I'm supposed to be working on Chapter 11, and now I'm running around trying to figure out Image Applying and Blending blue and green channels, until I'm blue in the face!!
I'll see if I can work up some of the portraits from your gallery, and I'll try to avoid the overcooked lobster impressions.
EDIT: stepping back and comparing the two edits, I realize we ended up with fairly different results. I cannot say that the different result is necessarily because of the difference in technique because I ended up with how I wanted the image to be. It's warmer, the face is brighter and the sky is not as bright. I purposely did not look at your result while I was working on mine because I wanted to see what my normal retouch thinking would produce.
I think shadow/highlight is a good tool to use but once you go beyond the 15% range it really starts to wreck images IMO - jfriend-John - your edit is not pleasing to me - but I'm still not yet an advocate yet of LAB in practice - though I find the study and discussion fascinating!
First 2 photos completed. I'll upload them with comments in the AM.
I'd like to try a little different approach. Hold onto them until we have some others to compare them with. Then we'll compare them all at the same time and nobody will have seen the other attempts before showing his/her own.
I think shadow/highlight is a good tool to use but once you go beyond the 15% range it really starts to wreck images IMO - jfriend-John - your edit is not pleasing to me - but I'm still not yet an advocate yet of LAB in practice - though I find the study and discussion fascinating!
Here's a different approach. I'm not trying to disprove the LAB approach - in fact I'm learning a lot from it - but I am trying to figure out when you really need all those steps and complication versus the more traditional approaches and/or how close the traditional steps will get to the LAB approach.
In this attempt, I used an inverted blue channel (biggest delta between sky and lady) with a few modifications to make a mask for the lady and then just used a curve to brighten her up without affecting the background. Then added USM local contrast enhancement and smart sharpen. No shadow/highlights.
Is this one any better? How's it compare to Rutt's LAB retouch?
My retouch is on the left, the Rutt LAB retouch on the right: Larger Version Here
Here's a different approach. I'm not trying to disprove the LAB approach - in fact I'm learning a lot from it - but I am trying to figure out when you really need all those steps and complication versus the more traditional approaches and/or how close the traditional steps will get to the LAB approach.
In this attempt, I use
There are two issues here:
How much time is it worth spending to learn new techniques?
How much time is it worth spending to improve any single image?
I think they are both interesting questions, and I'd like to address them separately.
[size=+1]How much time is it worth spending to learn new techniques?[/size]
This entirely depends on what you like to do. If you are Dan Margulis, the answer is: a lifetime. For some photographers, the answer is: 0. For example, two of my all time favorite photographers, Henri Cartier-Bresson and Annie Leibovitz, never went into the darkroom, leaving it to specialists. Two other of my favorites, Irving Penn and Ansel Adams, spent limtless time in the darkroom, taking some of their most famous images through numerous editions incorporating new ideas, technology, and innovations. Man Ray completely broke down the barriers between darkroom and photography per se.
Dan's particular approach particularly appeals to me because it builds upwards for a foundation of theory. Dan's methods work because they should work. The numbers work out and the image works out. It's so great when that happens. At the risk of a little pretension, I'm going to quote Isaac Newton:
When the numbers come out right like that, you know as Pythagoras did that a secret of nature is open in the palm of your hand.
OK, OK, learning how to color correct digital portraits isn't exactly the same thing as descovering the law of universal gravitation, but the experience Newton was describing isn't that different from the moment I figured out why LAB doesn't need a fourth channel for the cyan and red color opponents and why those were the two right colors to leave out.
That's pretty abstract. What about the potrait technique? Before JFriend started asking about simpler methods, I was already thinking that my explanations and step-by-step examples weren't doing justice to Dan's technique. I'm so interested in the "why" behind what's going on that I'm often guilty of giving too much information.
