Color Vision

Color Vision

Of mammals, only primates have well-developed color vision. Canines and felines do possess some color vision, although its quality and quantity is much reduced than that of humans. Human color vision, however, is rivaled (and even surpassed) by many species of birds, reptiles, amphibians and even insects!

Gary Larson -- color.jpg (55208 bytes)

Hue

Hue is that psychological dimension of color which roughly corresponds to wavelength. Hue is the common meaning of the word "color". Common color names, such as "red", "green", "yellow", etc., are all descriptive of hue, as the table below illustrates.

Wavelength (nanometers)	Associated Hue	Appearance
370-470	Reddish-Blue (Purple or Violet)
470-475	Blue
475-480	Greenish-Blue
480-485	Blue-Green (Aquamarine)
485-495	Bluish-Green
495-535	Green
535-555	Yellowish-Green
555-565	Green-Yellow
565-575	Greenish-Yellow
575-580	Yellow
580-585	Reddish-Yellow
585-595	Yellow-Red (Orange)
595-730	Yellowish-Red

The Number of Colors

National Bureau of Standards (US) lists 267 color names. Berlin & Kay (1969) studied color terms in 100 languages and proposed that there exist 11 universal color terms: white, black, red, green, yellow, blue, brown, purple, pink, orange, gray. This result was confirmed experimentally in American observers by Boynton & Olson (1987), and in both Japanese and American observers by Uchikawa & Boynton (1987). The number of discriminable colors probably exceeds 10,000. Not all discriminable differences are separated into different color categories (names).

Brightness

Colors are also described by their brightness. Brightness is the psychological dimension of color which most closely relates to physical intensity. For a given intensit, however, yellow lights appear brighter than greenish hues. The hue associated with a given light will also change with its intensity: greenish-yellows and yellowish-reds will become more uniquely yellowish as intensity is increases. Likewise, violet and blue-green will tend to appear more uniquely bluish as intensity increases. This phenomenon is known as the Bezold-Brucke hue shift. The table below illustrates the concept of brightness.

Brightness	Appearance
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Saturation

Saturation refers to the amount of hue a color possesses, and is most closely correlated with the physical attribute of spectral purity. A light consisting of a single wavelength, called monochromatic light, is spectrally pure, and appears very saturated with colored. The addition of other wavelengths to a monochromatic light makes the light appear more washed out, or pastel in color, thus desaturating the mixture. The table below illustrates the concept of saturation.

Saturation	Appearance
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Among monochromatic lights, short wavelengths (blue) and long wavelengths (red) appear the most saturated. Wavelengths of around 575 nm (yellow) appear the least saturated. To illustrate this phenomenon, ask yourself which hue, blue, red or yellow, is the most similar to white? Which is hardest to detect against the white screen? Most observers would agree that yellow is most similar to white, and the similarity is largely determined by the fact that yellow is the least saturated color of the three.

The Color Circle

Hues can be arranged in a circle, where the psychologically unique hues (e.g., red, green, blue and yellow) are represented at cardinal points. Intermediate positions are occupied by combinations (mixtures) of the unique hues in various proportions. Many other colors (not shown) would be located inside the circle, and would correspond to progressively desaturated colors; white is represented at the center of the circle.

Colors which are 180^o apart on this color circle are called complimentary colors. When complimentary colors are mixed together (additively, as lights, not subtractively, as pigments), they form a sum that approximates white (or a shade of gray). Mixtures of non-complimentary colors produce hues that are intermediate to the colors which are mixed.

The Color Spindle

The relationship between all three psychological dimensions of color (hue, saturation and brightness) can be conveniently and graphically described by the color spindle. The brightness axis runs vertically, hue is represented around the perimenter of a circle (the Color Circle, above). Saturation decreases toward the center of the spindle, where shades of gray fall along the brightness axis.

Color Mixture

While the Color Circle provides a shorthand way of evaluating color mixture, a more scientifically useful method involves determining the mixtures of specific wavelengths of light (called primaries) which match various monochromatic lights. Monochromatic light of various wavelengths is presented in the Test Field. The viewer adjusts the intensities of three other monochromatic lights (called Primaries) in the Comparison Field until the same hue is obtained. The two distributions of light energy, which are not equivalent, but which neverless look identical to a human observer, are called Metamers.

Color Matching Paradigm.JPG (54525 bytes)

Color Matching Functions

The result of performing the kind of experiment described above across a wide range of test wavelengths produces Color Matching Functions, shown below for primaries of 436 nm, 546 nm, and 700 nm. The color matching function reveal the fundamental fact about human color vision, that it is trichromatic. Trichromacy simply means that any light can be matched by a mixture of three primary lights. Depending on the particular primary lights used to derive the color mixture functions, a general equation for color appearance can be written: (C) = xR + yG + zB, where x, y and z are the coefficients (i.e., relative proportions) of the three primary lights necessary to match any color (C).

