Chapter 3, Professional Photoshop, 5th Edition
DavidTO
Registered Users, Retired Mod Posts: 19,160 Major grins
Chapter 3
Color by the Numbers
Let's start with a couple of definitions.
Those definitions don't appear in this chapter, and Margulis doesn't talk about those terms, but they're a good starting point for this summary. Digital photography is the art or practice of using discrete values to take and process photographs. These discrete values are numbers, and that's what this chapter is all about.
To get properly balanced, color corrected images calls for little artistic judgement. Here are the basics of color correction. The building blocks.
As Margulis says in this book, "The by-the-numbers rules can be stated in a single sentence.
Margulis has actually had good results with teaching a color-blind man to correct images that held up favorably against professional color correctors. This is possible because the numbers don't lie.
When we first start taking pictures we have little knowledge or interest in the magic at work in producing a fine image. But that changes as our love for photography and our need to reproduce our artistic vision grow. We become aware of the numbers that shape our images. We find out what neutral means, in numbers. In RGB we know a color is neutral when all values are equal, as in R117 G117 B117. That is a neutral gray. White, black and shades of gray are all called neutrals. In CMYK a neutral color would be expressed by equal numbers for magenta and yellow, but a slightly higher number for cyan.
But let's back up. I'm actually going to be spending little time in this summary going over some of the hard details in the chapter. That's what the book is for. Margulis suggests how to set up Photoshop and why. He talks about how to use the book. We're going to cover none of that.
What we are going to cover are the concepts. And we've already gotten past the most important one, the concept from which everything flows. That is: the numbers don't lie. Your monitor lies to you. Your printer lies to you. Or at least they each give you a different vision of reality. It's like watching Rashômon, a story told from different perspectives. And with each retelling, the story is changed, and told in a different light. Great movie, by the way. Rent it. You can learn a lot about color theory from that black and white film (my thought, not Margulis').
So if you can't trust your monitor, your printer, or any other mechanical device to tell the same story, all that's left is the numbers that underlie all of those devices.
But there's a catch. There is a way in which all of these devices, and especially the camera tell the truth in a way that our eyes don't. Our vision adapts and adjusts to our surroundings. That's why when we're indoors we don't see the world as having an unbearably yellow cast to it, as our camera might. This happens all the time. The camera faithfully records the light present in the scene, but that's not how WE see it, since our vision system adjusts.This concept is covered in Chapter 1, and in this chapter we start to put it to use.
In this chapter, Margulis gives an example of a macaque (monkey, to you and me) deep in the jungle. Being surrounded by the yellow-green leaves of the foliage, the macaque ends up with a yellow-green cast. A macaque is largely neutral, but not in the midst of all that yellow-green light. Deep in the jungle our eyes adjust, and would see the macaque as neutral, compensating for the overall cast of the scene. But the camera faithfully captures the yellow-green cast.
So what we want to accomplish, usually, is a picture that is as we would have seen it with our eyes, not as the camera sees it. The camera might be more faithful to the scene (showing the green reflected by the foliage onto the macaque's face), but we are human, with vision that adjusts in ways the camera never could, so a good by-the-numbers color correction will show us what we expect to see.
Margulis explains a lot of details that we won't go into here, about shadows, highlights, ink limits and the settings one should use in Photoshop. Again, we're not going there.
The important concepts in this chapter are: Looking for the Sure Thing, Picking Significant Target Points, The LIghtest Significant Part, and Betting the Image
The Sure Thing: Neutrals. If you KNOW a portion of an image should be neutral, or darn near to it, that's a sure thing. Fleshtones are a nearly sure thing. You know that there will be more yellow than magenta, and you certainly know what a skin tone is NOT. For instance, it's NOT green. Margulis goes over some rules of thumb for different shades of skin. The process of looking for sure things is a matter of being sensible. A cat's tiger markings could be gray, but we can't be sure of it, all on it's own. We know that the markings can't be blue. Well, that's something.
