and so we don't have to worry about issues such as minimum possible dot, but the CMYK values are still a reliable guide for the overall color of the skin, no matter what the actual color space.

    5. The MFM version was definitely cooler so I tried putting that curve layer on top of the A/B layers and reduced its opacity to 15%. I put the A and B overlay layers into a Layer set and pulled the set's opacity down to 60%. I think the combination works better.

    Combo.jpg

    6. Then on to the sharpening. In the full-size version I merged everything to a new layer, and called on my black channel mask once more to mask the merged layer. I applied Unsharp Mask at 211 amount, 2.5 radius and 0 threshold. This more or less approximates the Black Channel sharpening move that Dan recommends for CMYK images. But at this reduced size image, the effects are negligable, so I simply sharpened the merged layer's luminosity channel. I like to select the channel in the channel palette, then click the eye icon "On" for the composite layer. This allows me to see the actual image and any modifications I make, even though the changes are only affecting the luminosity channel. I pulled the sliders around a bit and wound up with 118 amount .9 radius.

    Sharpened.jpg

    7. HIRALOAM: don't like it. I'm certainly willing to be converted, but I don't see any benefit, particularly in the reduced size image. My original at full size didn't offend me as much, but I still couldn't see it as the kind of essential move that MFM curves have proven to be.

    8. At this point I converted back to RGB and still worried about my Cyan levels. Maybe I'm compulsive, but I felt it was still a bit harsh and hot, a problem that seems to charactize this entire process. So I placed the original image in a layer on top, in color mode, at 20% opacity. I also placed a Selective Color layer above that and added +5 Cyan to the Reds. This seems to do the job, which is to improve the image without making it look like a major move was made into a different color context. This strikes me as the image the way it was intended as opposed to a completely different image, one simply based on the original.

    Final_RGB.jpg
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • aporiaaporia Registered Users Posts: 145 Major grins
    edited November 24, 2005
    Once again, great discussion and examples in this thread.

    Thanks for the detailed description and commentary, Crawford. It's very helpful to see other color perspectives well rendered. thumb.gif
    Tom in Niagara (CAN/US)
    Real Body Integrated Arts
    GMT -5
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 24, 2005
    Crawford (AKA Edgework), it's great to have a real pro here. I especially like the color blend with the original and selective color at the end.

    For the record, I captured the curves which are equivalent to Dan's self overlays. You can pick them up here: http://www.chezrutt.com/rutt/overlaycurves.acv

    Both the A and B curves look like this:

    45856605-S.jpg

    Very steep near the center and flattening out at the ends. An alternative to using the overlays is to start with these curves and customize. I suppose using the MFM technique is similar to this.
    If not now, when?
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 25, 2005
    rutt wrote:
    Crawford (AKA Edgework), it's great to have a real pro here. I especially like the color blend with the original and selective color at the end.

    For the record, I captured the curves which are equivalent to Dan's self overlays. You can pick them up here: http://www.chezrutt.com/rutt/overlaycurves.acv

    Both the A and B curves look like this:

    45856605-S.jpg

    Very steep near the center and flattening out at the ends. An alternative to using the overlays is to start with these curves and customize. I suppose using the MFM technique is similar to this.

    I wonder. Since overlay screens and multiplies depending on whether the tone is above or below 50%, I'm not sure about the flattening off at the edges. Seems like brighter and darker colors would affect each other more severely. I saw the actual mathematical calculations that are performed on the pixel values once: didn't understand them for squat, other than to realize that there's nothing linear or anything that plain arithmetic can explain.

    The thing that attracts me so much to the MFM curves is the way they can be customized, all the way to forcing greens and blues into an image that's primarily hot and dry.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 25, 2005
    Idiot that I am. I take all that back. I've been using Overlay noise layers for years, precisely because Overlay mode stops working as you move to highlights and shadows. So of course the curves look like that. Sometimes you end up with technique confusion.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 25, 2005
    edgework wrote:
    Idiot that I am. I take all that back. I've been using Overlay noise layers for years, precisely because Overlay mode stops working as you move to highlights and shadows. So of course the curves look like that. Sometimes you end up with technique confusion.

