LAB Color - Chapter 4
Nikolai
Registered Users Posts: 19,035 Major grins
It's All About ...
The Food! (Andy)
The Centerfold? (me)
The Center Point (Dan, actually:-)
Chapter 1 showed that some easy symmetrical steepening of A and B curves (and well as some sharpening in L channel) can go a long way, making otherwise dull landscapes bright and removing fog and haze.
Chapter 2 brought some theoretical background to that magic.
Chapter 3 went a bit further and explored the possibilities of A and B curves having different angles.
In all these cases thus far we let one thing to be the same: all our curves were always going through the same original center point (0A0B). Now it's time to see what happens if we unlock this magic point of this mysterious space.
As you figured out, Chapter 4 dead centers on "center point".
First of all, executive version:
1) you need to move A or B curves off center if your image has a "cast".
2) you need to add a "center point lock" if you need to modify "halves" independently, for instance adjust red (i.e. magenta) but leave green intact (or apply independent green tweak).
Following the brave pioneers of this discussion (Rutt and DavidTO) I'm going to provide my own image and try to follow the brilliance of the author.
My first obstacle was to find a casted image that would carry enough familiar patterns to recognize the problem and would be joy to fix. I was trying to remember when was the last time I forgot to switch white balance from incandescent to auto:-), but nothing came to mind. Finally I decided to cheat and use of of the RAW files (in which I could spoil the white balance - hey, it's for the better good!).
Here's an example we'll be working with:
This picture was taken during my May trip to Yosemite, when I met with Sid and Michiel in SF. This is actually how 828's RAW data were recorded, so I didn't even have to spoil it on purpose. As it's easy to see, this picture features certain yellow cast.
I started with exploring the picture. As it was said before me, LAB is all about to make the picture closer to what you THINK you saw. Also, one of Dan's mantras says: "don't give them colors they wouldn't believe". So I put my markers on the 4 most visually important colored areas:
Here's how info looked on our markers:
It's easy to see that both clouds (#1) and water (#4) have positive B values (yellow), while they, generally speaking, should have had negative (blue) ones - it was a clear morning after a rainy night, air should not have added any yellow cast to the clouds.
Following Dan's advice, I decided not to concentrate on potentially neutral points. Landscape photographers know only too well that there is no such thing as 100% reliable neutral color in nature. Clouds could take whatever color sun decides to throw at them, skies can be anything from deep blue (up in the mountains) to pale green (way up North), water, even if not colored by some minerals, reflects both skies and clouds..
So I dropped the idea of "betting" my image on anything and simply took off with what I thought was right (supported by some channel info:-).
First I made sure that I need to offset only one curve, namely B, since B controls Blue-Yellow and the image is yellowish. Simple tweaking the center of A both ways proved me right: any move away from center point made the cast worse.
After playing a little with B I reached "believable" blueness of the water (#4) while keeping sky (#2) and clouds (#1) also quite natural.
My last stop was marker #3 - the Golden Gate Bridge itself. There was no scientific info on its color, but I *thought* it was less orange and more magenta. However, I already knew that if I make any change to the center point I'll introduce a new cast. Therefore I used the Trick No 2 and locked the A-curve in the middle, which allowed me to experiment freely with the red hue.
After I added a bit more contrast to L channel by making a pretty much standard S-curve, my curves ended up looking like that:
And my new marker info looked like follows (new data below, I kept the original info on top for an easy comparison):
As it's easy to see, B values #1 and #4 were changed in a most drastic manner - they "crossed the border" and moved from Yellow Country to Blue Country.
Here is the final image:
And, for another easy comparison, half-blended version:
The difference is quite obvious, and, I'd say, the final version is much closer to what I saw that lovely morning.
[SNIDE REMARK]
Actual editing in RAW without any LAB magic gives you this:
which, to my taste, is even closer to "my reality" than my LAB-processed version. So, my take on this: if you have RAW - by all means, use it first!
However, if you don't - LAB can give you a good hand in cast correction.
Besides, when it comes to independent color adjustment (trick #2) - LAB is a hands down winner, since raw editors cannot do this (yet).
[/SNIDE REMARK]
OK, what did we learn?
We learned to remove the cast (offset the curve away from center), and we also know how to fix one color without affecting its counterpart (lock the curve in the middle).