I watched Dan "correct" 10 portraits in as many minutes giving a running description of the reasoning behind each step while doing so. At least a few of those were new to him; they were my images which I had just given him. So there is a proof that, once mastered, the technique can be applied very quickly. In fact, that's part of the point. It's a recipe. Follow the steps. Do it the same way every time. Learn which variations are called for depending on the properties the input images and the goals for the image.
If you use my Photoshop actions, the Dan's LAB portrait technique is really only a few steps:
Run the first action, ending up in LAB with a luminosity blend layer.
Decide if that layer is helping. Tweak its blending modes.
Run the second action, ending up with a couple of overlay layers. Tweek their relative opacity based on some simple rules about people's complexions.
Merge the overlay layers and tune the amount of saturations and color differentiation to taste. Optionally apply an L curve to open up the contrast of the image.
Conventional USM sharpening, with some extra care to keep away from skin.
High radius sharpening.
I don't think that's any more steps than you ended up using, actually. But it is different from what you already know how to do. So it seems more complicated and less simple until you invest the time to learn it. After you do that, well, it's a workflow, a structure for dealing with portraits. That greatly simplifies working with portraits because you can do basically the same thing every time, yet it has more than enough room for variations. Since it arises from a deep understanding of the structure of images, it also has logical points for specific variations. For example, the highlight half of highlight/shadow should be applied very early in RGB where the highlights occupy the greater part of the contrast range, whereas the shadow half should be done in LAB right after the conversion with the luminosity blend layer selected. Blown highlights can be fixed with an "impossible retouch" after the overlay blend when color differentiaion is greater and the borders of impossible color blend won't be magnified by the overlay blend.
In sum, the technique is simple to learn, takes time to master, is quick to apply once mastered, and provides a framework within which almost anything (or almost nothing) can be done to improve a portrait.
[size=+1]How much time is it worth spending to improve any single image?[/size]
This is one of Dan's favorite questions. He addresses it at great length in Chapter 7, but he put it best in his class which I recently attended. I'll have to paraphrase:
Suppose you have just shot 1000 pictures of an event, perhaps a wedding, and you are making a proof sheet for your customer. How much time should you spend on the individual images? Very little. You don't even know which ones the customer will choose. Get the color balance right. Lighten or darken so they are visible. Apply by some batch process. Less than 1 second of your time per image assuming you took less than 15 minutes to set up the batch job. On the other hand, what if for the first time, you have a photo chosen for the front cover a major magazine? How long is it worth spending on that image? A day? A week? How many times will you try again before you convince yourself than you cannot do better with your current level of sophistication.
I recently saw a great quotation on dgrin, I think it was Aporia's signature and that it's been changed in the last day. Something about most of your iamges sucking and even the great phogotrahpers are remembered for only a few images. How long is it worth working on those images? Adams and Penn seem to have decided that the answer was that it was worth a lifetime.
Anyone else have a version of the first portrait from the practice set, or planing to make one in the next few days. We have three so far, Pathfinder's, Aporia's, and I've done one. That's enough to have a bake off, but more is better. Speak up.
"To be honest, most of my pictures suck . . ."
While often mind-boggling, Margulis has been very helpful for me in understanding not only the practice of color retouching but the critical theory that makes it work. As a beginner in the field, I find this particularly clarifying. Otherwise, I find myself just clicking buttons and hoping that trial-and-error leads me somewhere I wanted to go.
These are helpful reflections, Rutt. Thanks for all your time and energy to generate this LAB working group.
Aporia's signature . . . Something about most of your iamges sucking and even the great phogotrahpers are remembered for only a few images. How long is it worth working on those images? Adams and Penn seem to have decided that the answer was that it was worth a lifetime.
No magic bullet:
"To be honest, most of my pictures suck. The saving grace of that admission is that most of your pictures suck, too. How could I possibly know such a thing? Because most of everybody's pictures suck, that's how. I've seen Cartier-Bresson's contact sheets, and most of his pictures sucked.