    I used an empirical experiment to make the curves at first and then read the mathematics behind it somewhere. But I love the idea of substituting MFM curves at this point in the process. Really the overlay curves are just MFMs curves of the feint of heart.
    If not now, when?
  • warszawskiwarszawski Registered Users Posts: 33 Big grins
    edited November 25, 2005
    what does mean MFM curve?
    rutt wrote:
    ....But I love the idea of substituting MFM curves at this point in the process. Really the overlay curves are just MFMs curves of the feint of heart.


    Sorry men, please be gentle with me, I'm lost : what does mean MFM curve ? ...
    ... better late than never ...
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 25, 2005
    warszawski wrote:
    Sorry men, please be gentle with me, I'm lost : what does mean MFM curve ? ...

    It's chapter 12 talk: http://www.dgrin.com/showthread.php?t=21859
    It stands for Man From Mars. There is a good example there. Also Edgework showed a good example in his treatment of my portrait just above here.
    If not now, when?
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 25, 2005
    Edgework, I'm loving the suggestion of using MFM instead of overlay curves. I have a few ballet shots where it really did just what I wanted and more.

    I didn't think I could save this:

    45894200-L.jpg

    But the ugly duckling became this swan:

    45894354-L.jpg

    using essentially Dan's workflow with MFM substituted for overlay curves and a final trip to CMYK to establish black without plugging.

    One thing, here is an advertisement for HIRALOAM. Here it is just after conventional USM and before HIRALOAM:

    45894264-L.jpg

    And after HIRALOAM:

    45894083-L.jpg

    Download the originals (substiture O for L in the URLs) and compare A/B and you'll see. It's very dramatic at largish sizes.

    Here is another one I didn't think could be saved, before and after the same workflow:

    45896908-L.jpg

    45897002-L.jpg
    If not now, when?
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 25, 2005
    Edgework, I'm loving the suggestion of using MFM instead of overlay curves. I have a few ballet shots where it really did just what I wanted and more.

    Really nice shifts! The clothing, in particular, has a whole new lease on life.

    I think if I had only extended my use of curve enhancements (which really dates from Edition 2 of Professional Photoshop) into the full context of MFM curves, the book would have been worth every penny. Everything else is embellishment, good things to know and use as part of the work flow, when needed. But almost any image can profit from some measure of color enhancement (particularly the crap I get at work). I began using the technique immediately in my job and, along with cast removal techniques, I find I do the major portion of my color work in LAB these days.

    As your overlay curves show, that operation is perfectly symmetrical; whatever condition you start with will simply become more pronounced. MFM curves give you a chance to control the cast completely. The true goal, of course, is to simply be fluent with the curves, period, without having to subdivide them into this approach or that approach. I feel like this is a good start.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • aporiaaporia Registered Users Posts: 145 Major grins
    edited November 25, 2005
    Amazing save on these photos and great description of the essential features of color retouching. I'm looking forward to trying the MFM curves in the LAB process on some of my own photos. In the meantime, I'm still catching up in reading Margulis (Chap 6) and scratching my head around color theory -- particularly in reading a/b values and shifting the numbers appropriately.

    Tough learning curve here but I'm finding it extremely helpful in following the discussion and summaries as much as possible.

    Here's a link to my own attempts in the Margulis process from a few days ago. I don't know how helpful they will be. The more I read and practice color theory and photography, the more naive I think I was yesterday. headscratch.gif
    Tom in Niagara (CAN/US)
    Real Body Integrated Arts
    GMT -5
  • bugziebugzie Registered Users Posts: 30 Big grins
    edited November 26, 2005
    the image bugzie thought was overdone
    rutt wrote:
    Opinions?