Images for this post are also available here.
On to Chapter 5!
HTH
The Food! (Andy)
The Centerfold? (me)
The Center Point (Dan, actually:-)
Chapter 1 showed that some easy symmetrical steepening of A and B curves (and well as some sharpening in L channel) can go a long way, making otherwise dull landscapes bright and removing fog and haze.
Chapter 2 brought some theoretical background to that magic.
Chapter 3 went a bit further and explored the possibilities of A and B curves having different angles.
In all these cases thus far we let one thing to be the same: all our curves were always going through the same original center point (0A0B). Now it's time to see what happens if we unlock this magic point of this mysterious space.
As you figured out, Chapter 4 dead centers on "center point".
First of all, executive version:
1) you need to move A or B curves off center if your image has a "cast".
2) you need to add a "center point lock" if you need to modify "halves" independently, for instance adjust red (i.e. magenta) but leave green intact (or apply independent green tweak).
Following the brave pioneers of this discussion (Rutt and DavidTO) I'm going to provide my own image and try to follow the brilliance of the author.
My first obstacle was to find a casted image that would carry enough familiar patterns to recognize the problem and would be joy to fix. I was trying to remember when was the last time I forgot to switch white balance from incandescent to auto:-), but nothing came to mind. Finally I decided to cheat and use of of the RAW files (in which I could spoil the white balance - hey, it's for the better good!).
Here's an example we'll be working with:
This picture was taken during my May trip to Yosemite, when I met with Sid and Michiel in SF. This is actually how 828's RAW data were recorded, so I didn't even have to spoil it on purpose. As it's easy to see, this picture features certain yellow cast.
I started with exploring the picture. As it was said before me, LAB is all about to make the picture closer to what you THINK you saw. Also, one of Dan's mantras says: "don't give them colors they wouldn't believe". So I put my markers on the 4 most visually important colored areas:
Here's how info looked on our markers:
It's easy to see that both clouds (#1) and water (#4) have positive B values (yellow), while they, generally speaking, should have had negative (blue) ones - it was a clear morning after a rainy night, air should not have added any yellow cast to the clouds.
Following Dan's advice, I decided not to concentrate on potentially neutral points. Landscape photographers know only too well that there is no such thing as 100% reliable neutral color in nature. Clouds could take whatever color sun decides to throw at them, skies can be anything from deep blue (up in the mountains) to pale green (way up North), water, even if not colored by some minerals, reflects both skies and clouds..
So I dropped the idea of "betting" my image on anything and simply took off with what I thought was right (supported by some channel info:-).
First I made sure that I need to offset only one curve, namely B, since B controls Blue-Yellow and the image is yellowish. Simple tweaking the center of A both ways proved me right: any move away from center point made the cast worse.
After playing a little with B I reached "believable" blueness of the water (#4) while keeping sky (#2) and clouds (#1) also quite natural.
My last stop was marker #3 - the Golden Gate Bridge itself. There was no scientific info on its color, but I *thought* it was less orange and more magenta. However, I already knew that if I make any change to the center point I'll introduce a new cast. Therefore I used the Trick No 2 and locked the A-curve in the middle, which allowed me to experiment freely with the red hue.
After I added a bit more contrast to L channel by making a pretty much standard S-curve, my curves ended up looking like that:
And my new marker info looked like follows (new data below, I kept the original info on top for an easy comparison):
As it's easy to see, B values #1 and #4 were changed in a most drastic manner - they "crossed the border" and moved from Yellow Country to Blue Country.
Here is the final image:
And, for another easy comparison, half-blended version:
The difference is quite obvious, and, I'd say, the final version is much closer to what I saw that lovely morning.
[SNIDE REMARK]
Actual editing in RAW without any LAB magic gives you this:
which, to my taste, is even closer to "my reality" than my LAB-processed version. So, my take on this: if you have RAW - by all means, use it first!
However, if you don't - LAB can give you a good hand in cast correction.
Besides, when it comes to independent color adjustment (trick #2) - LAB is a hands down winner, since raw editors cannot do this (yet).
[/SNIDE REMARK]
OK, what did we learn?
We learned to remove the cast (offset the curve away from center), and we also know how to fix one color without affecting its counterpart (lock the curve in the middle).