. . . And what does this have to do with the 'magic bullet'? Just that there ain't one." (Mike Johnston. (The Magic Bullet. Sept 4, 2005. http://www.photo.net/mjohnston/column80/).
The first time I read this recipe it seemed a little over whelming, as I really didn't use Image>Apply and had never thought of overlay blending in an A or B channel. But after running through it a few times, it flows quickly, and I think I underrstand what is going on with the image in the various steps. I can think of some further variations also.
One thing about the images that I ran through this recipe and posted the URL for above, is that they would never be seen in this manner because we were asked NOT to spot and heal the images and some of them certainly need it.
I appreciate the time Rutt has spent setting up this thread about Margulis Portrait technique, and I want to thank him for it.
It has been very helpful in my understanding of the process. It really can't be called a recipe because there is quite a bit of decision making along the way - so I'll call it a process. Thanks again, John.
The process does seem to me to increase color noise if there is noise present in the original image at all. It is probably NOT visible in a print, but it is when viewed at 100% on a monitor screen. Anyone else notice this?
One thing about the images that I ran through this recipe and posted the URL for above, is that they would never be seen in this manner because we were asked NOT to spot and heal the images and some of them certainly need it.
At least for the first few image, I think it's very helpful to see how much you can do without any spot retouching. You can do an amazing amount with curves and channel based masks. Before I print for the photographers and retouchers, we'll take care of this kind of thing. For the second set, probably, we'll lift this restriction and see what happens.
It has been very helpful in my understanding of the process. It really can't be called a recipe because there is quite a bit of decision making along the way - so I'll call it a process.
"Workflow"? "Framework"? It's sort of like a recipe in that you learn the basics and then invent variations. But sometimes it doesn't work without the variations.
The process does seem to me to increase color noise if there is noise present in the original image at all. It is probably NOT visible in a print, but it is when viewed at 100% on a monitor screen. Anyone else notice this.
That's a good question. I do find myself struggling with noise at the very end. I've put it down to the two sharpening passes. Notice that Dan applies the conventional USM sharpening with a very light touch: on a low opacity layer, constrained by an inverted L mask selection, and with (for him) a high threshold.
If you take one where you find that you did end up with such noise, it would be very useful to snapshot the image at each stage and track the introduction of the noise to narrow it down.
I suspect the noise is introduced by the overlay blending of the A and B channel, but I have no proof. I'll look into it.
There is no real reason to use overlay blends. The exact same transformation can be described with a pair of curves. Dan thinks the overlay blends are easier. I made the curves that do the same thing once. I'll dig them up and post tomorrow and you can compare with the overlays.
I suspect the noise is introduced by the overlay blending of the A and B channel, but I have no proof. I'll look into it.
Here is an example from my last step-by-step. This is a pretty close crop. Notice that some noise appears on the tooth during the process. Looks like I let that happen during USM, though the immediately preceeding L curve move set things up by adding enough contrast for USM to pick up.
I was not cooking strictly by Dan's recipe here. He uses an inverted L channel as a mask during conventional USM to keep it away from flesh and he used blending options with the Blend if sliders on the A channel. I'm not sure how much difference this would have made. A higher threshold value might have been good, looking at that tooth and at her eyes.
Here's a portrait of Andy's which he sent me when I asked him for portraits to work on. This was the result of a simple raw conversion:
Andy sent me an edit, I think his goal was to fix the blown area on his nose. He wrote, "A little more than none at all. I spent maybe 3 or 5 minutes on it."
I did it twice. The first time my goal was to see how fast I could run it through the potrait recipe. This took 10 minutes.
The second time I spent 15 minutes instead of 1 minute on the nose to reconstruct it by cloning on a luminosity blend layer and painting on a color blend layer (this sort of thing works much better, more intuitively in LAB mode.) The second time I also made a final trip into CMKY to steepen the K curve, clean up the under colors in the blacks (selective-color). My real goal was to hold shadow detail to try to get some more of that left cheek back (I used L- channel shadow/highlight in the fast attempt.) But I was also pleased by the rich clean blacks.
