A question I've had for quite some time now when post-processing is, how do we know what "reality" is for color? When you "correct" in RAW, what are you "correcting?" How does one know the color temp is the "correct" one?
I'm not a graphics/digital camera/software expert, but I hope these answers help...
There's no magic to it. A RAW image is simply what is "seen" by the sensor in the camera -- without any compression at all. If you learn a little about Bayer arrays, you'll see how photos can have different levels of image noise in the various color channels (green vs. red vs. blue). Learning a little about
how S-, M-, and L- cones work in the human eye might also be an interesting diversion. If you're interested, Prof. John Dowling wrote a wonderful intro to neuroscience book entitled
Neurons and Networks. Here's a
webpage that demystifies sensor design. Nikon and Canon have their own proprietary RAW image data formats which you can imagine complicates the software programming for the various RAW image editors.
The RAW image data is nothing more than a huge matrix of numbers -- with light intensity of a particular wavelength associated with each sensor (red, green, or blue) at a given XY coordinate. Color correction via a graphical editor is just a transformation of those numbers.
Theoretically, the best method to ensure accurate color representation in post-processing would be to go through a calibration process. More specifically, one would bring along a color board on the dive and incorporate that color board in the picture at the same distance/depth/light angle as the subject. Then, in post-processing, the photographer would tweak Photoshop settings (channel mixer/levels/color balance) so that the color board at depth "matches" the true colors of the color board at the surface. I think we can all agree that this kind of calibration is probably logistically impractical (please hold still, Mr. Fishy) and more work than most casual photographers want to do.
Just for kicks, I downloaded the two unprocessed pics (above) and processed them in Elements using the "Underwater" filter that someone (can't remember who) made for PhotoShop.
I'm glad you did this, Peter. This supports the notion that JPGs, as well as RAW images, can be white balanced, doesn't it?
The Underwater "filter" you described probably isn't a "filter," per se. I'd be willing to bet that it's actually a Photoshop Action, a series of image manipulations that can be run by pressing the "play" button in the Action palette. The Action is probably a version of the Mandrake method which utilizes pixel info from the green channel to construct a replacement for the "missing" red channel.
BTW, this is a great example of how Photoshop Actions can be used to streamline repetitive tasks. I wish more people would look "inside" these Actions to see exactly what Photoshop is doing with each step. Much can be learned in this way...but very few people take the time to do so.
Here's a
nice YouTube video of a similar UW color correction technique that utilizes channel mixing and level adjustment. If you had the inclination, you could create your own Photoshop Action based on this technique. Enjoy the narrator's accent.