Picking Significant Target Points: Using the eyedropper is how we read the numbers at any given point. So you've found some points in your image that you would call a Sure Thing. A neutral point, maybe some skin. What you want to do in finding Significant Target Points is to find representative points in those areas, and not rely on a single point in your neutral portion, for example. When measuring skin tones you wouldn't pick out that zit as a good spot to measure, would you? Likewise you wouldn't pick a portion of the skin that's on any extreme of light, dark, red or yellow. Pick several representative points, and use all of them to determine the proper correction.
The Lightest Significant Part: This one is easy. Use the significant portions of your image to determine how light the image should be. In other words, don't use that window in the background of your portrait to determine the brightest point, because that will cause the fact to be too dark. And what, after all, is the subject of a portrait? The person.
Betting the Image: This is a "put your money where your mouth is" statement. What are you willing to bet the image on? Is a brownstone building really brown? Or is it gray? Would you bet your image on gray? If it made the horses in the shot pink, would you still stick with gray for the brownstone?
To illustrate, here's a shot I took in San Francisco. I've outlined in red the area that was so blown I couldn't even recover it in RAW conversion. Outlined in green is the area that should be ignored, as it's not significant to this image. In this version, however, I made that part significant. Do you see the cost that the rest of the image paid for that?
If I don't treat that part of the image as significant, then I've got more decisions to make. I can make those decisions based on things I'm certain of (the pipe is neutral, or nearly so, for example) and things I'm pretty sure of (the tow away sign is red). But which one's right?
The one thing you can be fairly certain of is this: the girl does not have purple skin. As for the rest...which looks best to you? Why?
Things to avoid (an abbreviated list): relying too much on a single sampling point, concerning yourself with highlights/shadows that aren't significant, getting stuck in making curves in "one way", thinking there's a magic number for fleshtones, thinking that there is "one" neutral point.
When you read the chapter, you'll see several specific examples of images given the by-the-numbers treatment, demonstrating how and when to bet the image, how and when to increase contrast and how to avoid the impossible colors while improving the image in ways you never even imagined when you started out.
There is also considerable detail on the mechanics of each color space, including ink limits for CMYK.
And just because I'm not covering the specifics of how to set up your system, using curves rather than levels, avoiding setting highlights/shadows with the eyedropper, avoiding the use of the master curve, etc., doesn't make them any less important. But like I said, that's what the book is for.
Color by the Numbers
Let's start with a couple of definitions.
Digital: relating to or using signals or information represented by discrete values (digits) of a physical quantity
Photography: the art or practice of taking and processing photographs.
Photography: the art or practice of taking and processing photographs.
Those definitions don't appear in this chapter, and Margulis doesn't talk about those terms, but they're a good starting point for this summary. Digital photography is the art or practice of using discrete values to take and process photographs. These discrete values are numbers, and that's what this chapter is all about.
To get properly balanced, color corrected images calls for little artistic judgement. Here are the basics of color correction. The building blocks.
As Margulis says in this book, "The by-the-numbers rules can be stated in a single sentence.
Use the full range of available tones every time, and don't give the viewers any colors that they will know better than to believe."
Margulis has actually had good results with teaching a color-blind man to correct images that held up favorably against professional color correctors. This is possible because the numbers don't lie.
When we first start taking pictures we have little knowledge or interest in the magic at work in producing a fine image. But that changes as our love for photography and our need to reproduce our artistic vision grow. We become aware of the numbers that shape our images. We find out what neutral means, in numbers. In RGB we know a color is neutral when all values are equal, as in R117 G117 B117. That is a neutral gray. White, black and shades of gray are all called neutrals. In CMYK a neutral color would be expressed by equal numbers for magenta and yellow, but a slightly higher number for cyan.
But let's back up. I'm actually going to be spending little time in this summary going over some of the hard details in the chapter. That's what the book is for. Margulis suggests how to set up Photoshop and why. He talks about how to use the book. We're going to cover none of that.
What we are going to cover are the concepts. And we've already gotten past the most important one, the concept from which everything flows. That is: the numbers don't lie. Your monitor lies to you. Your printer lies to you. Or at least they each give you a different vision of reality. It's like watching Rashômon, a story told from different perspectives. And with each retelling, the story is changed, and told in a different light. Great movie, by the way. Rent it. You can learn a lot about color theory from that black and white film (my thought, not Margulis').
