Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!cbatt!ucbvax!INGRES.BERKELEY.EDU!hatcher
From: hatcher@INGRES.BERKELEY.EDU.UUCP
Newsgroups: comp.graphics
Subject: Colour perception
Message-ID: <8703121101.AA09386@ingres.Berkeley.EDU>
Date: Thu, 12-Mar-87 06:01:49 EST
Article-I.D.: ingres.8703121101.AA09386
Posted: Thu Mar 12 06:01:49 1987
Date-Received: Fri, 13-Mar-87 06:13:53 EST
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: University of California at Berkeley
Lines: 104

In article <505@ubu.warwick.UUCP> rolf@warwick.UUCP (Rolf Howarth) writes:
>I thought I'd ask if anyone could explain to me how the human brain
>perceives colour.

>red + green light "gives you yellow", where "yellow" is also what you see
>at a particular position in the spectrum (when you shine white light
>through a prism). As far as I understand it, these two "yellows" are different
>spectroscopically , yet the eye perceives them to be the same colour.

The human eye has three kinds of color receptors; each one is sensitive
to a range of colors but has a peak near the colors red, green, and blue,
respectively. These are more accurately called long, medium, and short
wavelength receptors (since there is less implication that they respond
to only a single pure color). If a red receptor receives exactly two
monochromatic frequencies (say from a controlled source like a laser)
that cause it to produce a 1 picovolt signal, then the brain has no way
of distinguishing that signal from the same red receptor stimulated with
a single wavelength that is by itself strong enough to also produce a
1 picovolt signal. That particular example assumes that the wavelengths
in question fall outside of the spectrum that the blue and green receptors
respond to, to simplify the issue.

In general, if a receptor receives a mix of frequencies that cause it to
output a signal of strength N, then there will be an infinite number of
other mixes of similar frequencies that also produce a signal of strength
N. That's because it only gives output depending on overall signal strength,
there is no way for it to figure out exactly which frequencies caused the
stimulation. The cones that are used in human night vision behave the same
way, but are receptive to a broader range of colors. Still, they only give
out a signal that shows relative intensity. Thus colors fade as things
get darker, and when it gets too dark to stimulate our color sensitive 
rods, we depend on our cones for vision. Since there's only one kind of
cone, this gives us black and white images. They are more sensitive to
dim light than rods because they respond to any visible color. So you
get increased sensitivity at the expense of color resolution.

Back to rods: each rod behaves like a cone, but only over a narrow range
of frequencies. Thus the brain receives signals from the rods that indicate
the relative intensity of light, as broken up into the categories red, green,
and blue. The three types of rods are distributed around the retina somewhat
the way that a color tv has triads of red, green, and blue phosphors. (The
distribution of rods in the eye is not quite so geometrically perfect as
a tv, for interesting reasons, but that's a different story)

Imagine that a yellow image falls on the retina. It will stimulate both
the green and red rods (since they are both somewhat sensitive to this
wavelength, even though not as much as they are to green or red). Both
types of rods will give an output signal to the brain in response. Now
imagine that an image that is a mixture of green and red falls on the
retina. The green stimulates the green-sensitive rod, and the red
stimulates the red-sensitive rod. So again, both types of rods give an
output signal. If the intensity of the green and red lights are at the
right level, then these output signals will have exactly the same intensity
that they did when the yellow light was shining on the retina.

Given the same signal to the brain, the brain cannot tell the difference
between these two scenarios. That's why a mixture of green and red looks
yellow to a human being, despite the fact that a spectrometer will easily
show the difference.

You might think that we are missing out on a lot of interesting color
perceptions because of this. After all, the ear's ability to hear chords
of music is one of the things that makes music so interesting. A single
voice melody is not as richly interesting as a whole orchestra. But while
this seems logical in the abstract, it turns out that, in terms of what
is available to be perceived in the natural environment, we aren't missing
much.

This is because of the nature of the types of reflective surfaces in the
world. Any flower that reflects yellow light is pretty likely to reflect
red and green, too. So even if we saw "chords" of color, the flower would
still look yellow. If you do a 3D histogram of all the colors that appear
in a digitized image of a natural scene, you will find that they all tend
to clump around a plane running from one diagonal to the other, in a
fairly even distribution. In order for "chords" of colors to really
make a difference, you would need a distribution that had the same sharp
boundaries and clusters as the notes in a symphonic score, and this just
doesn't happen. Natural surfaces don't tend to behave that way. Thus
our color vision is well suited to our natural environment.

See:
	Human Color Vision, R.M. Boynton (1979)
	Visual Perception, T.N. Cornsweet (1970)
	Handbook of Perception, ed. J. Thomas (1986)
	Color Science, G. Wyszecki & W.S. Stiles (1982)

References from Prof. Brian Wandell, Stanford Psych. dept. who was one
of a panel of speakers on color perception at a recent SIGGRAPH local
chapter meeting.

As for your question about how a TV image might appear to someone with
a color defect, yes, there are sometimes differences in perception. However,
they are almost never as dramatic as the scenario you offer, because the
human brain is extremely good at adapting to whatever input it has. After
all, a banana still looks like a banana on a black and white TV (usually!).
But some of the simpler tests for color blindness do involve multicolored
images with controlled grey scale equivalences for the different colors.
Anyone lacking a particular receptor will not be able to distinguish the
corresponding color from the background. He will still SEE it, but it
won't look different than the carefully-chosen background. It is uncommon
to find natural images with this property, so people deficient in only
one type of rod are often not really aware that their vision is different
until they are tested, because their brain adapts so well.
	Doug Merritt