    45564702-L.jpg
    i'm sorry for the confusion, i'm new at posting and didn't know the posts wouldn't thread. this is the image i thought was overdone. too saturated. it's lost the childish new-skin softness and smoothness of the original. he looks like he's been spray on tanned. but i do understand you can blend back into the original. but it's something to keep in mind with kid's skins...
  • dandilldandill Registered Users Posts: 102 Major grins
    edited November 28, 2005
    Photoshop cheat sheet in support of Margulis's moves
    edgework wrote:
    Here's my approach to one of the early images posted by Rutt. ...
    Working through the steps detailed in this very helpful post, I realize that while the steps are clear, exactly how to carry them out sometimes is not always clear to me. Here are some example steps

    • Copy the green channel into a new layer and set the layer to luminosity mode.
    • Duplicate a layer to a new image
    • Copy the real black plate of a CMYK image to an active image and turn it into a layer mask (inverting first).
    • Overlay a channel onto itself
    I have figured out how to do each of these, but I suspect my solutions are not the most direct. For example, in another thread, rutt comments
    • "I made a copy of the red channel by dragging the red channel down into the duplicate icon. Then converted to LAB. Made a duplicate layer, luminosity blend. Apply image of that copy of the red channel to the duplicate layer's L channel."
    Now, that is really neat to know, but I wouldn't have thought of it myself.

    So, would it be worthwhile for people who are expert to contribute to a cheat sheet describing how to do the core Photoshop operations needed to carry out Margulis's schemes.

    That said, if how to carry out these kinds of moves is obvious to everyone, please forgive this post! Also, if this post is better in another thread, if someone can move it there, that would be appreciated.
    Dan Dill

    "It is a magical time. I am reluctant to leave. Yet the shooting becomes more difficult, the path back grows black as it is without this last light. I don't do it anymore unless my husband is with me, as I am still afraid of the dark, smile.

    This was truly last light, my legs were tired, my husband could no longer read and was anxious to leave, but the magic and I, we lingered........"
    Ginger Jones
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 28, 2005
    dandill wrote:
    Working through the steps detailed in this very helpful post, I realize that while the steps are clear, exactly how to carry them out sometimes is not always clear to me. Here are some example steps

    • Copy the green channel into a new layer and set the layer to luminosity mode.
    • Duplicate a layer to a new image
    • Copy the real black plate of a CMYK image to an active image and turn it into a layer mask (inverting first).
    • Overlay a channel onto itself
    I have figured out how to do each of these, but I suspect my solutions are not the most direct. For example, in another thread, rutt comments

    Anything that is copied onto the clipboard can be pasted anywhere else, in Photoshop. So in this case, in the Channels palette, clicking on the green channel will show only that grayscale image on the screen. Cmd-A selects the entire image frame. CMD-C copies the channel. Now, going to the layer palette, creating a new layer gives you a transparent frame, and when you CMD-V to paste the contents of the clipboard, the information that was your Green channel now appears in your layer. Of course, layers incorporate all three channels so now you have a neutral RGB version of your original grayscale Green channel. Going up to the blending modes button, which says "Normal" by default, and selecting Luminosity will apply the dark and light values of the new layer to the color information of everything below it.

    Using the black channel as a layer mask in a different file is actually a different application of the exact same principle. When I want to make a duplicate of a file to access some feature of a different color space, I'll first merge to a new layer in the existing file. Then, in the Layer palette options select Duplicate Layer. In the dialogue that pops up, under Document, you will see the current file name by default. Click on that and select "New" to create a new file from the layer. In this case, my original file was in LAB, but I wanted to access the black plate from CMYK, so I converted the duplicate file first to RGB, then to CMYK (you'll understand why that's important when you read the book). Now, in the channel palette, select the black channel. Same operation here as last time. Select the entire image, copy the channel to the clipboard and return to the LAB file. Since I already created a merged layer, that will become the layer that I sharpen. I hit it hard, harder than looks good, hard enough to look oversharpened. Now I create a layer mask which is white by default (nothing masked). I click on the mask icon and paste. That fills my mask with the contents of the black channel that I just copied. Keep in mind that ink channels are dark precisely where you want ink to appear, so the "positive" version of the black channel is correct, if it's being used as a genuine channel. In this case, however, we want to mask out everything BUT the black informtion. So now, after pasting the black channel into my layer mask, I invert it with CMD-I. This negative version reveals only the "black" information.