So if you can't trust your monitor, your printer, or any other mechanical device to tell the same story, all that's left is the numbers that underlie all of those devices.
But there's a catch. There is a way in which all of these devices, and especially the camera tell the truth in a way that our eyes don't. Our vision adapts and adjusts to our surroundings. That's why when we're indoors we don't see the world as having an unbearably yellow cast to it, as our camera might. This happens all the time. The camera faithfully records the light present in the scene, but that's not how WE see it, since our vision system adjusts.This concept is covered in Chapter 1, and in this chapter we start to put it to use.
In this chapter, Margulis gives an example of a macaque (monkey, to you and me) deep in the jungle. Being surrounded by the yellow-green leaves of the foliage, the macaque ends up with a yellow-green cast. A macaque is largely neutral, but not in the midst of all that yellow-green light. Deep in the jungle our eyes adjust, and would see the macaque as neutral, compensating for the overall cast of the scene. But the camera faithfully captures the yellow-green cast.
So what we want to accomplish, usually, is a picture that is as we would have seen it with our eyes, not as the camera sees it. The camera might be more faithful to the scene (showing the green reflected by the foliage onto the macaque's face), but we are human, with vision that adjusts in ways the camera never could, so a good by-the-numbers color correction will show us what we expect to see.
Margulis explains a lot of details that we won't go into here, about shadows, highlights, ink limits and the settings one should use in Photoshop. Again, we're not going there.
The important concepts in this chapter are: Looking for the Sure Thing, Picking Significant Target Points, The LIghtest Significant Part, and Betting the Image
The Sure Thing: Neutrals. If you KNOW a portion of an image should be neutral, or darn near to it, that's a sure thing. Fleshtones are a nearly sure thing. You know that there will be more yellow than magenta, and you certainly know what a skin tone is NOT. For instance, it's NOT green. Margulis goes over some rules of thumb for different shades of skin. The process of looking for sure things is a matter of being sensible. A cat's tiger markings could be gray, but we can't be sure of it, all on it's own. We know that the markings can't be blue. Well, that's something.
Picking Significant Target Points: Using the eyedropper is how we read the numbers at any given point. So you've found some points in your image that you would call a Sure Thing. A neutral point, maybe some skin. What you want to do in finding Significant Target Points is to find representative points in those areas, and not rely on a single point in your neutral portion, for example. When measuring skin tones you wouldn't pick out that zit as a good spot to measure, would you? Likewise you wouldn't pick a portion of the skin that's on any extreme of light, dark, red or yellow. Pick several representative points, and use all of them to determine the proper correction.
The Lightest Significant Part: This one is easy. Use the significant portions of your image to determine how light the image should be. In other words, don't use that window in the background of your portrait to determine the brightest point, because that will cause the fact to be too dark. And what, after all, is the subject of a portrait? The person.
Betting the Image: This is a "put your money where your mouth is" statement. What are you willing to bet the image on? Is a brownstone building really brown? Or is it gray? Would you bet your image on gray? If it made the horses in the shot pink, would you still stick with gray for the brownstone?
To illustrate, here's a shot I took in San Francisco. I've outlined in red the area that was so blown I couldn't even recover it in RAW conversion. Outlined in green is the area that should be ignored, as it's not significant to this image. In this version, however, I made that part significant. Do you see the cost that the rest of the image paid for that?
If I don't treat that part of the image as significant, then I've got more decisions to make. I can make those decisions based on things I'm certain of (the pipe is neutral, or nearly so, for example) and things I'm pretty sure of (the tow away sign is red). But which one's right?
The one thing you can be fairly certain of is this: the girl does not have purple skin. As for the rest...which looks best to you? Why?
Things to avoid (an abbreviated list): relying too much on a single sampling point, concerning yourself with highlights/shadows that aren't significant, getting stuck in making curves in "one way", thinking there's a magic number for fleshtones, thinking that there is "one" neutral point.
When you read the chapter, you'll see several specific examples of images given the by-the-numbers treatment, demonstrating how and when to bet the image, how and when to increase contrast and how to avoid the impossible colors while improving the image in ways you never even imagined when you started out.
