rutt
Dec-09-2006, 07:13 PM
Introduction
Here is a meta thought. People who are reading your books for the first time may take a while to understand what you are about. I have to explain all the time that you are not offering recipes the way other Photoshop authors do, but rather teaching people to think about color correction. In the process, you consider case studies and offer techniques. But the successful reader will come away with much more than just a catalog of techniques.
Dan wants us to know he is going to teach us to make photographs look very much better than the originals. But he is also going start us on the road to thinking about how to improve our images as much as possible, to make them look as good as or better than he, Dan Margulis, could make them look. No, it won't be enough to read this book once and read the L.A.B. book once and practice constantly and go to Dan's classes. That hasn't been enough to teach me to make corrections which are competitive with Dan's, at least not more often than once in a long while. Mostly when we both correct an image, his cuts mine.
But I have learned how to think about color correction in a way that allows me to innovate and solve problems I have never encountered before. I have learned to love processing images and have come to be proud of my results (usually.) Mostly, I've learned to approach image correction as a craft with a huge engineering component that repays clear thinking about light, color, and vision.
You can take this as far as you like. In the first five chapters or so, the book covers simple techniques which will help nearly every image. Learn these and you'll be able to make quick corrections which will offer dramatic improvements.
Venture further and you'll be on your way to retouching as a competitive sport. This is not for everyone, but I've found it very rewarding.
Chapter 1.
If we alter the photograph to make it look like what we would have seen if we had been in the position of the camera, it will look better. That's the whole case for aggressive color correction in a nutshell
http://colortheory.smugmug.com/photos/115777776-L.jpg
Of swine and vision The book starts out with a before/after of one of Dan's photos from a recent trip to Italy. He doesn't say so, but people familiar with Dan's ideas will know that he has used curves to set light and dark point and and remove remove a slight yellow cast and then used USM to sharpen. It isn't the best that Dan knows how to do; it is the minimum he knows how to do. Anyone who looks at the two versions of this image will agree it looks more realistic after these corrections.
Since Dan starts with a before/after of one of his shots, I'll do the same and offer a before/after of one of mine. I've done more different things to this shot than Dan did to his, but everyone should agree that it looks more realistic after these corrections.
More realistic? What does that mean exactly? The camera is a mechanical device "sees" in a predictable and objective way. The colors and contrast it records, well aren't those what was actually there when the shot was exposed. Supposing that the exposure and focus were set properly, how can we possibly improve on that?
Our vision is very different from a photograph.
We instantly and unconsciously allocate contrast to see more detail in the objects that are of more interest to us. This is called simultaneous contrast. The camera can focus. It can expose. It cannot easily allocate contrast to suit interest.
Our vision is self-calibrating. This is called chromatic adaption. We see ambient light as neutral even when it is quite distinctly colored. Look at a gray card under tungsten, fluorescent, halogen, daylight, strobe, and stage lighting. It will always look gray, not yellow, green, blue, whatever. Anyone who has ever struggled with color balance knows that's not what the camera sees.
We are very good at distinguishing the edges of objects even in low light and even when they are of like color to their backgrounds. How many times have you been disappointed by a shot where the subject didn't stand out well against the background or other elements of the shot.
We see in almost impossibly high dynamic range compared to any existing camera (or output device for that matter.) People back-lit by bright noonday sun? No problem, we can see enough facial detail to know if they are happy to see us or not. Reflections on skin (as from a flash)? We just don't see them. Even the best camera, on the other hand, is a poor instrument by comparison. Just try taking a picture into or out of a window in daylight.
The goal of color correction is to make the product of the camera better match what we would have seen if we'd been there. This is not completely a scientifically objective process.
Casts When photos show the impact of ambient light on known colors, this is called a cast. Cameras also see mixed casts from mixed lighting sources and reflections which we might not see. Dan says that green casts always look stupid, but that some warm casts are desirable.
Contrast Generally we like more contrast through the area of interest in a photo. It is possible to disagree about what is the area of interest.
Sharpening Just how much sharpening looks good where people have quite a bit of difference. Dan is sort of on one end of the spectrum -- he like a lot of sharpening and doesn't care if it's sometimes visible. But there is nobody who prefers no sharpening so long as the sharpening is effective.
Dan has his famous hog to illustrate the ways in which people agree and disagree. His statements about the hog are backed up by a lot of data which he has collected over the years. I have a few of my own examples, backed up by relatively little data.
Following an image which Ginger posted posted (left) and my correction (right). I was very confident that nobody would prefer her version to mine. I carefully conformed to Dan's rules. I got much better contrast, particularly on his face. I corrected the cast to make the jeans blue and the fleshtones closer to plausible values.
http://colortheory.smugmug.com/photos/115804350-M.jpg
Turns out I wasn't quite right. Ginger liked her yellow cast and at least one other person thought it was a more realistic take on artificial street light. Surprising. I can't see this, but that's what makes it interesting. I wonder what would happen if we took a poll?
