Anyone wanna take a stab at this?

MelP

I posted this in the People forum and it was recommended that I post it here to see if any could get the red tint out of it.

I shot this in RAW. But am not that good at post editing (though, I am getting better). I have both PSP X and PS Elements 2.

Feel free to play around with the image and see if you can make it better. I'm sure this is easy for you all. All I ask is that you post your steps of what you did so that I may learn a thing or two.

Here is the image:


Original can be found here


Thanks,
Mel

rutt

Pretty easy to do this kind of thing with curves, either in RGB or LAB. I don't know if your programs have curves. See: http://dgrin.smugmug.com/gallery/1091766

With PS, you can use LAB and then it's even easier.

gluwater

This took me about 2 minutes to do. Like Rutt said I used Lab curves but RGB curves would work just as well.


And this is the curve I used:

Rhuarc

Ok, how did you know to use that curve? When I go into the curves dialog I don't know what I'm doing and I just start clicking things to see what happens! Not exactly the most efficient way to work!

rutt

That's the point of the link I posted: http://dgrin.smugmug.com/gallery/1091766

Or you could begin a plan of study which has changed the lives of many right ehre on dgrin! Step right up, to the Dan Margulis LAB book study group! Now this won't cost a thing except some thinking on your part. And I have a whole chorus of people to testify. JFriend, step right up and tell him. DavidTO. Ginger. Even Andy Williams is coming around. Why the list is as long as your arm. Tell the boy what LAB has done for you...

gluwater

Rhuarc wrote:

Ok, how did you know to use that curve? When I go into the curves dialog I don't know what I'm doing and I just start clicking things to see what happens! Not exactly the most efficient way to work!

I've read most of the Lab book mentioned by Rutt and have followed along with some of the Lab book study group. It just takes some time and practice. You can hold down the alt key and click on the area you are trying to fix and it will show the coresponding area on the curve that you need to change.

rutt

gluwater wrote:

I've read most of the Lab book mentioned by Rutt and have followed along with some of the Lab book study group. It just takes some time and practice. You can hold down the alt key and click on the area you are trying to fix and it will show the coresponding area on the curve that you need to change.

That's another soul for LAB!

gluwater

rutt wrote:

That's another soul for LAB!

Your just loving this aren't you:D .

rutt

gluwater wrote:

Your just loving this aren't you:D .

Well, yes. Didn't get enough sleep last night and decided play my role to the hilt.

I'll get myself under control.

MelP

Rhuarc wrote:

Ok, how did you know to use that curve? When I go into the curves dialog I don't know what I'm doing and I just start clicking things to see what happens! Not exactly the most efficient way to work!

I second this. I have tried curves but come up hopeless. Not sure where or how to move the lines.

Mel

MelP

gluwater wrote:

This took me about 2 minutes to do. Like Rutt said I used Lab curves but RGB curves would work just as well.


And this is the curve I used:

WOW!!! A big difference!!!

Do all programs have LAB? If so how do I go about seeing this info? or is it something just certain programs have (like the expensive Photoshop)?

Thanks,
Mel

rutt

Only Photoshop supports LAB well (to my knowledge). It doesn't have to be that expensive. You can use PS 4 or later. PS 6 or 7 should be pretty cheap by now and both are pretty darned good. You can always upgrade it and I'm sure that would be much cheaper than just buying the full CS2 in one step.

rutt

MelP wrote:

I second this. I have tried curves but come up hopeless. Not sure where or how to move the lines.

Mel

I'll work on trying to get you started on this. The very best way is to start the pop tutorials. See:

http://dgrin.smugmug.com/gallery/1075277

or:

http://dgrin.smugmug.com/gallery/1108502

depending on whether you belive me that LAB is easier or Andy and David who aren't so sure.

System

different way-
hue/saturation
drop red sat
drop blue sat
bring up master sat and play with master hue to get color where you want it-
george

edgework

MelP wrote:

I posted this in the People forum and it was recommended that I post it here to see if any could get the red tint out of it.

I shot this in RAW. But am not that good at post editing (though, I am getting better). I have both PSP X and PS Elements 2.

Feel free to play around with the image and see if you can make it better. I'm sure this is easy for you all. All I ask is that you post your steps of what you did so that I may learn a thing or two.

Here is the image:


Original can be found here


Thanks,
Mel

Look, I'm a confirmed LABaholic, but I don't think lab is the place for this problem. If you look at the numbers, you see that it's not a red issue, it's a yellow problem: There isn't enough. Aside from the rosy cheeks, she's reading 10-20 points more magenta than yellow across her whole face. In addition, the whole thing is flat. So what needs to happen is yellow has to be pumped into the three quarter tones and magenta has to be sucked out of the quartertones and highlights. In fact, all the quartertons and highlights have to come down, to increase the contrast. The L channel isn't subtle enough for this move. You could borrow masks from a different color space, say a magenta mask from CMYK or a Green mask from RGB, to mute the intensity of LAB, but you can get the same results directly in either of those spaces. I read numbers in CMYK but did my moves in RGB.