Additive Color Mixture

Here are two demonstrations of additive color mixture. In the figure on the left the primaries are red, green and blue. Where they all overlap (the bulging triangular region in the center) they combine to produce white. The demonstration on the left illustrates that any three primary lights (in this case a cyan, a purple and an orange) can similarly be used to produce white.

Subtractive Color Mixture

Subtractive color mixture occurs when you mix pigments instead of lights. As school children know, mixing blue and yellow pigments yields green. We have just seen, however, that mixing blue and yellow light yields white. What's going on? When you mix light you are adding the spectral distributions of the lights together. When you mix pigments you are in essence "subtracting" one spectral distribution from another. Imagine you have a filter that passes only blue (short-medium wavelength) light. You have a second filter that passes only yellow (medium-long wavelength) light. Put the two together. Now both filters are "vetoing" light which they do not pass -- the only light that gets through both filters is light of wavelengths which are simultaneously short-medium and medium-long. The only light that survives passage through both filters has the Goldilocks characteristic of being neither too short nor too long -- only medium wavelengths get through, and these look green. Hence, blue pigment (or filters) mixed with yello pigment (or filters) looks green.

The Color Triangle and CIE Color Space

http://www.cs.rit.edu/~ncs/color/

Spectral Sensitivity

Two factors determine the probability that a photon will be absorbed by a given photoreceptor -- the optical density of the photopigment and the "match" between the photon's wavelength and the particular photopigment's tuning. There is one spectral sensitivity curve for rhodopsin (the rod pigment, sometimes called "visual purple") and three for the three different classes of cone (cone iodopsin), as shown in the figure below. The wavelength of maximal absorbance for the "blue" cone (cyanolabe) is 445 nm, for the rod it is 507 nm, for the "green" cone (chlorolabe) it is 535 nm, and for the "red" cone (erythrolabe) it is 570 nm. The fact that there are three different cone types in the human retina explains why normal human color vision is trichromatic -- and hence why it requires three and only three primaries to match any other color. This is because the preception of all colors is simply based on the relative excitations of these three classes of cone.

Cone Distributions

The three cone types are not distributed uniformly across the retina. Below is a computer-generated model of the cone mosaic in the vicinity of the fovea. Note that "blue" (short-wavelength or S) cones are absent from the fovea, and that "red" (long-wavelength or L) cones outnumber the "green" (medium-wavelength or M) cones by a ratio of approximately 2:1.

Chromatic Aberration

The reason that blue cones are absent in the fovea is because the images of objects formed from short-wavelength light are actually quite defocussed (by as much as 2 diopters) relative to the images formed from long-wavelength light. Remember, prisms and lenses refract short-wavelength light more strongly than long-wavelength light, which is why a prism can be used to disperse the spectrum. So, in order to avoid having the sharp images of objects in the high-resolution fovea degraded by chromatic aberration, the visual system simply has no receptors sensitive to the blurred short-wavelength image in the fovea. Literally out-of-sight, out-of-mind!

Post-Receptoral Processing: The Opponent-Process Model of Color Vision

The S (blue), M (green) and L (red) cones form the basis for trichromatic color vision, and their existence explains why any color can be matched using three, and only three, unique primary lights (because different colors closely correspond to different relative rates of quantal absorption by these three cone types). There are, however, several subjective features of color perception which are not readily explained by the three-receptor theory, namely, why are there four psychologically "pure" hues: red, green, blue and yellow. Why can't the complimentary hues coexist, in other words, why don't we see (or describe) a color with the terms "greenish-red", or "bluish-yellow"? In one version of the trichromatic theory the subjective experience of a greenish-red color would simply correspond to near-equal activation of the L (red) and M (green) cone types.

The answer to the questions above is that after photons are absorbed by the three cone types, their outputs are combined to form what are called opponent channels. This is shown diagramatically in the figure below.

The Luminance Channel

The combined outputs of the L and M cones form what is called the Luminance Channel. The level of activity in this channel codes colors along the white-black (brightness) dimension of color space (the vertical axis in the color spindle).

The Red-Green Channel

The outputs of the L and M cones are differenced (subtracted from each other) to form the Red-Green Opponent Channel. Roughly half of the neurons comprising this channel are +R/-G (i.e., they excite to long-wavelength "red" light, and inhibit to middle-wavelength "green" light), and half are +G/-R (i.e., they excite to medium-wavelength "green" light, and inhibit to long-wavelength "red" light). The level of activity in this channel codes for color along the red-green axis of color space (i.e., from red, through yellow, to green). Because the output of this channel signals either red or green (and cannot signal both simultaneously), it cannot code for the percept of "reddish-green".

