Using LAB for blending multiple RAW developments
jfriend
Registered Users Posts: 8,097 Major grins
As a side effect of all the recent LAB learnings, I've discovered that there are many great uses for the BlendIf layer settings. They can not only be used for many things in RGB, but sometimes LAB mode can really help you isolate the color you are looking to isolate so you can more easily make a modification without much or any masking. As such, I discovered that with some photos I could use the Blend If sliders in combination with LAB color to help me do a blending of two RAW developments to recover from a high dynamic range problem. This procedure is somewhat image-specific (since how you isolate colored objects depends upon what colors they are) but I thought I'd share a new way to use the Blend If sliders in LAB mode.
Here's what I started with (Salmon fishing in a glacer-fed river in Alaska):
In the RAW editor, when I tweak the image for good results on the face, I find that the reds get blown in both the red shirt and the yellow hat. When I tweak the image to save the highlight detail in those two areas, the rest of the image is quite dark and lacking contrast. Here's what that image looks like:
Notice that there's good detail in both the hat and shirt, but lots of shadow detail is much darker than we'd like. Now, it's probably possible to work directly on this underexposed image and raise the brightness in other areas, but I really liked the color and saturation in the brighter development so I decided I'd see if there was an easy way to blend the shirt and hat from the under-exposed image with the rest of the image. I started with a premise that because these two elements are so uniform in color that there ought to be an easy mechanism to blend them without extensive or careful masking.
The first thing I did was paste the two RAW developments into the same image as layers. Then, I experimented with the Blend If sliders in RGB mode. Since the shirt is entirely red, I started with the red sliders. Unfortunately, I couldn't make use of it in RGB mode. The problem is that the shirt contains a wide tonal range of red tones and the skin also contains a bunch of red tones. I couldn't find any settings that would let only the shirt blend, but not the face. The other channels weren't that useful.
So, I thought "lets go look at the image in LAB mode" and see how isolated the shirt is there. This time, I hit a jackpot. The Magenta side of the A channel is extremely strong in the shirt and not so strong in the face. A set of sliders like this would get me complete isolation of the red shirt without interfering with the face. With no masking at all, I got a perfect blend of the darker development of the shirt with the rest of the image.
That then left me with this image (note that the detail and a more reasonable color has returned in the shirt), but not yet in the hat.
OK, now I turn my attention to the hat. To prepare for this, I paste another copy of the dark image to be a third layer. I want to now find some blending settings that will let just the dark version of the hat come through. I'm optimistic that this should be easy because the hat is bright yellow and there's a yellow side to the B channel. So, I start with the yellow side of the B channel. Since I'm operating on the dark layer, I want to block everything from this layer except bright yellow. I grab the blue side of "this layer" and pull it to the right until the rest of the image comes back. What I find is that, unlike with the red shirt, there's no magic setting that gives me all of the hat from this layer, but lets all the rest of the image comes through from below. There are just too many other things in the image that have a significant yellow component, some very bright like the yellow splotches in the life jacket. So, the Blend If settings will never be perfect here. But, I do notice that nothing that isn't perfect is anywhere near the edges of the hat. That leads me to think I can set the Blend If settings to get the hat from this layer and then just do a quick and dirty mask to block the darker layer everywhere else away from the hat. That mask took all of 20 seconds because no careful edge is needed near the hat. So, now I have a blend of the hat from the darker image. The image now looks like this (detail restored in shirt and hat):
The rest of the retouch is basically just standard color enhancements, contrast curves and sharpening. I'm in LAB mode already, so I might as well add some contrast with a curve on the L channel and then experiment with color enhancements.
I start with a contrast improving curve on the L channel. When I brought my two RAW images in, I'd used the Linear curve in ACR because I wanted as much shadow detail as possible in the shaded areas of the face. But the image was a little flat. I added a quick S-curve on the L channel like this:
I didn't want to lose much shadow, so I didn't move that much. I did want to steepen the mid-tones. Doing this curve did make the hands just a bit too bright, so I just masked that area out from this curve with a few passes of a partial opacity soft brush. The curve isn't very strong so the mask didn't have to be very accurate, but the mask restored the detail in the hands. There are probably Blend If ways to protect the hands, but the mask was quick so that's what I used.
The contrast curve gave me this version:
Now I thought I'd play with color in the A and B curves. I'm not expecting to use it much because this image is already full of bright color, but I figure I'm here in LAB mode already so I might as well experiment. On the A channel, I quickly find that I have to leave the magenta side alone. Any move in magenta just makes the shirt go nuts. I lock the magenta side of the curve down and play with the green side. An extreme move in the green side clearly affects the foliage in the background and the back of the life jacket, but neither is really a needed enhancement, so no changes in the A channel.
Now I play with the B channel. I start with a simple symmetric move (from Chapter 1). With only a 10% push of the two ends, I find I like the subtle changes to the skin tone, but it's hammers the yellow in the hat and life jacket. Since I've already figured out how to isolate the hat once before, I figure I'll try using BlendIf settings for the B channel curve. This time I apply the settings to the the lower layers instead of this layer so I can block the curve's effect if the lower layer has strong yellow with this setting:
That gives me this slightly enhanced color, without effecting the hat.