But here is another comparison. Left is a naive B&W conversion of Ginger's image. I just did Image->Mode->Gray-Scale. The image right is the green channel with a pretty steep curve applied (and NO additional sharpening). Which is better?
http://colortheory.smugmug.com/photos/115804346-M.jpg
The point here is that better contrast is nearly always preferable even when there is room to disagree about proper color.
Channel structure
The primaries in light are red, green, and blue. We can mix these different colored lights to get any color by altering their relative amounts. Monitors and slide projectors work this way. But when we make prints, the situation is different. Red ink reflects just red light and blocks the other two primary colors. Similarly for green and blue inks. So what if we want a yellow area in our image? How can we mix our primary colored inks to make yellow? Yellow light is composed of red and green light. Suppose we try to mix red and green ink? What happens? Well, actually, we get black. Why? Because red only reflects red and not green and visa versa. Put them together and nothing is reflected, ergo black.
So for printing, we need inks that only block one and reflect two primary colors. We call these colors cyan, magenta, and yellow. We say they are the opponent colors of red, green, and blue (respectively). Cyan blocks red and reflects the other two colors, &etc. Mixing cyan and magenta inks results in an ink that blocks red and green light, reflecting only blue. So in a world of perfect ink, we'd be in business with these three colored inks. Unfortunately, inks aren't a perfect as light, but that's a topic for a later chapter.
For now, we need to know that almost all printing processes add a fourth black ink (called K to differentiate from blue). So the CMYK color space was initially designed to model printers.
The quiz Don't panic. It's easy to be spooked by this. Don't let it get you. Nobody gets this right the first time. Not many people get it 100% right the tenth time. The big point here is that every image comes with 10 different B&W versions, or channels which might or might not come in handy for that particular image. In the case of the flower example, which is the best B&W? The answer depends on how dark you think those flowers should be. The cousin channels, red and cyan have by far the best detail in the flowers, but the flowers are also light. How to get this good detail into the dark red flowers, well that's a good challenge that we'll be working on during later channels.
One important take-away from the quiz is how similar the opponent color channels are. The red and cyan, the green and magenta, and the blue and yellow channels look nearly the same. Get your mind around this and why it's true and I think you will have learned something.
A second important take-away is the black channel. This version of the image which captures the edges so well will turn out to be a very useful tool if you keep it in mind.
There is a lot of detail which I haven't covered here. But there is nothing that we won't revisit in greater detail later.
Homework
Examine the channels in RGB and CMYK of some of your images. People, skies, flowers, buildings are all good examples.
Try some of Dan's experiments. Copy the contents of one channel into another. What happens? Why? Post examples that baffle you.
Dig through recent dgrin posts and see what you think about the color and contrast. Are there any colors that look really wrong to you? What about contrast? Think about why
Here is a meta thought. People who are reading your books for the first time may take a while to understand what you are about. I have to explain all the time that you are not offering recipes the way other Photoshop authors do, but rather teaching people to think about color correction. In the process, you consider case studies and offer techniques. But the successful reader will come away with much more than just a catalog of techniques.
Dan wants us to know he is going to teach us to make photographs look very much better than the originals. But he is also going start us on the road to thinking about how to improve our images as much as possible, to make them look as good as or better than he, Dan Margulis, could make them look. No, it won't be enough to read this book once and read the L.A.B. book once and practice constantly and go to Dan's classes. That hasn't been enough to teach me to make corrections which are competitive with Dan's, at least not more often than once in a long while. Mostly when we both correct an image, his cuts mine.
But I have learned how to think about color correction in a way that allows me to innovate and solve problems I have never encountered before. I have learned to love processing images and have come to be proud of my results (usually.) Mostly, I've learned to approach image correction as a craft with a huge engineering component that repays clear thinking about light, color, and vision.
You can take this as far as you like. In the first five chapters or so, the book covers simple techniques which will help nearly every image. Learn these and you'll be able to make quick corrections which will offer dramatic improvements.
Venture further and you'll be on your way to retouching as a competitive sport. This is not for everyone, but I've found it very rewarding.
Chapter 1.
If we alter the photograph to make it look like what we would have seen if we had been in the position of the camera, it will look better. That's the whole case for aggressive color correction in a nutshell
http://colortheory.smugmug.com/photos/115777776-L.jpg
Of swine and vision The book starts out with a before/after of one of Dan's photos from a recent trip to Italy. He doesn't say so, but people familiar with Dan's ideas will know that he has used curves to set light and dark point and and remove remove a slight yellow cast and then used USM to sharpen. It isn't the best that Dan knows how to do; it is the minimum he knows how to do. Anyone who looks at the two versions of this image will agree it looks more realistic after these corrections.
Since Dan starts with a before/after of one of his shots, I'll do the same and offer a before/after of one of mine. I've done more different things to this shot than Dan did to his, but everyone should agree that it looks more realistic after these corrections.
More realistic? What does that mean exactly? The camera is a mechanical device "sees" in a predictable and objective way. The colors and contrast it records, well aren't those what was actually there when the shot was exposed. Supposing that the exposure and focus were set properly, how can we possibly improve on that?
Our vision is very different from a photograph.
We instantly and unconsciously allocate contrast to see more detail in the objects that are of more interest to us. This is called simultaneous contrast. The camera can focus. It can expose. It cannot easily allocate contrast to suit interest.
