White is Green/Electronic Imaging '99 (2024)

Abstract:

Keywords:

The human visual system is most sensitive to green light.The eye/brain system whereby we perceive color works counterto this sensitivity, so we do not perceive green to bebrighter than other colors. It is impossible to see the greencomponent in white and mixed colors, and thus very difficultto conceptualize the relationship of color to brightness.

A graph of the human sensitivity to brightness has a bellshape in relation to the spectrum. Even though the curve isnot symmetrical, it has a definite and pronounced center inthe green wavelengths. This center can be clearly seen whenlight is bent through a prism. When the spectrum is onlypartially spread, white light appears in the center, withcyan gradating to blue on one side and yellow gradating tored on the other. Green is the only spectral hue not toappear in the partially spread spectrum As you spread outthe spectrum, green is the last to be separated and the lastto appear.

As you watch the spectrum spreading out you can perceive therelationship of your own sensing mechanism to the physicalstimulus. You are seeing the bell curve from the 'inside'.The prism spreads the frequencies of light on the x axis,and your visual system sees brightness on the y axis.

This relationship is usually hidden and invisible, butspreading the spectrum out in stages reveals therelationship of hue to brightness in a dramatic way.

Artists and students have no way of perceiving orunderstanding the unequal proportions of RGB in white. Thelarge amount of green light in colors that appear tan,brown, pink or even sky blue does not seem logical or evenpossible. I use the example of a green leaf, a yellow leaf,and a brown leaf placed on a scanner and made into a PICTfile in RGB.

Most mammals see in monochrome with a single receptor thatis most sensitive to green. One would expect that thisreceptor would be useful in finding green leaves to eat, butsuch is not the case.

If we look at the RGB channels of the leaf image, we seethat the yellow and brown leaves are brighter in the greenchannel than is the green leaf. If we combine the green andblue channels, we see the yellow leaf as the greenest, assomeone trained in color would expect, but even the brownleaf is more green than the green leaf!

If we look at the red channel we see that the green leaf isvery dark in that channel, and the yellow and brown leavesare bright.

It is necessary to have a red receptor signal"not red" to distinguish the green leaf from the brown leaf,and to distinguish green from white. Scientists and artistswho have worked creatively with light, understand that theyellow leaf is reproduced on the screen with red and greenlight, but few people can appreciate that even the brownleaf in RGB has as much green component as red. Thesensation of green color is not produced by the presence ofgreen light, but by the absence of red light.

Not only that,but the green channel itself must be quite a bit less than100% brightness for us to perceive a green that we wouldassociate with leaves and other green objects from nature.The green light of the RGB system appears to the eye as ayellow-green because of its intensity. Green at 50% and 100%brightness compared side by side appear to be different huesrather than different values of the same hue.

I propose that the predominance of green light in thesensation of white establishes a hue for white. "White isGreen" is a useful concept for establishing the place ofwhite in the spectrum. The hue of a color is determined byits place in the spectrum, and white is in the same place inthe spectrum as green. The chart of color temperature isusually shown as white shading to red at the low end andblue at the high end. But the color temperature of light isalso a spectrum. As the light source increases intemperature, the color shifts from red to white because theamount of green light increases. When green predominates, wesee white. At a higher temperature, blue predominates and wesee the light turning blue-white. The holy grail of aperfect white is achieved when green predominates and theamount of red and blue on either side of green are inbalance.

When color is specified in the HSV system (hue, saturationand value), the tiniest difference in hue of the RGBcomponents is enough to define the hue of the color. Forwhite, the relative brightness of RGB are all 100%, but theabsolute brightness are different. In order to producewhite and grays, green must be the predominant hue in theabsolute brightness of the RGB component colors.

Another method of determining a hue for white is to make asample square of the three component colors of RGB on thecomputer screen. Each of these samples will indicate 100%brightness in the color picker. Now set the monitors todisplay grayscale. The different brightness values of red,green and blue become dramatically apparent. In the RGBsystem the maximum values of red, green and blue areadjusted so that the combination of the three colors willproduce white. The amounts of the three component colors arenot equal. Red is 30% of white, green 59%, and blue 11% Each of the three component colors is then divided into 100percentage steps or 256 digital steps so that white isdefined as 100% of red, green and blue. The RGB systemdepends on a calibration to match human vision so that equalpercentage settings of RGB values produce pure grays andwhite. A hue composed of the percentages of RGB representedby their gray values is definitely green by the definitionof hue, even if its appearance is white.

