Greg,
A couple of questions:
1) What is the relationship between white balance and color casts? If the white balance is set correctly for an image, will all (or most) neutrals already be correctly set? Or if that is too much of a generalization, are some portion of color casts caused by incorrect white balance?
2) I understand the RGB curves, and I understand the Lightness curve in Lab, but I don't understand Lab's a and b curves. Can you point me to something that explains them in detail?
-Jay
Questions
Jay,
Good questions...
White balance is the camera or correction adjusting for artificial light. For instance, if I am shooting an image and I have the white balance of the camera set for indoor tungsten light but I am outside, the image will look very blue. The White balance of the camera is thinking that the light hitting the imager is tungsten about 3200K where as the daylight actually hitting it is 5500K or so...
The image would have a heavy blue cast; so part of the answer is yes color balance and color cast can be related. When we are "generally" talking about color cast we are talking about a color leaking into the image that does not belong. In the image posted here http://www.curvemeister.com/forum/index.php?topic=2565.0 there is a strong green cast over the face of the child. It also bleeds down into other parts of the image but the face and skin is where we really hate to see it.
The trouble with this condition is that setting a neutral is going to get you close only if the neutral is experiencing the same color cast. If the neutral is not in the same light as the subject you could have a proper neutral but still have the cast on the subject. This is what happens to the child in the image above. If you fix the child the background goes "BOOM" if you fix the background the child is green...this is where masks and selective corrections come in.
Lastly if the entire image has a cast then setting a neutral should fix the major color issues in the image. the real trouble is getting a good neutral as you yourself have experienced. We make certain assumptions to get the neutral. It relies heavily on perception and sample points. The more you rely on the numbers sometimes the harder it gets. In the Abe video you saw that the neutral can be "shopped" around to find sweet spot that makes sense. If there is not clear neutral, that is the track I follow most often; and it is why I try to teach that shopping method to all of you. Your eyes can really tell you a lot if you trust them.
LAB: Lightness is the channel that makes the most sense. When I first hit LAB I was astounded at what I thought was the complexity of the whole thing...then I hit an image that Mike used to use to teach LAB color cast correction. It required that you be able to split the B channel in two with your correction. It was an eye opener. Here's how I now look at A and B...
In RGB you have three colors for correction, obvious and seemingly simple...but you cannot adjust RGB without messing with the color, all of it...you can select the R channel but you effect all of the R numbers when you make a move in there up to a certain point. The further the difference in brightness the less the effect, if you have two colors fairly close in brightness you can really mess with the color by adjusting only 1 channel.
In LAB you can isolate the color you want without messing all the others up. The A channel has two colors Magenta and Green...but at the very center of the grid is neutral L(x)(a0)(b0) this is the best part of LAB it allows you to separate the Magenta correction from the Green one. When you set a neutral in LAB you tell the image that the point I have chosen is now the center of the the A and B curves.
This is why you have only 1 neutral in LAB. If you place control points on the Magenta side of the grid...three equally spaced points should do, then crank on the other end of the curve you should be able to totally wack the greens out of this world without touching the magenta or reds. Week 4 has 2 images where you get to play with this and It will help if you understand what is happening first.
Now, what if I want to increase the reds and kill off a blue cast in the shadows? I would pin down the green end of the A curve and adjust the reds to my hearts delight...then pin down the yellow side of the B channel and kill off only the blue in the shadows.
As you look at LAB remember that the closer to vertical the curve is, the more saturated the colors are. Most of the action is near the center of the grid. There are techniques we are going to cover that increase the color variation of green vegetation because we isolate it in the A channel and then steepen the curve only in that area.
Lastly, LAB is capable of creating colors that cannot be produced by RGB. A extremely bright yellow for instance in LAB would be 80L (-5)A and 80B that would be = 255,255,255 in RGB because it cannot handle the colors, they are out of Gamut and RGB would try to come as close as possible but be unable to produce it. When you convert from LAB back to RGB it would assign pure white to the very bright yellow from LAB.
A extremely good book on LAB is "Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Color Space" By Dan Margulis. I have read it now twice and I am re-reading it again because every time I do I find something new to add to the class and my own personal knowledge. Dan can be very direct and sometimes confusing but if you relax, open your mind a bit and only try to get 1/4 of what is there, you will learn a ton.
Keep asking...I'll keep trying to give you an answer that makes sense...
PLEASE MAKE TIME for the conference call on near the end of the class. I think you will get a real benefit from it. Most of your questions are easily demonstrated and when you see that you will have totally different questions, but a much greater understanding. I hope... ;D
Greg
Good questions...
White balance is the camera or correction adjusting for artificial light. For instance, if I am shooting an image and I have the white balance of the camera set for indoor tungsten light but I am outside, the image will look very blue. The White balance of the camera is thinking that the light hitting the imager is tungsten about 3200K where as the daylight actually hitting it is 5500K or so...