This is a great thread and I am learning and trying to absorb this stuff like a sponge. I have tried these techniques on several of my photos, and I am sure that its probably just me but it seems that when I do this, my pics seem too saturated/ too contrasty. Here is a before and after of one that I took a shot at. Any ideas that you think could improve this?
it seems that when I do this, my pics seem too saturated/ too contrasty. Here is a before and after of one that I took a shot at. Any ideas that you think could improve this?
Somehow you lost contrast, detail, and softness on that right cheek. It doesn't stand out as well from the background as in the original. I'd say that the biggest obvious way in which your edit isn't better than the original. Look at the channels of the original RGB. How does that cheek look in the green channel? Was that the problem? The other place you might have caused this problem was with the L curve following the overlay blends. What does that curve look like?
That was straight from RAW conversion, resized, and converted to sRGB. Now here is the after.
That is what the blending sliders are for - so you can dial back the amount of the blending for the final image. Some of my edits, I only used the blend at 15 -20%
Here's a portrait of Andy's which he sent me when I asked him for portraits to work on. This was the result of a simple raw conversion:
Andy sent me an edit, I think his goal was to fix the blown area on his nose. He wrote, "A little more than none at all. I spent maybe 3 or 5 minutes on it."
I did it twice. The first time my goal was to see how fast I could run it through the potrait recipe. This took 10 minutes.
The second time I spent 15 minutes instead of 1 minute on the nose to reconstruct it by cloning on a luminosity blend layer and painting on a color blend layer (this sort of thing works much better, more intuitively in LAB mode.) The second time I also made a final trip into CMKY to steepen the K curve, clean up the under colors in the blacks (selective-color). My real goal was to hold shadow detail to try to get some more of that left cheek back (I used L- channel shadow/highlight in the fast attempt.) But I was also pleased by the rich clean blacks.
The greens of the background are much more vibrant in the last image and the face stands out and seperates from the background much better and with more detail visible.
The greens of the background are much more vibrant in the last image and the face stands out and seperates from the background much better and with more detail visible.
Thanks. I was pleased with that last version. Since Andy's street work was so nice to begin with, I thought it was an hour (well, maybe an hour and ten minutes) well spent.
Somehow you lost contrast, detail, and softness on that right cheek. It doesn't stand out as well from the background as in the original. I'd say that the biggest obvious way in which your edit isn't better than the original. Look at the channels of the original RGB. How does that cheek look in the green channel? Was that the problem? The other place you might have caused this problem was with the L curve following the overlay blends. What does that curve look like?
Thanks. I was pleased with that last version. Since Andy's street work was so nice to begin with, I thought it was an hour (well, maybe an hour and ten minutes) well spent.
Your green channel shows the problem. See how poor the contrast of the cheek is against the background. What does the blue channel look like? Maybe you want to use it instead, or make some blend with the blue and green channels.
That is what the blending sliders are for - so you can dial back the amount of the blending for the final image. Some of my edits, I only used the blend at 15 -20%
That's probably a big part of my problem. I have been working on the original layer the entire time I had been doing this technique, which I never do with other post-processing methods. Thanks for the advice.
Your green channel shows the problem. See how poor the contrast of the cheek is against the background. What does the blue channel look like? Maybe you want to use it instead, or make some blend with the blue and green channels.
I see it now, thanks. Would you make this blend via a black and white conversion using channel mixer or do you have another method that you would use to blend the blue and green channels?
I see it now, thanks. Would you make this blend via a black and white conversion using channel mixer or do you have another method that you would use to blend the blue and green channels?
If you are good at B&W conversions, then these make the best luminosity masks. Dan blends channels with "Image->Apply Image" and adjusts the blending modes and opacities with frightening virtuosity. You haven't shown us the blue channel, but is it too dark to use as is?