Images for this post are also available here.
On to Chapter 5!
HTH
"May the f/stop be with you!"
0
Comments
I love the split comparison view.
The greatest thing to my mind of all this isn't so much the removal of overall cast, as I would also rely on RAW for that (being a raw shooter), but removing the cast from specific parts of a shot.
I had a shot of my wife wearing a white t-shirt outside. The shirt picked up a blue cast from the sky--easily removed.
Here's a link to another example of the power of LAB.
Dgrin FAQ | Me | Workshops
I'd be very interested to know the steps you took in your raw conversion - you did a great job there of getting rid of that colour cast. Was it mainly a WB correction?
Cheers!
David
www.uniqueday.com
You are right about using getting color balance right in raw before starting to work on an image. It's easier if you start on the right planet. I also like to use the exposure and shadow sliders to reduce highlight and shadow clipping to the greatest extent possible (giving me more control in PS proper.) That's what I do in the raw converter.
But after conversion, a trip to LAB nearly always pays off. In the case of your bridge, I got started with your post-raw image (the last one in your post) and got this:
There was nothing really wrong with the starting point, but it was a little flat, no deep shadows and no full use of the highlights in the sky. My L curve was targeted to address this:
I bought some steepness in both the shadows and highlights by moving the endpoints in just a tad, as much as I could without blowing highlights or clogging shadows (good word, "clogging", eh? I've been looking for it and Dan supplied in an email.) Then I made steepend the highlight to bring out the detail in the clouds and again in the 3/4 tones to bring out the details in the bridge and far shore.
With better the image looked less flat, but now the colors seemed less vivid than the might be.
First, note that both curves keep their center points neutral. Nik's conversion has plausible color balance. Sky, water, clouds, vegetation, and bridge all have plausible readings. So I'm not trying to correct a cast. Instead, I'm just trying to get a little more "pop". So the B curve is just a stock symetrical steepening. I started off with an A curve like this, too. But while the B curve gave nice color to the sky and water, neither wanted more green. On the other hand, the vegation on the opposite shore did cry out for a little more green to balance the increased yellow from the steepend B curve. Thus the funny business ot the right side of center in the A curve. I'm pulling up the greens in the vegation and then flattening throught the greens in the water and sky. The point on the left of the A curve adjusts the magenta in the bredge to keep it under control. We don't want corvette red. This bridge has a famous rust color and that means the yellow and magenta have to be closely balanced.
But I want to go back to you LAB correction of the image with the cast. You said the raw convesion was " even closer to "my reality" than my LAB-processed version". Why? The answer is that it still has a cast.
I've added a few color sampler points to that final version:
The points read:
- A=7, B=(37)
- A=6, B=(11)
Can this be right? There is a lot of blue in that shadowed area on the opposite shore where I set point 1. I'd think this point should be neutral to slightly positive in the B channel since it's vegetation. And what's with the positve A value? That's magenta. Now look at the clouds. I'd expect them to be pretty close to neutral, perhaps slightly blue-cyan. But again, there's this postive A value. Magenta in the clouds is OK at dawn and dusk or in a storm, but on a clear San Francisco day? Seems unlikely. And that B value is more than a little blue.By comparison, the raw conversion had:
- A=0,B=(3)
- A=1,B=(18)
Much more like what we'd expect. So your LAB curves did lose the yellow cast, but left us with a magenta cast. I'd also likt to see at least a little green in the shadowed vegetation on the far shore. This might not really be possible. It's pretty dark in there.Here is my shot at cast removal from the original yellow shot. I have done just enough to make the lightest point of the clouds neutral and get a plausible values in the vegetation.
Now this isn't supposed to be a perfect finished product, just meant to show how easy it is to get the cast out in LAB. I set a couple of CS points and then only moved the right endpoints until the cast was gone. No magenta, no yellow. We could hope for more pop. And though the water of SF Bay might really be a little green, I'd personally like to see it more blue in this shot. And as I noted in the beginning, there is hay to be made in the L curve.
I got this with very simple curves:
I didn't manage blue water. Doing that would get more blue in the clouds than I wanted. But that would be an option...