    Many of the blending operations use the Apply Image command found in the Image menu. When you "apply" an image, whatever layer or channel is active at time becomes the base image. In the Apply Image dialogue you have the option of specifying both channel and layer for any image that is opened. That will be the image that is "applied" to your active image, which is to say, it will be blended according to whatever parameters you set. For example, 100% Normal mode will completely replace the Active image with the Apply Image. The operation I referred to involves applying the color channels in LAB to themselves in Overlay mode. Click on the channel, either A or B, to make it active. In Apply Image, choose the same channel as the apply image and choose overlay. The operation essentially pushes darks darker and lights lighter, but, since these are color channels in Lab, the result is to intensify the colors that are already there.

    Hope that helps.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited November 28, 2005
    I was thinking that this was a good place for a little mission statement. Our purpose here isn't to replace Dan's wonderful book. Our purpose is to help each other work through it and master its techniques (and improve on them where possible.) Go buy yourself a copy. You'll be glad you did. Dan is fun to read as well as being very useful.
    If not now, when?
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited November 28, 2005
    rutt wrote:
    I was thinking that this was a good place for a little mission statement. Our purpose here isn't to replace Dan's wonderful book. Our purpose is to help each other work through it and master its techniques (and improve on them where possible.) Go buy yourself a copy. You'll be glad you did. Dan is fun to read as well as being very useful.


    Yep. We're just scratching the surface of Dan's book. It's a taste and nothing more. For the meal you need to get a copy.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • dandilldandill Registered Users Posts: 102 Major grins
    edited November 28, 2005
    edgework wrote:
    ... Hope that helps.
    You guys are wizards! Thanks very much.
    Dan Dill

    "It is a magical time. I am reluctant to leave. Yet the shooting becomes more difficult, the path back grows black as it is without this last light. I don't do it anymore unless my husband is with me, as I am still afraid of the dark, smile.

    This was truly last light, my legs were tired, my husband could no longer read and was anxious to leave, but the magic and I, we lingered........"
    Ginger Jones
  • dandilldandill Registered Users Posts: 102 Major grins
    edited November 28, 2005
    rutt wrote:
    ... Go buy yourself a copy. You'll be glad you did. Dan is fun to read as well as being very useful.
    "Fun to read" and "useful" are understatements. Dan's two books are the only Photoshop books I read with a highlighter!
    Dan Dill

    "It is a magical time. I am reluctant to leave. Yet the shooting becomes more difficult, the path back grows black as it is without this last light. I don't do it anymore unless my husband is with me, as I am still afraid of the dark, smile.

    This was truly last light, my legs were tired, my husband could no longer read and was anxious to leave, but the magic and I, we lingered........"
    Ginger Jones
  • edgeworkedgework Registered Users Posts: 257 Major grins
    edited November 28, 2005
    rutt wrote:
    I was thinking that this was a good place for a little mission statement. Our purpose here isn't to replace Dan's wonderful book. Our purpose is to help each other work through it and master its techniques (and improve on them where possible.) Go buy yourself a copy. You'll be glad you did. Dan is fun to read as well as being very useful.