If you are good at B&W conversions, then these make the best luminosity masks. Dan blends channels with "Image->Apply Image" and adjusts the blending modes and opacities with frightening virtuosity. You haven't shown us the blue channel, but is it too dark to use as is?
I believe that it is too dark, but now I'm not so sure I completely understand the purpose of this step in the process. I know what it says it is for, but I'm not completely sure as to what I am looking for in the channels. I know that the red is usually blown out in the skin tones which doesn't provide any depth and usually the blue is too "rough" showing too much in the way of skin texture/blemishes. That's about all I look at. What else should I be looking for?
I believe that it is too dark, but now I'm not so sure I completely understand the purpose of this step in the process. I know what it says it is for, but I'm not completely sure as to what I am looking for in the channels. I know that the red is usually blown out in the skin tones which doesn't provide any depth and usually the blue is too "rough" showing too much in the way of skin texture/blemishes. That's about all I look at. What else should I be looking for?
There is a lot of theory behind the idea of using the green channel for a luminosity blend. Basically, the red channel is usually really bad for faces, no contrast, completely blown. Yet, it has a bigger part in computing the L channel than the blue channel. To the extent the red channel is used at all for this, the end result will lose detail and depth. So we try to find a better alternative. The best altrnative is a good B&W conversion, but not everyone knows how to do those and for faces, often the green channel is pretty close.
Take a look at each of the RGB channels.
Which is the better B&W of this particular shot? It sort of depends what you are looking for. The green is nice, but as we noticed before, it loses the right cheek against the background. The blue has a lot of nice depth, but yes, it's too dark. What to do?
I blended the green and blue channels as follows: start with the green and then Image->Apply Image to blend in the blue in lighten mode at 100% opacity. Here is the result:
This gave the face a little more definition and yet greater softness than the green channel, but we'll still have the problem of the right cheek cheek and the background. I decided I'd try to handle this later on.
I went ahead. I had to change the blending options after the conversion to LAB in order to exclude the shirt from the blend. Then I just followed the recipe and made the overlay layers and merged them and changed the opacity of the merged opacity layer to 40%. Then I applied this L curve to it:
This curve is very very steep across the transition from background to right cheek, and then flattens and lowers through the shadow on the left cheek. I wanted to make the right cheek stand out from the background and lighten that shadow and soften the skin on the left cheek. The result:
And after the two sharpening passes I arrived here:
So? I like this, but of course I did it, so I got more or less what I was after. Maybe the contrast between left and right side is too dramatic for you, but it retains a touch of the drama from the blue channel without reassigning his racial/ethnic identity. High radius sharpening further made his face stand out from the background. On the other hand it did darken that pesky shadow on the left cheek. You could use do high radius sharpening on a layer and use a layer mask to reduce the opacity across this part of the image if you like. You might go for a more yellow look, by reducing the opacity of the A overlay layer earlier in the recipe. You could have gotten a less tanned look by using a lower opacity for the merged overlay layer. This is really a matter of taste.
[size=+1]Date for practice set #1 comparisons?[/size]
Edgework, Pathfinder, Aphoria, and myself have done at least the first 3 shots of the practice set. Jstick80 is threatening to do some of them as well.
Let's make a date and all be online at the same time sometime in the next few days and see if we can compare results and pick best versions. I think this will go best if we do it this was so the response times are quick.
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.
Comments
I'll see if I can work up some of the portraits from your gallery, and I'll try to avoid the overcooked lobster impressions.
Moderator of the Technique Forum and Finishing School on Dgrin
I think shadow/highlight is a good tool to use but once you go beyond the 15% range it really starts to wreck images IMO - jfriend-John - your edit is not pleasing to me - but I'm still not yet an advocate yet of LAB in practice - though I find the study and discussion fascinating!
Portfolio • Workshops • Facebook • Twitter
Yup!!
Moderator of the Technique Forum and Finishing School on Dgrin
I'd like to try a little different approach. Hold onto them until we have some others to compare them with. Then we'll compare them all at the same time and nobody will have seen the other attempts before showing his/her own.
