OnFri, 10 Jan 2020 22:09:04 -0700, "Wayne Bretl" <waynebretl@cox.net <mailto:waynebretl@cox.net>> wrote:

There is a problem. "There is no *color* a camera [I would say sensor] can't see" implies that cameras see colors. This would imply they see colors, but still (as we have said) put out signals that aren't colors. (Why not?)

Cameras sense spectra. To me, to be consistent, you must say that sensors sense ("see") spectra, and when their signals are converted to a numerical color specification, may have certain ranges of color they cannot represent, and/or produce certain numerical values that are not colors (outside the human spectral locus), and/or represent colors that are outside a specific color space.

This is the crux of the problem. Graeme Gill touched on it as well:

Fri, 10 Jan 2020 09:36:00 +1100

You can of course draw parallel phenomena to color for other "beings", including other animal, artificial constructs etc., and call it "color". But it is probably not the same as human color, unless it has been deliberately constructed to be so.

For the purposes of most of the discussion here, we're talking about the tri-stumulus perception of light that humans experience somewhere in the retina of their eyes. Our overall vision is not so easily characterized as a set of spectral sensitivities, since it involves a lot of extra neurological processing.

Absolutely true. Graeme Gill also wrote: Fri, 10 Jan 2020 10:07:56 +1100

As before, it comes down to your definition of "gamut". For a Protanope/Deuteranope/Tritanope observer, their spectral gamut is one whole dimension less than most humans.

A two channel sensor certainly can't distinguish a full range of color (where "color" means a normal human observers perception of light). But this range can't be characterized by a neat bounding box in tri-stimulus space called a "gamut".

It can only be characterized by an accuracy limit boundary drawn in spectral space, that is a property of both the sensor itself, and whatever mechanism (i.e. profile) used to convert the sensor signals into color values.

It's unlikely that there is a corresponding 3 dimensional color volume, since the dimensionality of the spectral space is much higher, making a 1:1 mapping topologically impossible. The best you could do is color + extra dimensions. i.e. a particular color may be out of gamut when it has one spectral composition, but in gamut when it has another. The spectral composition forms the extra dimensions of such a "color gamut" representation.

Just as for for a Protanope/Deuteranope/Tritanope observer, whose spectral gamut is one whole dimension less than most humans, for other species, their spectral gamut can be an entire dimension larger than that of most humans. So yes, there are “colors" that exist that humans cannot see. They are outside the “gamut” of the human observer. Are they “imaginary?” Not to the species that see them. Birds have a tetrameric visual system, with cones dedicated to UV light. Can the colors they see be differentiated using a tristimulus-based camera sensor (or by humans)? Nope. But we can get an idea of what they see by using a multispectral camera with specially designed filters that mimic avian tetrameric vision. Birds Can See a 'Colour' Humans Can't. Now Scientists Have Revealed This Hidden World https://www.sciencealert.com/birds-can-see-uv-light-now-scientists-can-show-... (https://www.nature.com/articles/s41467-018-08142-5 if you want the details.) Do camera sensors always map “all input stimulus” to an output? Not if you are an astrophotographer and your camera sensor has an IR cutoff filter, as many do. Wayne and Graeme’s carefully worded descriptions are right on the money. But the real question remains - does this question of whether scanners (or cameras) have a "color gamut" have any practical usefulness? For most photographers or printers producing work product for consumers, on a practical basis, probably not. But I’m sure this MMA color management event will continue for many more rounds. ;-) —Rich Wagner