Gollnick,
This image is featured in the gallery on your Web site:
That picture is on my site, yes. But, like many of the pictures on my site, it was submitted by a reader. I didn't take it.
Now, let's talk about what happens when you adjust for a color cast in Photoshop.
Each pixel in a digital picture is a vector of three numbers which represent how much red, how much green, and how much blue make up that pixel. All colors of the rainbow can be made by mixing red, green, and blue. Most "better" "consumer" digital cameras have eight bits for each of those three numbers for each pixel. "Prosumer" gear gets up to ten or twelve bits. High-end pro gear like the PhaseOne backs that Dr. Darom is fortunate enough to use get up to 16 bits. But let's talk eight bits for a moment because that's what most of us probably have. An eight bit number can represent 256 levels. Think of the gas guage in your car. How many marks does it have between full and empty? An eight-bit number can represent 256 levels between full and empty. So, your camera can represent 256 levels between black and red, 256 levels between black and green, and 256 levels between black and blue. Put the three channels, red, green, and blue, together and you can represent 16,777,216 colors. That's a lot of colors. But you need that many because detail in pictures is -- what? -- variation in color. If you want the appearance of fine detail, then you've got to have lots of colors. If you want the appearance of fine focus in a color image, then you've got to have lots of colors.
So, let's say that I take a picture under poor light -- let's use old-fashioned florescent lights -- and the light is 25% too green. How does Photoshop fix that? Basically, it knocks 25% off the green value of every pixel.* But, because these are fix-point numbers, eight bits, that means we can no longer represent 256 levels of green. We can only represent 192 levels of green. And that means we can no longer represent 16,777,216 colors but only 12,582,912. That's still a lot of colors, but it's less than we had and that means that we will loose fine detail and apparent focus.
Let's say that I didn't use enough light and so the picture is a bit dark? Let's say that it's 25% dark. How does Photoshop fix that? It adds a fixed amount to each channel of each pixel.* In the case of a 25% increase to 8-bit fixed-point numbers, that means adding 64 to each channel of each pixel. But that essentially removes 64 values of variability from each channel. And that reduces the number of colors we can have in the picture to just 7,077,888, less than half of what we started with. And that, again, reduces fine detail and apparent focus.
If you did both of these operations to the same picture, then there'd only be 4,718,592 possible colors just a over a quarter of what we started with.
The old adage about free lunches holds very true.
There's an old proverb that says, "If clear water flows from the spring, then there is no need to filter it downstream."
You might say, "Well my camera has a feature that let's me adjust white balance." The way that control works is it moves the offset voltage going into the amplifiers that are in each channel, red, green, and blue, between the CCD and the Analog-to-digital converter. What comes out of the CCD for each pixel is three voltages. One voltage represents the level of red, one the level green, and one the level of blue. Those small voltages need to be amplified before going into the A/D converter which converts them into digital values. By moving the offsets on those amplifiers, you can compensate for poor whitepoint. But the free lunch concept bites us again. If you move those offsets to much, you can cause the amplifier to clip. That will also reduce your effective color depth, but it does something worse than that: when clipped signals get processed through such algorithms as compression and decompression, they mathematically introduce noise into the signal.
What happens in your camera when you try to compensate for low light by turning up the "ASA" setting? You increase the gain of those amplifiers. But the higher the gain of an amplifier, the more noise it adds to the signal.
So, the very best thing you can do is start with good light so that you don't have to compensate by moving the offsets on those amplifiers, use a lot of light so that you can keep the gain on those amplifiers low, do not compress the image, and do as little adjustment of it in Photoshop as you absolutely can.
There's another old adage: every action has an opposite and equal reaction. When you use a tool like Photoshop, every image-altering action you make has a reaction; every good comes with some bad. To use tools like Photoshop well, you have to understand what the reaction to every action is. I sometimes correct colors in Photoshop. But I know when I do that that I'm reducing color depth. If it's just a little change, then I may decide to take that tradeoff, especially when targetting the web where compression will eat into your color depth anyway. But I make that decision aware of what the tradeoff is.
A highly knowledgable pro like Dr. Darom knows all of this much, much better than I do and can make those tradeoffs even better than I can.
And that's part of what Lark is concerned about. They do not say that they are for some reason philosophically opposed to digital editting. No. They say that they're gonna edit the pictures. They just want their pro to do it and not some amateur. And they want someone who has done this targetting high-quality printed media.
* The actual algorithms used in better editting software such as Adobe Photoshop are slightly more sophisticated than this, but the principle is the same and the effect, that of reducing color depth, while reduced in magnitude some by better algorithms, is the same.