Here's a different approach. I'm not trying to disprove the LAB approach - in fact I'm learning a lot from it - but I am trying to figure out when you really need all those steps and complication versus the more traditional approaches and/or how close the traditional steps will get to the LAB approach.
In this attempt, I used an inverted blue channel (biggest delta between sky and lady) with a few modifications to make a mask for the lady and then just used a curve to brighten her up without affecting the background. Then added USM local contrast enhancement and smart sharpen. No shadow/highlights.
Is this one any better? How's it compare to Rutt's LAB retouch?
My retouch is on the left, the Rutt LAB retouch on the right:
Larger Version Here
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Moderator of the Technique Forum and Finishing School on Dgrin
There are two issues here:
- How much time is it worth spending to learn new techniques?
- How much time is it worth spending to improve any single image?
I think they are both interesting questions, and I'd like to address them separately.[size=+1]How much time is it worth spending to learn new techniques?[/size]
This entirely depends on what you like to do. If you are Dan Margulis, the answer is: a lifetime. For some photographers, the answer is: 0. For example, two of my all time favorite photographers, Henri Cartier-Bresson and Annie Leibovitz, never went into the darkroom, leaving it to specialists. Two other of my favorites, Irving Penn and Ansel Adams, spent limtless time in the darkroom, taking some of their most famous images through numerous editions incorporating new ideas, technology, and innovations. Man Ray completely broke down the barriers between darkroom and photography per se.
Dan's particular approach particularly appeals to me because it builds upwards for a foundation of theory. Dan's methods work because they should work. The numbers work out and the image works out. It's so great when that happens. At the risk of a little pretension, I'm going to quote Isaac Newton:
OK, OK, learning how to color correct digital portraits isn't exactly the same thing as descovering the law of universal gravitation, but the experience Newton was describing isn't that different from the moment I figured out why LAB doesn't need a fourth channel for the cyan and red color opponents and why those were the two right colors to leave out.
That's pretty abstract. What about the potrait technique? Before JFriend started asking about simpler methods, I was already thinking that my explanations and step-by-step examples weren't doing justice to Dan's technique. I'm so interested in the "why" behind what's going on that I'm often guilty of giving too much information.
I watched Dan "correct" 10 portraits in as many minutes giving a running description of the reasoning behind each step while doing so. At least a few of those were new to him; they were my images which I had just given him. So there is a proof that, once mastered, the technique can be applied very quickly. In fact, that's part of the point. It's a recipe. Follow the steps. Do it the same way every time. Learn which variations are called for depending on the properties the input images and the goals for the image.
If you use my Photoshop actions, the Dan's LAB portrait technique is really only a few steps:
- Run the first action, ending up in LAB with a luminosity blend layer.
- Decide if that layer is helping. Tweak its blending modes.
- Run the second action, ending up with a couple of overlay layers. Tweek their relative opacity based on some simple rules about people's complexions.
- Merge the overlay layers and tune the amount of saturations and color differentiation to taste. Optionally apply an L curve to open up the contrast of the image.
- Conventional USM sharpening, with some extra care to keep away from skin.
- High radius sharpening.
I don't think that's any more steps than you ended up using, actually. But it is different from what you already know how to do. So it seems more complicated and less simple until you invest the time to learn it. After you do that, well, it's a workflow, a structure for dealing with portraits. That greatly simplifies working with portraits because you can do basically the same thing every time, yet it has more than enough room for variations. Since it arises from a deep understanding of the structure of images, it also has logical points for specific variations. For example, the highlight half of highlight/shadow should be applied very early in RGB where the highlights occupy the greater part of the contrast range, whereas the shadow half should be done in LAB right after the conversion with the luminosity blend layer selected. Blown highlights can be fixed with an "impossible retouch" after the overlay blend when color differentiaion is greater and the borders of impossible color blend won't be magnified by the overlay blend.In sum, the technique is simple to learn, takes time to master, is quick to apply once mastered, and provides a framework within which almost anything (or almost nothing) can be done to improve a portrait.