The Blue-Yellow Channel

The combined outputs of the L and M cones (i.e, the Luminance Channel) and the S cones are differenced to form the Blue-Yellow Opponent Channel. The majority of neurons in this channel are +B/-Y (i.e., they excite to short-wavelength "blue" light, and inhibit to medium-long-wavelength "yellow" light). Again, the level of activity in this channel codes for color along the blue-yellow axis of color space (i.e., from blue, through white, to yellow). Because the output of this channel signals either blue or yellow (and cannot signal both simultaneously), it cannot code for the percept of "bluish-yellow".

The Opponent-Process Model of Human Color Vision

Receptive Fields

Single-Opponent Cells

Luminance Channel Receptive Fields
The receptive fields of the Luminance Channel have a center-surround organization, as described in an earlier module. The RF on the left is an "on-center" type, and the one on the right is an "off-center" type. The important point regarding color vision, however, is that the spectral sensitivity of both the center and surround reflect a combination of L and M cones

On-Center	Off-Center

Red-Green Opponent Receptive Fields
The receptive fields of the Red-Green Opponent Channel have a center-surround organization. The RFs on the left are "on-center" types, and those on the right are "off-center" types. The spectral sensitivity of the center and surround the reflect only the M or L cones.

On-Center	Off-Center

Blue-Yellow Opponent Receptive Fields
The receptive fields of the Blue-Yellow Opponent Channel have a center-surround organization. The RFs on the left are "on-center" types, and those on the right are "off-center" types. The spectral sensitivity of the center and surround the reflect only the S cones or the combination of the L and M cones.

On-Center	Off-Center

Double-Opponent Cells

Blue-Yellow Double-Opponent Receptive Fields

Further along the processing pathway, in V2, double-opponent cells are encountered. In addition to demonstrating the spectral opponency of the single-opponent cells, above, double-opponent cells are "pickier", and are also spatially opponent for color stimuli. Unlike the single-opponent blue-yellow cells above, these cells will not respond to a diffuse field of either blue or yellow light -- the stimulus must be a blue spot on a yellow background (left), or a yellow spot on a blue background (right).

Spectrally-Opponent Responses

Electrophysiological recordings of the responses of parvocellular ganglion cells in the retina or LGN to flashes of light delivered to the receptive field center reveals the spectrally-opponent nature of their responses. The dashed white horizontal lines indicates the level of "spontaneous" activity of these cells, i.e., the activity associated with broadband white light. +M/-L cells excite to short- and medium-wavelength light, and inhibit to long-wavelength light. +L/-M cells do just the opposite, they excite to long-wavelength light and inhibit to medium- and short-wavelength light.

Response of +M/-L spectrally opponent cell to light of various wavelengths in the receptive field center	Response of +L/-M spectrally opponent cell to light of various wavelengths in the receptive field center

+S/-(L+M) cells excite to short-wavelength light, and inhibit to medium- and long-wavelength light. +(L+M)/-S cells do just the opposite, they excite to medium- and long-wavelength light and inhibit to short-wavelength light.

Response of +S/-(L+M) spectrally opponent cell to light of various wavelengths in the receptive field center	Response of +(L+M)/-S spectrally opponent cell to light of various wavelengths in the receptive field center

Defective Color Vision

Incidence and Type of Color-Defective Vision

Classification	Incidence (%)
Classification	Males	Females
Anomalous Trichromacy	(6.3)	(0.37)
Protanomaly (L-cone defect)	1.3	0.02
Deuteranomaly (M-cone defect)	5.0	0.35
Tritanomaly (S-cone defect)	0.0001	0.0001
Dichromacy	(2.4)	(0.03)
Protanopia (L-cone absent)	1.3	0.02
Deuteranopia (M-cone absent)	1.2	0.01
Tritanopia (S-cone absent)	0.001	0.003
Rod Monochromacy (no cones)	0.00001	0.00001

Why do males have a higher incidence of protan and deutan color vision than females? It is because the genes which code for the photopigment proteins are located on the X-chromosome. Females have two X chromosomes, and so even if one of them has defective genes, there is a "back-up" copy on the other X chromosome. Males have only one X chromosome, which they picked up from their mother, and if the genes on that chromosome are defective, then they are expressed.

Rod monochromacy is an autosomal (not linked to the sex-chromosomes) recessive trait, so it is very rare, since it requires two sets of defective genes to be expressed.