Our vision is self-calibrating. This is called chromatic adaption. We see ambient light as neutral even when it is quite distinctly colored. Look at a gray card under tungsten, fluorescent, halogen, daylight, strobe, and stage lighting. It will always look gray, not yellow, green, blue, whatever. Anyone who has ever struggled with color balance knows that's not what the camera sees.
We are very good at distinguishing the edges of objects even in low light and even when they are of like color to their backgrounds. How many times have you been disappointed by a shot where the subject didn't stand out well against the background or other elements of the shot.
We see in almost impossibly high dynamic range compared to any existing camera (or output device for that matter.) People back-lit by bright noonday sun? No problem, we can see enough facial detail to know if they are happy to see us or not. Reflections on skin (as from a flash)? We just don't see them. Even the best camera, on the other hand, is a poor instrument by comparison. Just try taking a picture into or out of a window in daylight.
The goal of color correction is to make the product of the camera better match what we would have seen if we'd been there. This is not completely a scientifically objective process.
Casts When photos show the impact of ambient light on known colors, this is called a cast. Cameras also see mixed casts from mixed lighting sources and reflections which we might not see. Dan says that green casts always look stupid, but that some warm casts are desirable.
Contrast Generally we like more contrast through the area of interest in a photo. It is possible to disagree about what is the area of interest.
Sharpening Just how much sharpening looks good where people have quite a bit of difference. Dan is sort of on one end of the spectrum -- he like a lot of sharpening and doesn't care if it's sometimes visible. But there is nobody who prefers no sharpening so long as the sharpening is effective.
Dan has his famous hog to illustrate the ways in which people agree and disagree. His statements about the hog are backed up by a lot of data which he has collected over the years. I have a few of my own examples, backed up by relatively little data.
Following an image which Ginger posted posted (left) and my correction (right). I was very confident that nobody would prefer her version to mine. I carefully conformed to Dan's rules. I got much better contrast, particularly on his face. I corrected the cast to make the jeans blue and the fleshtones closer to plausible values.
http://colortheory.smugmug.com/photos/115804350-M.jpg
Turns out I wasn't quite right. Ginger liked her yellow cast and at least one other person thought it was a more realistic take on artificial street light. Surprising. I can't see this, but that's what makes it interesting. I wonder what would happen if we took a poll?
But here is another comparison. Left is a naive B&W conversion of Ginger's image. I just did Image->Mode->Gray-Scale. The image right is the green channel with a pretty steep curve applied (and NO additional sharpening). Which is better?
http://colortheory.smugmug.com/photos/115804346-M.jpg
The point here is that better contrast is nearly always preferable even when there is room to disagree about proper color.
Channel structure
The primaries in light are red, green, and blue. We can mix these different colored lights to get any color by altering their relative amounts. Monitors and slide projectors work this way. But when we make prints, the situation is different. Red ink reflects just red light and blocks the other two primary colors. Similarly for green and blue inks. So what if we want a yellow area in our image? How can we mix our primary colored inks to make yellow? Yellow light is composed of red and green light. Suppose we try to mix red and green ink? What happens? Well, actually, we get black. Why? Because red only reflects red and not green and visa versa. Put them together and nothing is reflected, ergo black.
So for printing, we need inks that only block one and reflect two primary colors. We call these colors cyan, magenta, and yellow. We say they are the opponent colors of red, green, and blue (respectively). Cyan blocks red and reflects the other two colors, &etc. Mixing cyan and magenta inks results in an ink that blocks red and green light, reflecting only blue. So in a world of perfect ink, we'd be in business with these three colored inks. Unfortunately, inks aren't a perfect as light, but that's a topic for a later chapter.
For now, we need to know that almost all printing processes add a fourth black ink (called K to differentiate from blue). So the CMYK color space was initially designed to model printers.
The quiz Don't panic. It's easy to be spooked by this. Don't let it get you. Nobody gets this right the first time. Not many people get it 100% right the tenth time. The big point here is that every image comes with 10 different B&W versions, or channels which might or might not come in handy for that particular image. In the case of the flower example, which is the best B&W? The answer depends on how dark you think those flowers should be. The cousin channels, red and cyan have by far the best detail in the flowers, but the flowers are also light. How to get this good detail into the dark red flowers, well that's a good challenge that we'll be working on during later channels.
One important take-away from the quiz is how similar the opponent color channels are. The red and cyan, the green and magenta, and the blue and yellow channels look nearly the same. Get your mind around this and why it's true and I think you will have learned something.
A second important take-away is the black channel. This version of the image which captures the edges so well will turn out to be a very useful tool if you keep it in mind.
There is a lot of detail which I haven't covered here. But there is nothing that we won't revisit in greater detail later.
Homework
Examine the channels in RGB and CMYK of some of your images. People, skies, flowers, buildings are all good examples.
Try some of Dan's experiments. Copy the contents of one channel into another. What happens? Why? Post examples that baffle you.
Dig through recent dgrin posts and see what you think about the color and contrast. Are there any colors that look really wrong to you? What about contrast? Think about why