There is also considerable detail on the mechanics of each color space, including ink limits for CMYK.
And just because I'm not covering the specifics of how to set up your system, using curves rather than levels, avoiding setting highlights/shadows with the eyedropper, avoiding the use of the master curve, etc., doesn't make them any less important. But like I said, that's what the book is for.
0
Comments
Since you asked, I'm pretty sure that the second image (top right) is the closest to the original scene. In the first, the blue jeans are almost cyan. I think it is obviously wrong, and I would bet that the numbers show the jeans are closer to cyan than to blue. In the third, the yellows all seem off, including the skin tones, but especially in the paint on the building to the left.
There is one point I think Margulis would disagree with you on. If you have a final output in mind, then whatever result you get from that final output does not lie to you. This, I think, is another of his cardinal rules and explains why he is so sceptical of calibrationism: an "uncalibrated" monitor matched to a similarly uncalibrated printer can lead to perfectly good prints (and if the print is the final output, then it can't lie).
Duffy
If he trusted calibration he wouldn't lose sleep at the printer's making sure it comes out right.
Dgrin FAQ | Me | Workshops
Hey, if you see them, point them out!
Dgrin FAQ | Me | Workshops
Not a typo, but the "a way" is redundant. You could just say "All of these devices, and especially the camera tell the truth in a way that our eyes don't"
replace "has" with "as"
replace "ones" with "one is" (or use a contraction "one's)
I remember seeing at least one other typo when I first read this, but I'm not finding it now.
There's one other point that I think is contrary to the ideas in the book. You say that our eyes adjust to remove casts, and that the camera "faithfully" records them. The trouble with this characterization, for someone who has not read the book, is that it sounds like our eyes are somehow "unfaithful." I don't think Margulis would have said it this way (and if he did, I missed it). Instead, the impression that the book leaves me with is that the camera sees the scene differently than people do. And since we are correcting pictures for people and not machines, we should correct according to the way people see. That's the neutral way to put it. But, given a choice, I think Margulis might prefer to say that the camera stupidly records the cast, because it doesn't know any better.
Duffy
Thanks, Duffy. Fixed the typos, left the rest for discussion.
Dgrin FAQ | Me | Workshops
I like how you selected the crucial high points, leaving the reader to fill in the details from the book. In my own experience, however, I think the material on impossible colors and how to decide what constitutes a good green, red, blue or fleshtone, particularly in the absence of an identifiable neutral, is one of the crucial high points. As I see these write-ups, they're the Cliff Notes version of Dan's original. His breakdown of fleshtones on p. 61 and the "Avoiding the Impossible" sidebar on p. 72 can save lots of frustration and confusion, and, if not dealt with in detail, should at least be pointed out to guide the unaware.
But, for all that, your explanations were clear and right on target.
—Korzybski
Thanks. I'll take a look at that when I'm back at the book. Or if you want to add what you think is missing...
Dgrin FAQ | Me | Workshops
= = = = = =
The macaque
= = = = = =
On pages 52 and 53 Dan shows us the curves he used. On page 52 he merely corrects for the green/yellow cast. On page 53 he add S curves for contrast.
Could somebody please explain the selected place markers he uses?
On Page 52 are he appears to just pick the center point on each channel.
On Page 53 he selects 2 points that may or may not be the range of the area of interest. If that's the case, what are the red lines?
Thanks.
Second one's just fine.
Third one's too yellow.
Fourth one's too magenta.
What do I win?
Catapultam habeo. Nisi pecuniam omnem mihi dabis, ad caput tuum saxum immane mittam
http://www.mcneel.com/users/jb/foghorn/ill_shut_up.au
Dgrin FAQ | Me | Workshops
Catapultam habeo. Nisi pecuniam omnem mihi dabis, ad caput tuum saxum immane mittam
http://www.mcneel.com/users/jb/foghorn/ill_shut_up.au
Nice writeup, great job!
Before starting talking the content, may I make a suggestion? The four versions of the dark alley shot you put side by side made your post exteremely wide even for my wide screen monitor, it's extremely inconvenient to read :-( Can you please make them 2x2? This should narrow the post width down to a reasonable value.