    In the same vein, I think anyone who's worked with Dan knows that he's more than willing to be shown a better way to do something. Of course, the catch is you need to actually know what you're talking about because he does not suffer fools gladly. But this entire book is cutting edge in an area that no one realized was even there. His own ideas about LAB have evolved dramatically through the 4 editions of Professional Photoshop and now this book. The point is, there is a mutuality to the explorative process here. This is not so much a final product that we're trying to understand, but an ongoing process that has hasn't begun to be exhausted. Answers aren't so much waiting around for us to catch up with them; rather, they're generated in the act of stumbling around, trying this, trying that, trying something else, failing badly and trying again. Photoshop technique is like music theory or chess theory: it's simply the stuff that people have tried that, for one reason or other, works. Margulis is an undisputed grandmaster, but there's room for lots of useful activity at all levels.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • bugziebugzie Registered Users Posts: 30 Big grins
    edited December 1, 2005
    hey, i'm a convert.
    have 80 or so portrait shots to work on... so i tried some of the techniques on some of them. i blended them back into the original, mind you, and varied some of the amounts of blending. they look great. i need to take the few i've done to my digital print place and test them before i continue. (i'm in australia). and i've ordered the book. :-)

    thanks!
  • jfriendjfriend Registered Users Posts: 8,097 Major grins
    edited December 2, 2005
    Margulis portrait workflow dissected
    After several attempts at a retouch on the photo in this thread, I decided to finally give Margulis' portrait workflow a try. I had read about it in the book and here, but had never actually tried it myself.

    I did not succeed in using the whole workflow on this photo for a variety of reasons, but I did learn a ton and I now have a new workflow that uses the same techniques he writes about, but I find easier to understand, follow and adapt to various images. I'll first describe how his workflow went on this photo, then discuss how I execute it differently now.

    First, this is a retouched version of the photo I was working on:
    46704293-M.jpg


    Margulis' recipe calls for these steps. I've broken them out into more steps than he labels in his book - he just combines several steps into one labeled thing.
    1. Dup layer
    2. Apply green channel to the dup layer
    3. Change blend mode to luminosity
    4. Fix any areas that are overdone with blendif
    5. Convert to LAB and flatten
    6. Dup layer
    7. Apply A to itself in overlay at some percentage that works for this image
    8. Apply B to itself in overlay at some percentage that works for this image
    9. Lower opacity to taste
    10. Consider contrast enhancing L channel curve depending upon the image
    11. Sharpen
    In summary, steps 2-4 are to enhance contrast. You pick a contrasty channel and do a luminosity blend with that contrast.

    Steps 5-9 are a different implementation of somewhat normal LAB color enhancements.

    Step 10 is a normal contrast enhancement.

    Step 11 is a set of sharpening techniques he has.

    Here's how this went with my image. When I followed steps 2-4, I ended up with an overcooked part of my image on the red hair piece of the girl. It became dark. I tried Margulis' recommendation of using blendif to hide the changes from the dark red parts of the image, but that blocked the effect from parts of the face too (where there is lots of red) and I had ot get the blendif slider almost to the middle in order to restore the red hair piece. I concluded that that basically didn't work because I'd also removed the contrast enhancing effect from more than half the face and that's what we're trying to do in the first place. After futzing around a bit and looking a the other channels, I saw that the hair piece was extremely dark in the green channel (that's why it made the hairpice dark in the luminosity blend). So, if I was willing to give up on contrast enhancements in the dark part of the images (the hair mostly), I could block the luminosity blend if the lower layer had dark green. That restored the red hairpiece, though it blocked contrast enhancement from the dark colored hair (which also has a dark green component). But, it didn't block the effect from the face.

    I had never used the Apply Image command and it can be a little daunting to understand exactly what it's doing when you first try to follow this. After playing with this for awhile, I started to understand it and what this step is really doing. In a nutshell, it's the same as creating a new layer with only the green channel in it and setting the blend mode to luminosity. I don't know if that will be clearer for you, but this was an aha for me. OK, now I understand what we were doing. We're picking a contrasty color channel and simply blending it with the original image in Luminosity mode. That will just add contrast. You can try this yourself without using Apply Image. Take an image. View the green channel. Select All. Copy to clipboard. Go back to RGB view (all channels). Back to layers. Paste the clipboard into a new layer. Change the blend mode to luminosity. I think this is exactly the same thing as Apply Image in this case (please correct me if I'm wrong) except I can see what this is doing and the result is in a visible layer and I can change the opacity on it any time. And, more importantly, I can keep this change independent of other changes later.

