Options
Correcting Lynn's nudes
rutt
Registered Users Posts: 6,511 Major grins
While Lynn was at the height of her panic about printing her nudes, we had some PM correspondence which I think would be more generally interesting. Lynn kindly gave me permission to post. She has since decided to go with B&W for this project, so in a sense, this is mute for this project. But I think it still illustrates a lot of interesting points, so here goes.
Lynn started out posting this image:
Looks great on the monitor, right? It's such a great image and so humorously shocking that it's hard to look carefully at the colors. And of course, the eye is very forgiving with images on the monitor. But Lynn quickly ran into trouble when she tried to print it. She had different descriptions of what was wrong, but she wrote about "orange patches".
So let's take an objective look at Joyce's skin color. I moved the image in CMYK, mostly because I'm most comfortable correcting portraits in this color space, but also because it is closely tied to printing. (Also, Lynn is reading Dan Margulis, and his first few chapters focus on CMYK. Because black is separate, it is often easier to correct in CMYK than RGB and so that's where Dan starts.)
After moving to CMYK I set a few color sampler points:
Now we can easily see why Lynn is having trouble making good prints of this. Joyce is an impossible color! The rule of thumb for Caucasian skin is that it should have Y&M about equal and C about 1/5 of that. This is not a hard and fast rule. Sun tan/burn, lighter and darker skin, light, can all vary the balance, but points 1 and 2 are way off the mark. For those values of magenta and yellow, we'd want a cyan value somewhere between 20 and 30. Orange is right. The point at 3 shows another fault of this image. The shadows are very bright red. The values of Y and M there are extremely high for skin. We'd need enough cyan to balance them that the result would be pretty black.
So in short, not enough cyan in the flesh midtones, and too much mageneta and yellow in the shadows. Not so hard to fix this (though it did take me two tries because I didn't notice the very red shadows at first.) Here are the curves I used:
The cyan, magenta, and yellow curves do just what we set out to do - raise the amount of cyan in the flesh midtones and clip magenta and yellow in the shadows to keep these from being too red. The black curve is just somthing that often works well in portraits. It lightens the skin tones and increases the drama of dark hair, eyes, &etc. In this case, since we incrased the cyan in the midtones, we especially need to reduce the black there in order to keep the result from being too dark.
Here how our samplers look after this adjustment:
These are myc more plausible values. 1 and 2 are red all right, but they have enough cyan to be light pink. 3 is now a reasonable skin shadow, instead of a deep red.
And here is the entire image after this move:
Joyce looks better doesn't she? Compare with the original. Now we can see how badly we started off. But even with a well adjusted monitor, none of us caught it. We had to look at the numbers!
In the course of this correction, we have screwed up the color balance of the flag. This makes sense, since the flag wasn't actually in the orignal shot under the original light. Lynn added it in photoshop. That's why the wite in the flag could be neutral while Joyce was so wrong. Of course Lynn has the flag in a separate layer, so she can just correct the Joyce layer without damaging the flag.
As long as we have the image in CMYK, there is another trick we can play. We'd like to sharpen to get more snap in the eyes, hair, hat, gun, etc., but we don't want to add wrinkles and otherwise make Joyce look less glamourous. We are in luck here, because nearly all the detail we don't what to bring out (in the flesh) is in the magenta and yellow channels, while most of the detail we do want to bring out is in the cyan and black channels. So we can USM to our heart's content in the C+K channels without ruining the fine skin texture which is a large part of the point of this shot.
Here how I did this:
Because the sharpening was so accurately targeted, I could have gotten away with much more. In Lynn's full resolution image, we might use a radius more like 3 pixels perhaps a much higher amount. It's just a question of making the hat and eye look natural at the intended viewing size.
So after restoring the original balance to the flag and sharpening, we have this:
What do we think about it? Compare to the original. This version will also accurately on any printer that is on the right planet.
Lynn started out posting this image:
Looks great on the monitor, right? It's such a great image and so humorously shocking that it's hard to look carefully at the colors. And of course, the eye is very forgiving with images on the monitor. But Lynn quickly ran into trouble when she tried to print it. She had different descriptions of what was wrong, but she wrote about "orange patches".
So let's take an objective look at Joyce's skin color. I moved the image in CMYK, mostly because I'm most comfortable correcting portraits in this color space, but also because it is closely tied to printing. (Also, Lynn is reading Dan Margulis, and his first few chapters focus on CMYK. Because black is separate, it is often easier to correct in CMYK than RGB and so that's where Dan starts.)
After moving to CMYK I set a few color sampler points:
Now we can easily see why Lynn is having trouble making good prints of this. Joyce is an impossible color! The rule of thumb for Caucasian skin is that it should have Y&M about equal and C about 1/5 of that. This is not a hard and fast rule. Sun tan/burn, lighter and darker skin, light, can all vary the balance, but points 1 and 2 are way off the mark. For those values of magenta and yellow, we'd want a cyan value somewhere between 20 and 30. Orange is right. The point at 3 shows another fault of this image. The shadows are very bright red. The values of Y and M there are extremely high for skin. We'd need enough cyan to balance them that the result would be pretty black.
So in short, not enough cyan in the flesh midtones, and too much mageneta and yellow in the shadows. Not so hard to fix this (though it did take me two tries because I didn't notice the very red shadows at first.) Here are the curves I used:
The cyan, magenta, and yellow curves do just what we set out to do - raise the amount of cyan in the flesh midtones and clip magenta and yellow in the shadows to keep these from being too red. The black curve is just somthing that often works well in portraits. It lightens the skin tones and increases the drama of dark hair, eyes, &etc. In this case, since we incrased the cyan in the midtones, we especially need to reduce the black there in order to keep the result from being too dark.
Here how our samplers look after this adjustment:
These are myc more plausible values. 1 and 2 are red all right, but they have enough cyan to be light pink. 3 is now a reasonable skin shadow, instead of a deep red.
And here is the entire image after this move:
Joyce looks better doesn't she? Compare with the original. Now we can see how badly we started off. But even with a well adjusted monitor, none of us caught it. We had to look at the numbers!
In the course of this correction, we have screwed up the color balance of the flag. This makes sense, since the flag wasn't actually in the orignal shot under the original light. Lynn added it in photoshop. That's why the wite in the flag could be neutral while Joyce was so wrong. Of course Lynn has the flag in a separate layer, so she can just correct the Joyce layer without damaging the flag.
As long as we have the image in CMYK, there is another trick we can play. We'd like to sharpen to get more snap in the eyes, hair, hat, gun, etc., but we don't want to add wrinkles and otherwise make Joyce look less glamourous. We are in luck here, because nearly all the detail we don't what to bring out (in the flesh) is in the magenta and yellow channels, while most of the detail we do want to bring out is in the cyan and black channels. So we can USM to our heart's content in the C+K channels without ruining the fine skin texture which is a large part of the point of this shot.
Here how I did this:
Because the sharpening was so accurately targeted, I could have gotten away with much more. In Lynn's full resolution image, we might use a radius more like 3 pixels perhaps a much higher amount. It's just a question of making the hat and eye look natural at the intended viewing size.
So after restoring the original balance to the flag and sharpening, we have this:
What do we think about it? Compare to the original. This version will also accurately on any printer that is on the right planet.
If not now, when?
0
Comments
"The Edge... there is no honest way to explain it because the only people who really know where it is are the ones who have gone over."-Hunter S.Thompson