Thanks!
Huh. Mine show up 2x2?
But sure, I'll put a space in there.
Dgrin FAQ | Me | Workshops
One question I have on this is whether a two pass approach would work better (pg 52 curves for color, and then a more targetted group of curves to steepen the face)? Another question I've always had with Dan's curves is how, exactly, he chooses the points that he does. Sometimes, the explanation for the placement is not in the text. This appears to be one of them. He says that the exact placement of the points is a more a matter of art and science, and that you might be able to improve on the page 53 image.
Duffy
My contention is that anyone who has only gotten through chapter 3 cannot explain it, since Dan hasn't explained it himself. Either that, or he's done a lousy job of explaining it. I'm pretty sure he has too much experience explaining for the latter to be true.
But I don't see this chapter as having explained how to do anything. It's more of a primer on what's to come, as far as I can tell. Explaining the concepts, but not the details.
Dgrin FAQ | Me | Workshops
= = = = = =
Thanks Duffy.
Lunchtime today (east coast US) I reread (again) the macaque stuff and for page 52 he does mention he started with the center point.
I assume when you wrote "Page 52 corrects the cast and steepens the area of interest considerably" you really meant to write page 53 and just did that to confuse matters (ha!).
I agree with you that I'm more comfortable working in stages: first correct the cast using curves and then a new curve to enhance. I look at this as to almost unrelated goals. Fix then enhance.
On page 53, his end points don't coincide with his red lines. I think I read somewhere that the end points of the area of interest don't match exactly the end points of the curve since you want as much as possible of the area of interest to be steeper.
Well, I'm getting a bit more comfortable in RGB (I'm a LAB convert now) but this CMYK stuff seems almost useless to me since I don't see myself outputing to CMYK. I guess I'll just force myself.
Thanks.
I guarantee you'll feel differently by the end of chapter 5.
Thanks! Now they are 2x2 for me, too:-)
= = = = =
I'm having a lot of trouble with a few images.
First, I find that Dan seems to have things stacked in his favor. When a color needs work (whether to increase or decrease) I find that color does not usually appear in the image in another location that should be left alone. That is unlike wmy own images.
Second, on some images (such as the woman in front of the window) I see nothing more than a bit of LAB required to give it the pop we want.
How can I change something that really looks fine. Even the numbers I see aren't very bad.
He picks his images to illustrate and explain what he is trying to do. It's not surprising to me that his examples fit well into what he is trying to explain. This is the first time I've heard someone complain that Dan's books are too simple!
If you can't find anything wrong wtih a picture, and you are your own client, then of course you don't have to do any correcting. Dan thought the woman in front of the window was too pink. You are free to disagree. If you do disagree, then you should also think that the original is better than Dan's correction on the next page. If you do think the original is better, I'm willing to bet you are in the minority, but that doesn't mean you are wrong.
However, if you think Dan's correction is better than the original, then the question is how can you learn to see the problem in it that Dan saw. I'm not sure what the answer to that question is, but in large part I think its a process of developing a more refined color sensitivity.
Duffy
Duffy,
You are not being fair. I don't like the choices you gave me. In your opinion, since I don't see all of the problems Dan sees in that image, then I've decided it doesn't need any work and the original is better than his final effort.
It reminds me of the college professor who over-analyzed the writings of some author. I didn't read the same book he did.
Yes, that image needs some work.
Some simple standard LAB enhancement making sure not to increase the reds does fine for me.
Oh yes, I didn't say (or mean to say) that Dan's book is too simple. Far from it. I read no more than a couple of pages at a sitting and then quite often reread them before moving ahead.
I just find that quite often when he's working on a color, that color doesn't appear elsewhere in the same image. That way when he adjusts it, there is no other area that gets ruined.
Don't get me wrong. I'm a big fan of Dan's. Even if I got nothing directly out of his books (which is far from the case), what I've learned about channel masking, blending, etc is absolutely fantastic.
And I even read these threads with great interest. One of these days I'll move back to learning how to take good photos.
