[psgbg]

Can someone explain how this works? As far I understand the glasses contain a filter for each lens.

So if a completely colour blind (who can't tell any colour just brightness) will see the filtered image, but that image should be like subtle a change in the depth.

Sounds a little weird to me.

[beagle3]

Real world color is not actually 3 channels. It's an infinite number of channels (not all equally informative, of course - we've evolved to see 3 of the most informative ones as far as humans living in nature is concerned). Healthy people see only 3 channels[1], which are usually referred to as "X", "Y", and "Z", but which can quite faithfully be represented with additive combinations of Red, Green and Blue .. RGB (as done in CRTs and DLP projects, and individual LEDs) - or subtractive combinations of Cyan, Meganta and Yellow .. CMY, as done in print[0] and on LCD screens.

Color blind people see two channels (different causes for color blindness select different channels). Very few people can see 4 daylight channels (google "tetrachromat"). IIRC there is also a one-channel color blindness (at least theoretically) but it is ultra rare.

Now, what these glasses probably do (video won't work for me, can't see if they give a description), is that the filters make sure you see three different channels among your 4 input ones (2 eyes * 2 channels). It doesn't necessarily give a depth effect - more like a semi-transparent overlay saying "this greenred here is actually green" or "this greenred here is actually red".

And, yes - these filters necessarily select one (or two) channels which normal people can't actually see.

[0] In print you also add a Black (K) because it makes life easier, and because CMY subtraction dynamic range is harder to get right.

[1] there's another channel for darkness optimized for moonlight, which is not one of the daylight three.

[agumonkey]

So, so... our color system is, pardon my maths, some kind of simplified Fourier encoding of the frequencies (real) into a finite number of fixed frequency component ?

[beagle3]

It's just a 3-d subspace of the continuous (infinite dimensional) input space. If you start in the em frequency domain (which is the only sensible thing to do) nothing is cyclic, so it is not customary to call them Fourier coefficients - but I think you got the concept.

[agumonkey]

A tiny bit. From the point of view of my ignorance I'm still thinking you might troll me :)

[geon]

The photo cells in the eyes aren't just sensitive to red green and blue, but a range of reds, greens and blues.

As I understand it, people with mild red-green color blindness have trouble seeing color because the sensitivity range of their red/green photo sensitive cells overlap too much.

The filter in the glasses cuts out a portion of the red/green spectra, so that the overlapping is reduced. That makes the red photo sensitive cells receive only the deeply red light, and vice versa, creating better color channel separation.