    OK, so now I understand what Apply Image is doing and I have a (clearer and easier for me) alternate way of doing it.

    How we go to LAB mode. At this point, I've found that I don't like to flatten my image. I love adjustment layers. I love keeping separate adjustments independent. This may be because I'm just not good enough to avoid having to redo or adjust some previous adjustments and keeping them all on separate layers saves me a lot of work or maybe it's just a good practice to follow. In any case I want to preserve my layers and masks that I already have. But, there are many types of layers and blend modes you can't keep when you convert to LAB mode. So, you usually have to flatten.

    I've decided that I won't convert this image to LAB mode. I know that in my workflow I'm coming back to RGB when we're done so instead of flattening, converting this image to LAB and then flattening and converting back, I'm going to keep this image by itself, make a dup image that I flatten and convert to LAB and then when I'm done with the LAB edits, I'll paste a merged copy of that result back into my RGB file. Further, I'll even keep the LAB version of the image around until I'm really done with the image. That means if I ever need to redo some before or after the LAB edits, I can accomplish that without a lot of rework. And, if I want to tweak the LAB edits after I've gone back to the final RGB steps, I still have that work and all the adjustment layers in LAB mode so I can tweak it and repaste into the main image.

    So, I make a dup of the image, flatten it, convert to LAB and proceed to steps 7-9. This again calls for Apply Image, only this one is a little more complicated. Rather than apply a channel to the whole image, we're trying to apply a channel to itself in overlay mode. Margulis does not explain this in the first part of the chapter, but later on, he explains that this Apply Image step is really just the same as steepening the A and B curves by pulling in the ends a symmetrical amount (the easiest concept we learned in Chapter 1). This makes some sense because blending anything with itself in overlay mode is going to leave the neutrals the same, make the brights brighter and make the darks darker. Lo and behold, that's basically what you do when we push in the ends on an A or B curve. For some reason that I don't agree with, Margulis thinks that using Apply Image is easier than creating a curve adjustment layer and pushing in the ends. His only reasons for this are that he can more easily adjust the opacity of the result when he's done and that he likes the single slider in the Apply Image dialog rather than pushing in both ends in a curve dialog. I say that it's easier and more straightforward to build on what we already mastered in chapter one (and remember I like adjustment layers too), so I no longer use apply image here. Instead I create three adjustment layers and name them A, B and L. In the A and B layers, I push in the ends just like in chapter one. You have two variables for how much of an effect to use. How far to push in the ends (just like in chapter one) and the opacity of the resulting adjustment layer.

    Now, here's the real benefit of this method over Margulis'. Margulis' technique requires you to guess what percentages to use in Apply Image, sometimes using different ones for the A and B blend. If you follow his technique and your first guess isn't right, you can't change it. You have to undo and redo it. If you use the adjustment layers, you just tweak either the A channel, the B channel or both. No undo. No starting over.

    Anyway, back to my image. I ended up pushing the ends of both A and B about 10% and then adding a slight L contrast curve (basic chapter one work). Rather than overdo the effect and then dial back the opacity, I just don't overdo the effect in the first place. I dial in the effect until I like the result.

    I now have an action that creates these three named A, B and L layers. There's even a flavor of the action that automatically pushes in the ends of the A and B layers 10% each. I was already using this for non portrait pics too as general LAB color enhancement.

    Copy a merged version of this image to the clipboard and paste it in as a new layer into the original image. Finish the image there.

    Then we come to Margulis' techniques for sharpening which are worth learning. I particularly like how he dials in a good radius and threshold for local contrast enhancement with USM. That works well.