If you stand at the back of a typical student computer lab,you will see markedly different white backgrounds on everyscreen. Each student, however, will perceive his own screento be white unless the calibration is extremely far off.This adaptation of the human eye to accept the light sourceas white makes calibration by eye extremely difficult.

The goal of color systems is to create a uniform colorspace. The hue/value relationships are adjusted by thescientist using precise measuring instruments so that theuser can deal only with hue relationships. This is veryuseful up to a point, but the usefulness breaks down whenthe hue/value relationship changes from one display deviceto another. The gradation of each of the RGB components isset so that equal numerical settings for RGB will yield apure gray. Unfortunately, this calibration will create otherhues that are not equal in brightness.

The human visual system is optimized for perceiving hues. Wecan see quite tiny hue differences, so the task of matchingcolors to the satisfaction of humans is an endless scramble.

Look a typical spectrum of colors in a 'color picker'.Hues of equal saturation are in a straight line. Imaginethe task of charting a line, solely by observation, thatrepresents 50% brightness. It is very hard for even atrained artist to do. This line moves up and down on the yaxis and is only traced with careful comparison andbracketing using a grayscale for measuring.

I have created the Percept Color Chart that displays huesalong the X axis and values on the Y axis. The Chart is onthe home page of this site.

It was presented at Color '97 in Kyoto

Each fully saturated hue is in a different place on the yaxis. On this chart equal value hues are on the same line onthe y axis. Look at the task of tracing hues of maximumsaturation on this chart. It is very easy to do because ourvisual-perceptual system is optimized for just that task.

The Percept Colorwheel

Hue in the RGB system is described in a circular arrangementwith hues designated as degrees of a circle. Unfortunatelythe traditional relationships of opposites, analogous andtriadic colors do not yield what an artist would callharmonious color because the value differences are muchgreater in the RGB system.

I have arranged the Percept Color Chart into a color wheelso that the radial axis is a gradation from black in thecenter to white at the outside. There are three steps ofhue, corresponding to the primary, secondary and tertiarycolors of the artist's color wheel. There are 10 steps ofvalue because graphic arts grayscales are done in 10% steps,and blue is close to 10%, red is at 30% and green is veryclose to 60%, so the pure hues match the steps of the scale.Each pure color has a different place on the radii from thecenter black 0% to the outside white 100%. The radialarrangement of colors was first used by Newton, and is stillused to select hues in RGB, so the wheel arrangement mightmake the Percept system easily accessible to more users.

In the traditional color wheel, yellow is the brightest purehue, so yellow is at the top of the wheel. In the PerceptColor Wheel, green (in its highest value of white) is thebrightest hue, so it is at the top of the wheel. When the colorwheel is shown on a grayscale monitor, it displays as agrayscale of concentric circles. Each ring of the circle isthe same value of gray. When the same circle is converted tograyscale using conversion algorithms, the results varyconsiderably. Blue is usually lightened, sometimes to asmuch as 25% gray. On some grayscale laser printers thecircles are a uniform gray, on others the blue is lightened.

Undoubtedly the hue/value relationship is not the onlyproblem in calibration of different devices, but it iscertainly a major problem. The end user has no way of seeingwhat is being done to this particular relationship whencalibrations are altered. I am hoping that the Percept chartand color wheel will give users a handle on this importantaspect of color display and some intuitive tools to adjustcolors to achieve their desired result.

The Percept color chart and color wheel also allow the userto select colors that are harmonious in the same way ascolors are harmonious in the artist's red yellow blue colorwheel. I believe the traditional artist's color wheel worksbecause it is adjusted to limit the value differencesbetween hues. In the RYB color wheel there is only a singlegradation of value from yellow at the top to purple at thebottom. Colors opposite each other in the horizontal planeare of almost equal value. Unfortunately, the artist's colorwheel does not work with the maximum saturation of huesavailable even in pigment form. Large numbers of possiblecolors are left out. The opposites in RYB do not correspondto either the artist's definition of opposites as colorsthat mix to produce gray, or the scientist's definition ofopposites as afterimage colors.

In the Percept system, opposites can be chosen with anydesired value relationship. The wheel is useful forselecting opposites, triads and analogous colors.

The rectangular chart is preferred for describing color aslight/dark and warm/cool. The chart seems to work better asan intuitive tool for selecting color. What I have foundteaching computer graphics is that students using thestandard color picker will use hues of maximum intensity.When students are given the Percept Color Chart withoutexplanation, they will select colors with less intensity andintuitively arrive at compositions with closer valuerelationships.