These curves go a long way towards fixing the problem:



and this selective color setting eases up some of the redness that's left over:



Here's the result:

gluwater

Very nice edgework. How did you know to add yellow rather than subtract magenta?

edgework

gluwater wrote:

Very nice edgework. How did you know to add yellow rather than subtract magenta?

Usually (and I say that guardedly), Magenta isn't the problem unless it's a really bad scan or the camera settings were schizoid. The problem will probably be too little cyan, in which case the skin will look like it was filtered through plutonium, or too little yellow, in which case there will be a cool red cast to the skin, and blues will be REALLY blue, since a weak yellow will leave it unavailable to counter the blues and give them shape and definition.

First clue in this image is the overall similarity in the backdrop, her shirt and her jeans. All are too blue, and the blues are too close to each other. But the numbers in her face are the clincher. We can assume that this is not overly tanned caucasian skin, in which case magenta and yellow should be about equal. Rules are meant to be broken all the time, and there have been some threads here questioning this "standard", which I think have raised valid points. But I'm reading 5C, 39M, 10Y under her eye and that's just not going to happen in this universe.

The magenta problem, almost always, is not quantity, but distribution. Often, even when the numbers suggest that the tones are in line, the face will still look too red. Usually that means that there is insuufficient contrast in the magenta channel. That's why I pulled the quartertones down so drastically in my curves. Bringing some shape into the mix can remove a lot of the redness, simply by making areas that we expect to be lighter conform to our expectations.

And by focusing on her face, which is all I really addressed with my curves, note how the blues have now pulled farther apart, and how the shirt and jeans pop out from the background. Which just proves that if a color is out of whack in one area, it's out all over.

dandill

I am trying to get the hang of using CMY values for skin adjustment. Here is a crop of my first go, showing the smaple points I used.



The before/after CMY values for the three sampled points are

#1 C31/21 M45/39 Y38/49
#2 C8/5 M45/39 Y18/24
#3 C1811 M38/32 Y25/33

I left the high megenta in #2 because of the rosy cheeks. The before(left)/after(right) full images are



The curves are



Thanks for any suggestions.

Grendel2000

edgework wrote:

Usually (and I say that guardedly), Magenta isn't the problem unless it's a really bad scan or the camera settings were schizoid. The problem will probably be too little cyan, in which case the skin will look like it was filtered through plutonium, or too little yellow, in which case there will be a cool red cast to the skin, and blues will be REALLY blue, since a weak yellow will leave it unavailable to counter the blues and give them shape and definition.

First clue in this image is the overall similarity in the backdrop, her shirt and her jeans. All are too blue, and the blues are too close to each other. But the numbers in her face are the clincher. We can assume that this is not overly tanned caucasian skin, in which case magenta and yellow should be about equal. Rules are meant to be broken all the time, and there have been some threads here questioning this "standard", which I think have raised valid points. But I'm reading 5C, 39M, 10Y under her eye and that's just not going to happen in this universe.

The magenta problem, almost always, is not quantity, but distribution. Often, even when the numbers suggest that the tones are in line, the face will still look too red. Usually that means that there is insuufficient contrast in the magenta channel. That's why I pulled the quartertones down so drastically in my curves. Bringing some shape into the mix can remove a lot of the redness, simply by making areas that we expect to be lighter conform to our expectations.

And by focusing on her face, which is all I really addressed with my curves, note how the blues have now pulled farther apart, and how the shirt and jeans pop out from the background. Which just proves that if a color is out of whack in one area, it's out all over.

Edgework,

I'm curious why you think LAB isn't the way to go here.

I'm no expert (I've been a PS hack for years but only recently began learning it in earnest - I'm currently on Chapter 4 of the LAB book) but this seemed to work out nicely for me in LAB.

Here's my result:




I used the following curves in LAB:



As a relative newby, I understand that just because I think this looks good doesn't mean it actually is. If there is a problem, I'd love to identify it to improve my own work, however.

So lemme have it!

Thanks.

BTW - why won't the images show up when I link them? I use the "img src="http://blah blah blah.jpg" with <> instead o f"" yet it doesn't link the pics??? Any help for non dgrinners?

DISREGARD - PBASE HAD CRASHED!

edgework

Grendel2000 wrote:

Edgework,

I'm curious why you think LAB isn't the way to go here.

I'm no expert (I've been a PS hack for years but only recently began learning it in earnest - I'm currently on Chapter 4 of the LAB book) but this seemed to work out nicely for me in LAB.

