Retinex is an example of how the colours we perceive, are not the same colours as we're actually experiencing. Hopefully the site is back. The following image is composed of only red and white: http://www.wendycarlos.com/colorvis/gridbig.gif

There is NO green or blue - load it into photoshop, and check with the dropper tool!

re calibration - if your scans look OK on screen, then they're OK on screen. On the other hand there's no reason to assume they'l look OK when you print them, or send them to someone else. There are (at least) three problems reguarding calibration. First off we have the (false) notion of red, green and blue. Let's skip past the problems of three colour representation, and then ask WHICH three colours? "red" is a pretty vaugue concept. sRGB (a standard for computer monitors) defines the colour of "red" on a monitor. Thats not the same as the red sensor in your camera or scanner. Your printer doesn't even have a red cartridge - it has CMY and K. All of these colours are defined in terms of cieXYZ which are a set of "standard" primary colours. If you know what your red channel represents in terms of cieXYZ you can map it to any other set of primaries.

However we still have the problem of the relative intensity of the three channels. White is defined as "equal" amounts of (say) RGB. Unfortunatly our eyes will accept just about anything as "equal" - thats why we can call daylight "white" and a lightbulb "white", even though they're actually very different colours. Again there are standards defining "white" as a cieXYZ colour mix. Your monitor probably has a "color temperature" setting, allowing you to select things like 5900K and 9000K. These are the differences between tungsten lighting and daylight - try chainging it: your images will look radically different, yet they're all just as "right". The different temptratures on a monitor look very "wrong" for about 30 seconds, then your eyes adjust, and they all look very similar. The idea of white also mixes with the idea of Gamut - monitors,printers, photographic paper, slides etc all have a physical limit on the colours they can represent. This can throw the colour balance off.

Finally there's gamma. you would think that a pixel value of 1 would be twice as bright as one of value 0.5. This is only true in a linear colour space. Most digital image files are stored in a logarythmic format, which hopefully matches your monitor's logarythmic response. There are LOTS of reasons to store images in a linear format, but it's rarely done. You can however measure the gamma of your monitor, and adjust your images so they display correctly on it (for example the machine I'm on at the momment has a gamma of 3 - rediculously high, so all images are VERY dark).

Sorry if this drifted into slightly digital territory - the principles are all actually analouge, and I'd guess most people who print colour, encounter this stuff. It's imply the for most of us the guy down the lab deals with it all (or not). Part of scanning, and printing colour at home is that you bring home all the problems...

(colour is one of my pet rants...)

Ian