Worked well, didn't it?:-)
I agree. But, as in the book, I decided to work on a fully casted image to make the (center:-) "point" clear. And this followup would allow us to go into further, AP-level details:-)
Cheers!
In RAW it did the following:
WB: from "as shot" to "daylight" (that took care of the 90% of cast)
Tweaked the other settings from Auto to whatever was more pleasing:-), esp shadows.
Curves - moved to Strong contrast and touched the curve to enhance it a bit more.
Details: Set all three (sharpness, luminance smoothing and color noise reduction) to 100%.
That was it!
Thanks for looking!
Thank *you* for letting me work on this chapter, it was fun:-)
I agree. Each tool has its own merit and domain. The main point I got from all the PS books I've read (Scott's, Kevin's and Dan's) is that if you learn all of them you'll be able to use the most appropriate one to the task at hand in each particular case, therefore saving time and efforts and getting the exact results you want.
Of course, the important (and one of the hardest) part is to *know* what you want in advance :-) I agree again. After all, the LAB is all about what you want the image to be (or what you think you saw), and cold camera's AI is just too darn "objective", while what we want is it to be "personal", "subjective". In this example, camera (or PS) has no idea that we want the bridge to "pop", for its just a set of reddish pixels..
Yep, you're absolutely correct. I can see it clearly now
Yes, I didn't pay much attention to the upper A half, my bad, sorry mastah
Hey, nobody's perfect. Important thing is that we showed how one can affect various color areas using assymetric and off-centered LAB curves.
Locking the center point (at least maybe after the "center cast") generally helps with color dependency. And if not - there are always masks, selections and other tools:-)
Thank you for a great feedback!
I bought Dan's book this afternoon so watch out!!
Cheers!
David
www.uniqueday.com
As it's been pointed out already, WB is one thing, but actually adjusting some partial colors to what we THINK they were (as opposed to what camera's cold eye saw) is where the real strength of the LAB lies...
The more the merrier:-)
Cheers!
Watch out yourself. We are always looking for someone to write a chapter summary.
Here's my attempt for chap 4. This isn't so much an example of an overall color cast as it is certain casts in certain colors. This might actually belong more in a future chapter. Feel free to move it if that turns out to be the case.
Here's the original:
Not too bad as a starting point.
I first did the basics from chap 1:
The colors are much more vivid. I should have spent more time on the L but the focus here this time was the color.
Things I don't like here:
-Skin yellow
-Leaves more yellow than I like
-Water bottle should actually be more purple where here it is more blue
Side note: I've never done much in curves so I didn't know the trick of holding the mouse button and scanning over the photo to see where each spot falls on the curve. Very simple yet very useful tip!
I used that tip to find out where my problem spots were and steepened or shallowed the a/b curves in the spots where I needed more/less of the particular color.
The colors here seem much more pleasing to me. I'm sure it could be even better with more knowledge and time. Obviously this took me a little while but I'm at the very bottom of the learning curve (pun intended) and I can see where it can be much quicker.
Feel free to play with this image.
http://philu.smugmug.com
I like it how you presented the color areas in the curves, very visual and easy to read!
Darn, I wish PS had some built-in tool for publishing the curves..
Hey, Phil. I would bet that you could use the B curve to eliminate the bluish cast on the shadow area of the face....
Dgrin FAQ | Me | Workshops
You're probably right - and now that you've pointed it out I notice it more
Besides just being a shadow area he was in a blue blow-up pool which I'm sure added to it.
I'll have to play with it more.
http://philu.smugmug.com
Try this one. Got rid of the blue cast around his eye and the green cast on his shoulder. I made his skin less grey which I think (hope) is better. Pushed out some more purple in the water bottle - it really looks right now. Also did some tweaking on the L to bring out different areas. Something still doesn't feel right with the skin tone though.
Man - talk about the 80/20 rule. 80% of the time on 20% of the details - such is life.
http://philu.smugmug.com
Better. Hard to tell without working on it if you could go even farther with getting rid of the casts, but it definitely helps. Subtle, but good.
Dgrin FAQ | Me | Workshops
Phil - I think you are very close. I would remind you that Dan talks about how skin needs to be on the yellow side of red, and even though your little boy is very fair, I think you may have taken out just a tad too much yellow. Otherwise it looks very good. Great shot of a cute kid, by the way.
-- Russ