Finally, I apply sharpening to the L channel and convert back to RGB and end up with this before/after result:
I hope people found this useful.
Here's what I started with (Salmon fishing in a glacer-fed river in Alaska):
In the RAW editor, when I tweak the image for good results on the face, I find that the reds get blown in both the red shirt and the yellow hat. When I tweak the image to save the highlight detail in those two areas, the rest of the image is quite dark and lacking contrast. Here's what that image looks like:
Notice that there's good detail in both the hat and shirt, but lots of shadow detail is much darker than we'd like. Now, it's probably possible to work directly on this underexposed image and raise the brightness in other areas, but I really liked the color and saturation in the brighter development so I decided I'd see if there was an easy way to blend the shirt and hat from the under-exposed image with the rest of the image. I started with a premise that because these two elements are so uniform in color that there ought to be an easy mechanism to blend them without extensive or careful masking.
The first thing I did was paste the two RAW developments into the same image as layers. Then, I experimented with the Blend If sliders in RGB mode. Since the shirt is entirely red, I started with the red sliders. Unfortunately, I couldn't make use of it in RGB mode. The problem is that the shirt contains a wide tonal range of red tones and the skin also contains a bunch of red tones. I couldn't find any settings that would let only the shirt blend, but not the face. The other channels weren't that useful.
So, I thought "lets go look at the image in LAB mode" and see how isolated the shirt is there. This time, I hit a jackpot. The Magenta side of the A channel is extremely strong in the shirt and not so strong in the face. A set of sliders like this would get me complete isolation of the red shirt without interfering with the face. With no masking at all, I got a perfect blend of the darker development of the shirt with the rest of the image.
That then left me with this image (note that the detail and a more reasonable color has returned in the shirt), but not yet in the hat.
OK, now I turn my attention to the hat. To prepare for this, I paste another copy of the dark image to be a third layer. I want to now find some blending settings that will let just the dark version of the hat come through. I'm optimistic that this should be easy because the hat is bright yellow and there's a yellow side to the B channel. So, I start with the yellow side of the B channel. Since I'm operating on the dark layer, I want to block everything from this layer except bright yellow. I grab the blue side of "this layer" and pull it to the right until the rest of the image comes back. What I find is that, unlike with the red shirt, there's no magic setting that gives me all of the hat from this layer, but lets all the rest of the image comes through from below. There are just too many other things in the image that have a significant yellow component, some very bright like the yellow splotches in the life jacket. So, the Blend If settings will never be perfect here. But, I do notice that nothing that isn't perfect is anywhere near the edges of the hat. That leads me to think I can set the Blend If settings to get the hat from this layer and then just do a quick and dirty mask to block the darker layer everywhere else away from the hat. That mask took all of 20 seconds because no careful edge is needed near the hat. So, now I have a blend of the hat from the darker image. The image now looks like this (detail restored in shirt and hat):
The rest of the retouch is basically just standard color enhancements, contrast curves and sharpening. I'm in LAB mode already, so I might as well add some contrast with a curve on the L channel and then experiment with color enhancements.
I start with a contrast improving curve on the L channel. When I brought my two RAW images in, I'd used the Linear curve in ACR because I wanted as much shadow detail as possible in the shaded areas of the face. But the image was a little flat. I added a quick S-curve on the L channel like this:
I didn't want to lose much shadow, so I didn't move that much. I did want to steepen the mid-tones. Doing this curve did make the hands just a bit too bright, so I just masked that area out from this curve with a few passes of a partial opacity soft brush. The curve isn't very strong so the mask didn't have to be very accurate, but the mask restored the detail in the hands. There are probably Blend If ways to protect the hands, but the mask was quick so that's what I used.
The contrast curve gave me this version:
Now I thought I'd play with color in the A and B curves. I'm not expecting to use it much because this image is already full of bright color, but I figure I'm here in LAB mode already so I might as well experiment. On the A channel, I quickly find that I have to leave the magenta side alone. Any move in magenta just makes the shirt go nuts. I lock the magenta side of the curve down and play with the green side. An extreme move in the green side clearly affects the foliage in the background and the back of the life jacket, but neither is really a needed enhancement, so no changes in the A channel.
Now I play with the B channel. I start with a simple symmetric move (from Chapter 1). With only a 10% push of the two ends, I find I like the subtle changes to the skin tone, but it's hammers the yellow in the hat and life jacket. Since I've already figured out how to isolate the hat once before, I figure I'll try using BlendIf settings for the B channel curve. This time I apply the settings to the the lower layers instead of this layer so I can block the curve's effect if the lower layer has strong yellow with this setting:
That gives me this slightly enhanced color, without effecting the hat.
Finally, I apply sharpening to the L channel and convert back to RGB and end up with this before/after result:
I hope people found this useful.
--John
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
0
Comments
I know I did.
Dgrin FAQ | Me | Workshops
Real Body Integrated Arts
GMT -5