    So, what seemed really complex to me about a week ago has now been reduced to some fairly simply concepts all of which I already knew (except for his sharpening technique). Here's how I describe the portrait workflow now:
    1. Add contrast. If you need more facial contrast, find a good channel (often the green channel for faces) that you can copy and paste into a new layer and set the blend mode to luminosity. You can do this more than once. You can use several different channels if you want.
    2. Move to LAB. I don't convert my master image, I make a copy and convert that one - keeping both my original RGB and the LAB version around until I'm fully done with the image so it's easy to go back and tweak things at any time even things in LAB mode.
    3. Make A, B and L layers and tweak them just like you would in Chapter 1. If you ask me, Margulis portrait workflow just makes this step sound way too hard. It's why I didn't even attempt this workflow for awhile. But when you describe it as basically the same steps as Chapter 1, this is downright easy. Also you avoid the Apply Image command which works fine but is not understood by many.
    4. Sharpen.
    5. Paste result back to RGB image as a layer.
    I now feel like I really understand the portrait workflow and, fortunately for me, it's not a ton different than the canyon workflow which I found fairly easy to understand. I hope this is useful to someone else.
    --John
    HomepagePopular
    JFriend's javascript customizationsSecrets for getting fast answers on Dgrin
    Always include a link to your site when posting a question
  • Tom K.Tom K. Registered Users Posts: 817 Major grins
    edited December 2, 2005
    Thank you very much for this explanation and work flow method. Superb work on all counts.
    Visit My Web Site ~ http://www.tomkaszuba.com/
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited December 2, 2005
    jfriend wrote:
    I tried Margulis' recommendation of using blendif to hide the changes from the dark red parts of the image, but that blocked the effect from parts of the face too (where there is lots of red) and I had ot get the blendif slider almost to the middle in order to restore the red hair piece.

    Convert to LAB before adjusting the blending options. In this case, the magenta blendif slider can capture the hair piece without touching the faces. Your really have to do this anyway, as you are planning not to flatten until you move to LAB and your blending options will be lost after you convert.
    jfriend wrote:
    I think this is exactly the same thing as Apply Image in this case (please correct me if I'm wrong) except I can see what this is doing and the result is in a visible layer and I can change the opacity on it any time. And, more importantly, I can keep this change independent of other changes later.

    Yes and no. It is exactly the same thing as Apply Image followed by changing the blending mode to luminosity. Once you do this your future options are exactly the same as if you had used Apply Image. No differernce at all. You have not opened any additional options for yourself.
    jfriend wrote:
    So, I make a dup of the image, flatten it, convert to LAB and proceed to steps 7-9.

    There you go again, flattening in RGB instead of LAB. You really want the luminosity blend to take place in LAB so you get the benefit of impossible colors if possible. Also, I find the LAB blendif sliders more versitle as discussed above.
    jfriend wrote:
    Now, here's the real benefit of this method over Margulis'. Margulis' technique requires you to guess what percentages to use in Apply Image, sometimes using different ones for the A and B blend. If you follow his technique and your first guess isn't right, you can't change it. You have to undo and redo it. If you use the adjustment layers, you just tweak either the A channel, the B channel or both. No undo. No starting over.

    My PS action basically make the layers you describe but with overlay technique. See my initial post on this thead and search for "action". Also, I actually worked out curves that are identical to the overlay blends. See: http://dgrin.com/showpost.php?p=202259&postcount=71

    Edgework has some even more compelling reasons to do this. Read his posts to the portrait thread.
    jfriend wrote:
    Copy a merged version of this image to the clipboard and paste it in as a new layer into the original image. Finish the image there.
    No, sharpen in LAB or perhaps in CMYK if you know what you are doing and why. There are very good reasons for this. Perhaps you actually did this, bu it's not what you described.

    Anyway, sounds like you basically get it and I agree with most of what you wrote. Edgework and I have been making the point that the basic framework integrates very well with other techniques from the book. For example, Edgework suggests substituting the Man from Mars technique from Chapter 12 for the overlay layer approach. If there are blown spots on the face, you can use the Impossible Retouch from Chapter 8 just after the conversion to LAB (and after flattening if you do so.)

    OK to have the moderators move these two posts into the main discussion thread of this chapter? I think that way they won't be lost later on.
    If not now, when?
Sign In or Register to comment.