[size=+1]How much time is it worth spending to improve any single image?[/size]
This is one of Dan's favorite questions. He addresses it at great length in Chapter 7, but he put it best in his class which I recently attended. I'll have to paraphrase:
I recently saw a great quotation on dgrin, I think it was Aporia's signature and that it's been changed in the last day. Something about most of your iamges sucking and even the great phogotrahpers are remembered for only a few images. How long is it worth working on those images? Adams and Penn seem to have decided that the answer was that it was worth a lifetime.
While often mind-boggling, Margulis has been very helpful for me in understanding not only the practice of color retouching but the critical theory that makes it work. As a beginner in the field, I find this particularly clarifying. Otherwise, I find myself just clicking buttons and hoping that trial-and-error leads me somewhere I wanted to go.
These are helpful reflections, Rutt. Thanks for all your time and energy to generate this LAB working group.
No magic bullet:
"To be honest, most of my pictures suck. The saving grace of that admission is that most of your pictures suck, too. How could I possibly know such a thing? Because most of everybody's pictures suck, that's how. I've seen Cartier-Bresson's contact sheets, and most of his pictures sucked.
. . . And what does this have to do with the 'magic bullet'? Just that there ain't one." (Mike Johnston. (The Magic Bullet. Sept 4, 2005. http://www.photo.net/mjohnston/column80/).
The whole article is worth reading.
Real Body Integrated Arts
GMT -5
One thing about the images that I ran through this recipe and posted the URL for above, is that they would never be seen in this manner because we were asked NOT to spot and heal the images and some of them certainly need it.
I appreciate the time Rutt has spent setting up this thread about Margulis Portrait technique, and I want to thank him for it.
It has been very helpful in my understanding of the process. It really can't be called a recipe because there is quite a bit of decision making along the way - so I'll call it a process. Thanks again, John.
The process does seem to me to increase color noise if there is noise present in the original image at all. It is probably NOT visible in a print, but it is when viewed at 100% on a monitor screen. Anyone else notice this?
Moderator of the Technique Forum and Finishing School on Dgrin
You're welcome. My pleasure.
"Workflow"? "Framework"? It's sort of like a recipe in that you learn the basics and then invent variations. But sometimes it doesn't work without the variations.
That's a good question. I do find myself struggling with noise at the very end. I've put it down to the two sharpening passes. Notice that Dan applies the conventional USM sharpening with a very light touch: on a low opacity layer, constrained by an inverted L mask selection, and with (for him) a high threshold.
If you take one where you find that you did end up with such noise, it would be very useful to snapshot the image at each stage and track the introduction of the noise to narrow it down.
Moderator of the Technique Forum and Finishing School on Dgrin
There is no real reason to use overlay blends. The exact same transformation can be described with a pair of curves. Dan thinks the overlay blends are easier. I made the curves that do the same thing once. I'll dig them up and post tomorrow and you can compare with the overlays.
Here is an example from my last step-by-step. This is a pretty close crop. Notice that some noise appears on the tooth during the process. Looks like I let that happen during USM, though the immediately preceeding L curve move set things up by adding enough contrast for USM to pick up.
I was not cooking strictly by Dan's recipe here. He uses an inverted L channel as a mask during conventional USM to keep it away from flesh and he used blending options with the Blend if sliders on the A channel. I'm not sure how much difference this would have made. A higher threshold value might have been good, looking at that tooth and at her eyes.
PF, is this a representative example?
Andy sent me an edit, I think his goal was to fix the blown area on his nose. He wrote, "A little more than none at all. I spent maybe 3 or 5 minutes on it."
I did it twice. The first time my goal was to see how fast I could run it through the potrait recipe. This took 10 minutes.