The image would have a heavy blue cast; so part of the answer is yes color balance and color cast can be related. When we are "generally" talking about color cast we are talking about a color leaking into the image that does not belong. In the image posted here http://www.curvemeister.com/forum/index.php?topic=2565.0 there is a strong green cast over the face of the child. It also bleeds down into other parts of the image but the face and skin is where we really hate to see it.
The trouble with this condition is that setting a neutral is going to get you close only if the neutral is experiencing the same color cast. If the neutral is not in the same light as the subject you could have a proper neutral but still have the cast on the subject. This is what happens to the child in the image above. If you fix the child the background goes "BOOM" if you fix the background the child is green...this is where masks and selective corrections come in.
Lastly if the entire image has a cast then setting a neutral should fix the major color issues in the image. the real trouble is getting a good neutral as you yourself have experienced. We make certain assumptions to get the neutral. It relies heavily on perception and sample points. The more you rely on the numbers sometimes the harder it gets. In the Abe video you saw that the neutral can be "shopped" around to find sweet spot that makes sense. If there is not clear neutral, that is the track I follow most often; and it is why I try to teach that shopping method to all of you. Your eyes can really tell you a lot if you trust them.
LAB: Lightness is the channel that makes the most sense. When I first hit LAB I was astounded at what I thought was the complexity of the whole thing...then I hit an image that Mike used to use to teach LAB color cast correction. It required that you be able to split the B channel in two with your correction. It was an eye opener. Here's how I now look at A and B...
In RGB you have three colors for correction, obvious and seemingly simple...but you cannot adjust RGB without messing with the color, all of it...you can select the R channel but you effect all of the R numbers when you make a move in there up to a certain point. The further the difference in brightness the less the effect, if you have two colors fairly close in brightness you can really mess with the color by adjusting only 1 channel.
In LAB you can isolate the color you want without messing all the others up. The A channel has two colors Magenta and Green...but at the very center of the grid is neutral L(x)(a0)(b0) this is the best part of LAB it allows you to separate the Magenta correction from the Green one. When you set a neutral in LAB you tell the image that the point I have chosen is now the center of the the A and B curves.
This is why you have only 1 neutral in LAB. If you place control points on the Magenta side of the grid...three equally spaced points should do, then crank on the other end of the curve you should be able to totally wack the greens out of this world without touching the magenta or reds. Week 4 has 2 images where you get to play with this and It will help if you understand what is happening first.
Now, what if I want to increase the reds and kill off a blue cast in the shadows? I would pin down the green end of the A curve and adjust the reds to my hearts delight...then pin down the yellow side of the B channel and kill off only the blue in the shadows.
As you look at LAB remember that the closer to vertical the curve is, the more saturated the colors are. Most of the action is near the center of the grid. There are techniques we are going to cover that increase the color variation of green vegetation because we isolate it in the A channel and then steepen the curve only in that area.
Lastly, LAB is capable of creating colors that cannot be produced by RGB. A extremely bright yellow for instance in LAB would be 80L (-5)A and 80B that would be = 255,255,255 in RGB because it cannot handle the colors, they are out of Gamut and RGB would try to come as close as possible but be unable to produce it. When you convert from LAB back to RGB it would assign pure white to the very bright yellow from LAB.
A extremely good book on LAB is "Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Color Space" By Dan Margulis. I have read it now twice and I am re-reading it again because every time I do I find something new to add to the class and my own personal knowledge. Dan can be very direct and sometimes confusing but if you relax, open your mind a bit and only try to get 1/4 of what is there, you will learn a ton.
Keep asking...I'll keep trying to give you an answer that makes sense...
PLEASE MAKE TIME for the conference call on near the end of the class. I think you will get a real benefit from it. Most of your questions are easily demonstrated and when you see that you will have totally different questions, but a much greater understanding. I hope... ;D
Greg
-
derekfountain
- Posts: 251
- Joined: Fri Jan 26, 2007 1:24 pm
1) What is the relationship between white balance and color casts? If the white balance is set correctly for an image, will all (or most) neutrals already be correctly set? Or if that is too much of a generalization, are some portion of color casts caused by incorrect white balance?
2) I understand the RGB curves, and I understand the Lightness curve in Lab, but I don't understand Lab's a and b curves. Can you point me to something that explains them in detail?
Greg has given you his usual thoroughly detailed answer. Allow me to give the short versions:
1) Yes, incorrectly set white balance is one of the reasons that an image might contain a colour cast. But it's far from the only one, so no, correct white balance doesn't guarantee a cast-free image. (Think about coloured light hitting the subject, reflections from green sunshades (as per my child image Greg references), and so on.)
2) There are countless articles that describe LAB in detail. See Wikipedia for one, or Google for hundreds more. I have a feeling that what you need is a much simpler idea of how the A and B channels work.
Take the A channel. This describes how much green and how much magenta an image has. (In LAB those colours are "opposites" - a pixel can't be both green AND magenta, it must be one or the other.) The mid-point of the A channel curve, value 0, represents no green and no magenta. Value 0 means grey. Absorb that for a moment - the *midpoint* of the curve has value 0, and that means zero colour.