Here's my result:
<img src="http://www.pbase.com/grendel2000/image/55612672/large.jpg"&gt;


I used the following curves in LAB:
<img src="http://www.pbase.com/grendel2000/image/55612670.jpg"&gt;

As a relative newby, I understand that just because I think this looks good doesn't mean it actually is. If there is a problem, I'd love to identify it to improve my own work, however.

So lemme have it!

Thanks.

BTW - why won't the images show up when I link them? I use the "img src="http://blah blah blah.jpg" with <> instead o f"" yet it doesn't link the pics??? Any help for non dgrinners?

First of all, let's remind ourselves, yet again, that there are many paths to a desired destination, and Photoshop is tolerant of most of them.

Second, the only reason people don't consider me as much of a LAB advocate as Rutt is because I don't post as much. I love LAB and it's totally transformed my work flow.

That being said, one of the things I've had to (and continue to) learn is when to use it and when not. It's not the replacement space for everything that I used to do elsewhere. It's an awesome embellishment and expansion of what I still do.

In this case, I don't think LAB is the place to repair seriously out of whack colors, at least not without a neutral benchmark to measure the cast overall. It's guesswork otherwise, and part of that is because (I freely admit) I have no intuitive feel for subtle number differences in lab readings. I can tell you if the numbers are showing Red, Green, Yellow or Blue, but I couldn't give you the recipe for a pink little baby's cheek if you held a gun to my head. Having come of age in the CMYK world, I'm not even all that fluent in RGB numbers, but looking at the CMYK readout is particularly relevant for faces and skin since the three ink colors have such specific roles to play in skin tones.

In this image I wanted to direct more yellow into the darker tones, and pull more magenta out of the lighter tones as well as lighten all three channels in the quartertone and highlights, but to varying degrees with each channel. This is something that LAB doesn't do easily, at least not with a single curve. The L channel won't darken one color and avoid modifying another, unless you use blending sliders. But that just makes my point: It's a complicated move in LAB, and a simple move in RGB or CMYK.

I'm not totally satisfied with my result, but if you compare it to yours, you will see that I've gotten more contrast out of my curves, and I've also managed to retain some color variation due to my increased magenta in the upper values. There's a believable rosiness in my cheeks; they feel flat, color-wise, in yours. Note also the difference in the color of my jeans. I think that might be close to real than yours, which feel a little washed out.

As for your thinking, it seems dead on. You've pushed the relevant range of magenta to a less saturated area, and your yellow curve is pulling down below the midpoint, adding it overall. Fortunately there are no clouds, salt or clean sheets of paper in the image, because they would be pushing yellow. Such is the power of the LAB midpoint.

Sometimes I use LAB beause it makes things easier; most cast problems fall into this group, but I've encountered files that worked seriously better in RGB for cast removal. I love the ability for LAB's imaginary colors to fill i blown out highlights... except those times when I want my highlight to stay neutral. Any channel blend that involves luminosity is almost certainly going to produce a better result in LAB, but if you venture into more arcane blends using Softlight or Overlay, LAB generates too many colors that get in the way, whereas with RGB, it's easy to keep it real. When I use the healing brush, I will always work in LAB, due to that space's superior handling of gradients and color blends. But dodging and burning with a Hard Light layer is far safer in RGB. Usually, the ability to attack color and luminosity separately is a good thing. In this case, it is the ability to modify both color and luminosity in different directions at once with Green and Blue RGB curves that makes this a candidate for RGB. (although if I wish to push things further in the way of enhancing the colors, LAB here I come.)

Image links: enclose your URL withing these tags:
Works everytime for me.

jfriend

Chapter 14 mentions Edgework's point

edgework wrote:

In this image I wanted to direct more yellow into the darker tones, and pull more magenta out of the lighter tones as well as lighten all three channels in the quartertone and highlights, but to varying degrees with each channel. This is something that LAB doesn't do easily, at least not with a single curve. The L channel won't darken one color and avoid modifying another, unless you use blending sliders. But that just makes my point: It's a complicated move in LAB, and a simple move in RGB or CMYK.

To add to what Edgework is saying: As Margulis himself discusses in Chapter 14, one of the reasons to do things in RGB instead of LAB is when you need to do color correction differently at different tones. Since LAB so perfectly separates color from luminosity, it's very difficult to make a different color correction in a quartertone than a midtone than a threequartertone. As edgework says, it can be done with masks or blend-if sliders in LAB, but it's way easier in RGB where you can manipulate color differently vs. tone.

I myself probably wouldn't have been able to identify what Edgework did (that he wanted to change yellow and magenta differently at different tones), but given that conclusion RGB is definitely an easier place to do that.

dandill

jfriend wrote:

... but given that conclusion RGB is definitely an easier place to do that.

John and edgework, not CMYK (see my post above)?

jfriend

CMYK or RGB

dandill wrote:

John and edgework, not CMYK (see my post above)?