Anomalous Trichromatic Color Vision

Anomalous trichromats possess three types of photopigment in their cones, and so are truly trichromats. What makes their color vision "anomalous" is that it differs from normal trichromacy in that one pigment type is shifted along the wavelength axis from the normal position toward shorter wavelengths (dashed red line) for the protanamolous observer, or toward longer wavelengths (dashed green line) for the deuteranomalous observer, as shown in the figures below. Because anomalous trichromats possess different visual pigments from normal trichromats, they do not accept normal color matches, and vice versa. There is some variation in the exact position of the pigments along the wavelength axis even among normal trichromats.

Tritanomalous Cone Spectral Sensitivities	Deuteranomalous Cone Spectral Sensitivities	Protanomalous Cone Spectral Sensitivities

Dichromatic Color Vision

The defect in dichromatic color vision is more severe than in anomalous trichromacy. Dichromats simply lack one type of photopigment altogether, hence they can match any color with a mixture of only two primaries. Interestingly, for dichromats also there is also a spectral light which is indistinguishable from white -- a wavelength which is called the Spectral Neutral Point. While we can't know precisely what dichromats see (any more than we can know how different colors subjectively appear to other people), we can get some idea of what dichromatic vision is like by determining which colors look the same to them.

Protanopic Cone Spectral Sensitivities	Confusion Colors (down columns)

Because they lack the long-wavelength sensitive photopigment, protanopes confuse colors which appear different to normal trichromats. Although they appear quite distinct to normal trichromats, the cyan, gray and pinkish colors in the columns of the grid above look the same to protanopes.

Deuteranopic Cone Spectral Sensitivities	Confusion Colors (down columns)

Because they lack the middle-wavelength sensitive photopigment, deuteranopes confuse colors which appear different to normal trichromats. Although they appear quite distinct to normal trichromats, the three cyan, gray and pinkish colors in the columns of the grid above look the same to deuteranopes.

Tritanopic Cone Spectral Sensitivities	Confusion Colors (down columns)

Because they lack the short-wavelength sensitive photopigment, tritanopes confuse colors which appear different to normal trichromats. Although they appear quite distinct to normal trichromats, the yellow, gray and violet colors in the columns of the grid above look the same to tritanopes.

Color Vision Tests

A number of standard tests for color vision have been developed. The Ishihara pseudoisochromatic plates are one such test. They can be used as a quick screening tool to determine whether an observer has a red-green color vision deficiency (i.e, deuteranomaly, protanomaly, deuteranopia, protanopia, or monochromacy). The Ishihara test does not screen for blue cone (tritan) defects. Below are a few of the 38 plates which comprise this test.

Ishihara Plate 2
Normals read as 8. Red-green deficient observers see 3. Monochromats see no number at all.

Ishihara Plate 15
Normals read as 7. Majority of color deficient observers see no number at all.

Ishihara Plate 24
Normals read as 35. Protanomalous and protanopic observers see 5. Deuteranomalous and deuteranopic observers see 3.

Ishihara Plate 21
Here's a fun one: The majority of observers with red-green deficiencies see the number 73. The majority of observers with normal color vision see nothing at all! This is because normals' sense of color is actually masking the subtle brightness differences which color deficient observers use to see the number.

Color masks brightness.JPG (32815 bytes)

Here is a another demonstration of how chromatic information can interfere with our perception of luminance differences, which explains why, in Ishihara plate 21 (above), normal observers cannot see the number 73 (which is defined only by subtle luminance differences), whereas color-deficient observers can "break the camouflage", so to speak, and detect the hidden numeral. On the upper row of the figure above are 5 disks, four of which have identical luminances, and one which is about 5% darker. Can you tell which it is? The lower row also contains 5 disks, four of which have identical luminances, and one which is 5% darker. Because the disks vary in color as well as luminance, however, it is much more difficult to tell which disk is darker. (The yellow disk is the darker one).

Central Processing of Color Information

The color vision pathway consists of the following: Parvocellular retinal ganglion cells (midget) ® Parvocellular layers of the thalamus (LGN) ® "Blob" regions of striate cortex (V1) ® Thin stripe and interstripe regions of V2 ® V4 ® Inferior temporal cortex (IT)

Cerebral Achromatopsia

Damage to visual area 4 (V4) sometimes results in an acquired colorblindness called cerebral achromatopsia. Here, it is important to keep in mind that the victim's color visual system is intact at the level of the retina, LGN and striate visual cortex. Apparently, the conscious awareness of color depends upon intact neural responses in this area of extrastriate cortex (V4).

Experiment With Color

By clicking the hyperlink you can download Hans Irtel's Color Vision Demonstration software. The machines in the PURC already have this software installed (search for the file CVD.exe), but you might want a copy on your own machine to fill those empty hours.

Up one level