As the value of the A increases, the colour gets more magenta. A pixel with an A value of 1 is just about grey, tinged with a tiny hint of magenta. A value of 10 is a stronger shade, right up to the top magenta value of 128 - the very top right of the curve.
Going the other way, the value of A decreases. Since it starts at 0 that means negative numbers. A pixel with an A value of 0 has no green and no magenta; a value of -1 is very slightly green and a value of -128 is very green.
The same applies to the B curve, only that carries the other two colours in LAB: yellow and blue. The midpoint, 0, is grey, positive values mean yellow and negative values mean blue.
So a pixel with an A value of 0 and a B value of 0 (0A/0B) will be grey (whether it's light grey or dark grey depends on the L value for that pixel). A pixel with a value 10A/0B is a bit magenta and a pixel with 0A/-25B is a bit blue. A pixel with 100A/100B is strong in both magenta and yellow and will appear on screen as a shade of reddish orange. A pixel with -50A/-50B is quiite strong in both green and blue and will appear on screen as turquoise.
You can experiment with the PS colour picker. Double click on the foreground icon to bring up the colour picker, select the L radio button to put it in LAB mode, and set L to 90, A to 0 and B to 0. Now bump the A and B values up and down away from that 0 midpoint and see how the represented colour changes.
Derek et.al
Thanks for jumping in..This is one area where the more voices in the mix the better the answer. LAB is not very intuitive at first then it suddenly dominates a large amount of your time. I think it is the speed and power of the whole space that makes it so addicting.
If you only have five minutes...Vast improvements can be made in LAB...
Greg
Thanks for jumping in..This is one area where the more voices in the mix the better the answer. LAB is not very intuitive at first then it suddenly dominates a large amount of your time. I think it is the speed and power of the whole space that makes it so addicting.
If you only have five minutes...Vast improvements can be made in LAB...
Greg
-
mikemeister_admin
- Posts: 4927
- Joined: Fri Sep 20, 2013 8:29 pm
It is not a 1/4 thing...it is half of the A or B channel really and of that space, only the areas near the center vertical axis have most of the color information. Float you mouse around with the window enlarged and see how far from the axis you find color...
That 1/4 defines more blue or green the the entire RGB or WgCMYK color spaces. It is really that green is anti-magenta or blue is anti-yellow. Maybe a better way to describe it really is that the green color space is negative magenta hence the cooler colors are negative numbers...
If it is a size thing...make the window bigger..I myself frequently enlarge the window to over half of the screen so I can see what the heck I'm doing. Many people have asked for a "zoom" feature for the A and B curves so you can zoom in and see what you are adjusting. Harder to do than you would think...
Greg
That 1/4 defines more blue or green the the entire RGB or WgCMYK color spaces. It is really that green is anti-magenta or blue is anti-yellow. Maybe a better way to describe it really is that the green color space is negative magenta hence the cooler colors are negative numbers...
If it is a size thing...make the window bigger..I myself frequently enlarge the window to over half of the screen so I can see what the heck I'm doing. Many people have asked for a "zoom" feature for the A and B curves so you can zoom in and see what you are adjusting. Harder to do than you would think...
Greg
-
derekfountain
- Posts: 251
- Joined: Fri Jan 26, 2007 1:24 pm
One question,why are only a quarter of the graphs devoted to green and blue in lab?
There comes a point when you just have to accept the mathematical basis of the tool you're using. RGB uses 3 colours, CMYK uses 3 different ones plus black. LAB uses 4 colours, but puts 2 per graph. So you end up with half a graph per colour. It's the nature of the beast - there's plenty of mathematical analysis of why it works as it does, and people like Mike and Dan Margulis will be able to give you way more information on the theory than you would probably like. :)
The key point to remember with LAB colour curves is that the zero-colour point isn't in the bottom corner like most "sensible" curves, it's in the middle. You effectively have one curve going from the midpoint to the top right, and another one going from the midpoint to the bottom left. CM gives you an option (right click on the A or B curve) to pin down one half of the curve so it doesn't move while you play with the other side (allowing you to, say, fix the greens, while leaving the magentas just as they are). Just looking at the screen shot you (gremurphoto) posted, you appear to have used this to adjust the greens in the child image while leaving the magentas be. That's a good technique for that image.
-
mikemeister_admin
- Posts: 4927
- Joined: Fri Sep 20, 2013 8:29 pm
-
mikemeister_admin
- Posts: 4927
- Joined: Fri Sep 20, 2013 8:29 pm
Short answer...yes..
Longer answer...Yes but.
The effective range of input and output values is limited by the final destination.
If your output values that cannot be printed by today's pigments there is no real value in doing so.
It is part of the reason the action occurs closest to the axis in the first place. The further out from the center the less likely they are to be printable.
Greg
Longer answer...Yes but.
The effective range of input and output values is limited by the final destination.
If your output values that cannot be printed by today's pigments there is no real value in doing so.
It is part of the reason the action occurs closest to the axis in the first place. The further out from the center the less likely they are to be printable.
Greg
Return to “Curvemeister 101 March 2009”
Who is online
Users browsing this forum: No registered users and 8 guests