The second time I spent 15 minutes instead of 1 minute on the nose to reconstruct it by cloning on a luminosity blend layer and painting on a color blend layer (this sort of thing works much better, more intuitively in LAB mode.) The second time I also made a final trip into CMKY to steepen the K curve, clean up the under colors in the blacks (selective-color). My real goal was to hold shadow detail to try to get some more of that left cheek back (I used L- channel shadow/highlight in the fast attempt.) But I was also pleased by the rich clean blacks.
Somehow you lost contrast, detail, and softness on that right cheek. It doesn't stand out as well from the background as in the original. I'd say that the biggest obvious way in which your edit isn't better than the original. Look at the channels of the original RGB. How does that cheek look in the green channel? Was that the problem? The other place you might have caused this problem was with the L curve following the overlay blends. What does that curve look like?
That is what the blending sliders are for - so you can dial back the amount of the blending for the final image. Some of my edits, I only used the blend at 15 -20%
Moderator of the Technique Forum and Finishing School on Dgrin
The greens of the background are much more vibrant in the last image and the face stands out and seperates from the background much better and with more detail visible.
Moderator of the Technique Forum and Finishing School on Dgrin
Thanks. I was pleased with that last version. Since Andy's street work was so nice to begin with, I thought it was an hour (well, maybe an hour and ten minutes) well spent.
If you are good at B&W conversions, then these make the best luminosity masks. Dan blends channels with "Image->Apply Image" and adjusts the blending modes and opacities with frightening virtuosity. You haven't shown us the blue channel, but is it too dark to use as is?
Here's the blue channel.
There is a lot of theory behind the idea of using the green channel for a luminosity blend. Basically, the red channel is usually really bad for faces, no contrast, completely blown. Yet, it has a bigger part in computing the L channel than the blue channel. To the extent the red channel is used at all for this, the end result will lose detail and depth. So we try to find a better alternative. The best altrnative is a good B&W conversion, but not everyone knows how to do those and for faces, often the green channel is pretty close.
Take a look at each of the RGB channels.
Which is the better B&W of this particular shot? It sort of depends what you are looking for. The green is nice, but as we noticed before, it loses the right cheek against the background. The blue has a lot of nice depth, but yes, it's too dark. What to do?
I blended the green and blue channels as follows: start with the green and then Image->Apply Image to blend in the blue in lighten mode at 100% opacity. Here is the result:
This gave the face a little more definition and yet greater softness than the green channel, but we'll still have the problem of the right cheek cheek and the background. I decided I'd try to handle this later on.
I went ahead. I had to change the blending options after the conversion to LAB in order to exclude the shirt from the blend. Then I just followed the recipe and made the overlay layers and merged them and changed the opacity of the merged opacity layer to 40%. Then I applied this L curve to it:
This curve is very very steep across the transition from background to right cheek, and then flattens and lowers through the shadow on the left cheek. I wanted to make the right cheek stand out from the background and lighten that shadow and soften the skin on the left cheek. The result:
And after the two sharpening passes I arrived here:
So? I like this, but of course I did it, so I got more or less what I was after. Maybe the contrast between left and right side is too dramatic for you, but it retains a touch of the drama from the blue channel without reassigning his racial/ethnic identity. High radius sharpening further made his face stand out from the background. On the other hand it did darken that pesky shadow on the left cheek. You could use do high radius sharpening on a layer and use a layer mask to reduce the opacity across this part of the image if you like. You might go for a more yellow look, by reducing the opacity of the A overlay layer earlier in the recipe. You could have gotten a less tanned look by using a lower opacity for the merged overlay layer. This is really a matter of taste.
Opinions?
Edgework, Pathfinder, Aphoria, and myself have done at least the first 3 shots of the practice set. Jstick80 is threatening to do some of them as well.
Let's make a date and all be online at the same time sometime in the next few days and see if we can compare results and pick best versions. I think this will go best if we do it this was so the response times are quick.
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.