Edgework will know better than I, but I would presume you could make the Edgework-style changes in either RGB or CMYK. Both allow you to adjust color at different tones. I usually try to avoid the color mode switch to CMYK just for convenience if it's not hard to do the adjustment in RGB, but sometimes, it is just easier to do the adjustment in CMYK.

edgework

dandill wrote:

John and edgework, not CMYK (see my post above)?

Odd that I overlooked your CMYK moves. As an old prepress grunt, it's still my space of choice for fine tuning skin tones, and if I'd encountered this image in a print job, I'd have gone straight to CMYK and done all my moves there.

I suspect that my CMY curves would pretty much resemble the RGB curves that I posted, in that they would need something of an "S" shape. I think your moves are accurate; the numbers look good but your curves don't address the need to pull overall contrast down in the highlights and quartertones.

What I find about CMY is that, lacking the black channel information, you can make more subtle moves with them than with RGB. Even if I'm going end up in RGB I sometimes find myself stopping off in CMYK to fine tune a yellow highlight. And there is one move in CMYK that I still love: a luminosity curve adjustment layer with a contrast move in the Magenta curve. Flat skin tones can almost always be traced back to a magenta channel lacking contrast. Stretching the magenta highlights in a luminosity curve can bring out a wealth of shape without messing with an otherwise good balance between the colors.

In some ways CMYK is the poor cousin of RGB and LAB, but it has some tricks available that keep it current in my workflow.

dandill

Here is a further very slight tweak of the CMY curves (though not yet with an S move): first try (earlier post) on the left, current of the right.



The second seems to make the blouse more three dimensional and also bings out some sparkle in the hair. It is amazing to me the effects of such small adjustments. By the way, it is a charming picture.

rutt

I played with this image for a few minutes and I stared to form a radical idea. What if the curtain in the background is actually black? Anyway, what if it's a lot more neutral than we are giving it credit for. The implication then is that the image has a pretty hefty blue cast, which also explains where the yellow in the face got to.

I played around with this assumption, but if you buy it, you have to also buy that the lighting is mixed because the straight line LAB curves won't neutralize all the curtain. But they do bring the image pretty close pretty fast, though her hair ends up a little too yellow.

Just wondering how it got the way it is. We'll do a better job if we know.

MelP

rutt wrote:

I played with this image for a few minutes and I stared to form a radical idea. What if the curtain in the background is actually black? Anyway, what if it's a lot more neutral than we are giving it credit for. The implication then is that the image has a pretty hefty blue cast, which also explains where the yellow in the face got to.

I played around with this assumption, but if you buy it, you have to also buy that the lighting is mixed because the straight line LAB curves won't neutralize all the curtain. But they do bring the image pretty close pretty fast, though her hair ends up a little too yellow.

Just wondering how it got the way it is. We'll do a better job if we know.

Hi Rutt,
The color of the background is actually a dark navy blue. In Dandill's post, his 2nd picture is getting pretty close to the actual color of the background.

It is really exciting to see how everyone has taken the picture and made their changes to it. Keep 'em coming!!!

Thanks,
Mel

rutt

Yeah, navy blue. That make a lot of sense, not all the way to black. I was figuring that out anyway.

Anyway, because this image really does have a pretty serious blue cast, it's a natural for LAB in spite of the fact that it's fleshtone. Here's what I ended up with:



I used these curves to fix the cast and add back in a little contrast:



The major move is in the B curve to fix the blue cast (adding quite a bit of yellow to the flesh tones.) The A curve is just fine tuning the hair and the rosy cheeks.

Once I got done fixing the cast I cheated and pushed it through the Chapter 17 portrait recipe.

Total time? < 10 minutes once I decided the curtain should be a lot more neutral.

I really wanted to fix in another color space, just to show that I'm really not a religious zealot. Really, really, really. Maybe there is still hope for me, but not this time.

Moral: Really bad casts are easy to fix in LAB and the Chapter 17 portrait recipe works once you get the hang of it.

MarkR

Here's my attempt...
Here's one I did using simple CMYK curves and an USM...


Original image (compressed for space) here.

Mike Lane

LAB will do your dishes, it will remind you of your mother's birthday. It will change the laundry and walk your dog. Thus is the power of LAB. If you're feeling blue, LAB will buy you ice cream and it'll send your girlfriend flowers in your name with a haiku to go with it.

LAB is your best friend and will always be there for you. Can I get an amen?

rutt

Mike Lane wrote:

LAB will do your dishes, it will remind you of your mother's birthday. It will change the laundry and walk your dog. Thus is the power of LAB. If you're feeling blue, LAB will buy you ice cream and it'll send your girlfriend flowers in your name with a haiku to go with it.

LAB is your best friend and will always be there for you. Can I get an amen?

Throw away your crutches, stand up, and walk!