Later in the book, there are some images that deal with the more complicated situation you describe: the chinese lady with the fan on 358-368, the hotel room bed on and around 448. I've read the 4th edition of this book about 3 times. If Dan had put either of these examples into Chapter 2 or 3 of this edition, I probably would have thrown the book away in disgust. It's one thing to force people to dive into the deep end right away, but that would be the equivilent of having beginners dive into a shark infested tank. As it stands, I think I will probably start to become comfortable with the corrections used on the fan lady and the bed on my 3rd or 4th go round with this book. I think he stacks the images in his favor because the ideas he is dealing with are complicated enough that it is best to present them as cleanly as possible.
Duffy
Duffy,
I'm glad to see you now gave me more choices than either the original or his final image.
I keep sticking with his approaches because I figure if LAB keeps me happy I'll miss some concepts and be more lost later.
I was amused when you referred to images in the range of pages 350-450. Well, I'm in the 65-70 range and each small breakthrough is cause for celebration.
Unfortunately I'm in this alone (so my wife is not willing to celebrate my little victories nor does she understand them) and I don't drink so celebrating is strictly a personal thing.
There must be something wrong when we put ourselves through this.
You're not alone in all senses:
first of all, many of us dgrinners are going through those pages at approximately same rate;
and, second, not too many spausal units (there are some lucky exceptions:-) understand what we do, except they see us throwing gobbles of totally usable money at some gizmos and wonder how nice would be of us to put all that cash at their immediate disposal instead...
Cheers!
So, when I talk about images later in the book, I'm also talking about corrections that made my head spin. Between reading the 4th edition three times, and the beginning of this book twice, I finally think I'm developing a better understanding of curves. At least, I'm capable of doing things now with RGB and CMYK curves now that would not have been possible for me six months ago.
And I feel alone in this as well, except for the phantoms that live in the internet and talk to me from time to time. My wife also has little to no understanding of why I spend so much time on this stuff (even though she likes the pictures).
Duffy
It's quite interesting to see the different approaches to Dan's books as well as spousal attitudes.
I feel I have to understand what Dan's doing and writing. I've read the LAB book once. But in a way that's not true. I've actually gone through it once but reread many pages several times. In the first 5 (app) chapters I didn't read the last portion. During the next reading I'll probably read the whole thing and expect it won't take me as long.
My wife feels photos should be printed as they come out of the camera. I guess the less she knows the better. She thinks nobody suffers from red-eye when I take pictures. It never occurs to her why people have red-eye when others take the pictures on their cameras.
I maintain I'm not changing the photo nor do I add different effects.
I only want the printed photo to look like what I saw when I took it.
do you get 'work on this picture' or 'why don't you work on that photo'?-
I tell my wife I'll be happy to show you how to do it-
I think I've been saying that since I got on here-
anyway, I had to make the concession of moving the dd (digital darkroom) upstairs off the kitchen because if I wasn't at work (and that is out-of-town 2-4 days a week) I was downstairs closed up in my office on the computer-
and make that a bump for this thread-
I'd rather that than the one with the castle. To me that just need a standard enhancement (LAB) and some sharpening.
Duffy
That's exactly my approach to his books. I've gone through LAB a good four times as well. With this book I run through the chapters and then go back and read them more slowly while I go over the photos he gives us to do. I also write down some notes. I never end up using those notes much but simply writing them down helps me learn the stuff better.
I just got my copy on friday and have just made it to chapter 5 with my quick read. I'm in chapter 2 with my slow read. After I finish the book with at least two reads then I will probably go back and read it again just to make sure I remember what he said and understand it enough to where it doesn't confuse me as I go over it. I try to explain what he is talking about in my own words as I read each section of each chapter.
That seems to work alright with me. It takes quite a while to get through all this stuff for sure. I often have to read specific sections multiple times when I am in my first or second read.
Overall, I do think that he could make things easier on us poor saps. Part of me loves his writing and am perfectly happy with his book and then there's the other part that wants to kill him. I think in this chapter he uses the term north to describe where to nudge the point in the curve. That is just a plain awful word to use to explain things. In RGB he meant down and then in CMYK he meant up and when I think of north, I think of up regardless of the colorspace. If he would just refrain from trying to be witty someimes and just say "move it up" then there would be no problem. This book is hard enough to add that crap.