In response to inquiries from various list members, Ron Bean put together the following informative list of quotations illustrating the thinking of certain leading authorities on color. He posted this in several segments to the list in 1999.
[Yule]
Principles of Color Reproduction, by J. Yule
(Wiley, 1967)
A very technical book about color printing, including a full discussion of the Neugebauer equations and the masking equations, photographic masking, and halftone dot structures. You can skip the math and still get a lot out of it, though. Also contains a brief description of early electronic scanners. Yule worked for Kodak Research Labs in Rochester.
[Billmeyer & Saltzman]
Principles of Color Technology, by Fred Billmeyer & Max Saltzman
(ISBN 0-471-03052-X, 2nd ed, 1981)
A book about specifying color in industry (for things like paint, plastics, and textiles). Heavy emphasis on metamerism. Nothing specifically about printing or photography, but worth reading anyway.
[Hunt]
The Reproduction of Color, by R. Hunt
(ISBN 0-85242-356-X, 1957, 3rd ed 1975)
Mostly about photography, but also covers television and color printing. Odd mixture of topics, but there is some good stuff here. Hunt worked for Kodak Research Labs in Harrow, England.
[Sipley]
A Half Century of Color, by Louis Sipley
(Macmillan, 1951)
A book about the history of color printing and color photography, up to 1950. Many color inserts, made from original printing plates. Before Kodachrome, color photography was very difficult and expensive, so it was more suited to printing (where the costs could be spread over many copies) than to making individual photographs to hang on a wall.
[Coote]
The Illustrated History of Color Photography, by Jack Coote
(ISBN 0-86343-380-4, 1993)
Another book about the history of color photography. Includes discussion of the various Technicolor processes, automatic snapshot-printing machines, and film-coating machines. Many historic color photographs.
Before I knew anything about printing, I thought duotones were actually 2-color separations (that is, as close to the actual color as you could get with just two inks). That's not the way it's done now, but at one time 2-color printing was considered a serious alternative to 3- and 4-color printing. The first three Technicolor processes were two-color, but they were not successful and were replaced by a three-color process in 1932.
2-color separations never caught on because they're too dependent on the content of the image. Ironically, it would be easier now with scanners and computers, but 4-color printing is so common that it's probably not worth it anymore.
I knew that Hexachrome wasn't the first 6-color printing system, but I was surprised at how common 6-color litho was up until the 1950s-- so common that Sipley felt it necessary to remind his readers (in 1951) that 4-color litho was also commonly used. I assume 6-color litho died out because color correction methods had become good enough that they didn't need the extra colors anymore.
[Coote, p75]
[describing Ives' two-color "Polychrome" system]
"From the orange-red record negative an iron-toned (blue-green) print was made using Defender Ivora, a bromide emulsion coated on a white pigmented acetate base. To this image was cemented a red/magenta positive printed from the blue-green record negative on to Kodak Wash-Off-Relief film. The gelatine relief image was dyed with a mixture of orange and magenta dyes which tended to produce yellowish highlights that favored flesh colours and shadows that remained reasonably neutral."
[Sipley, p84]
"An interest in two-color printing was manifested by photoengravers and gravure workers. In June of 1929 *The Inland Printer* printed a picture of red cherries in both three colors and in two colors. In one case the standard red, yellow, and blue were used and in the other orange and green. The originals came from *The Bridewell and Bromley Magazine*, house publication of the Grout Engraving Company of London, England. The results were amazing, and the Grout company suggested that where economy was necessary the two-color idea was far better than using three-color plates with cheaper inks on low-priced and inferior paper. The Sungravure company in England turned out some fine examples of two-color gravure work around 1930. In America the Beck Engraving Company have produced some extraordinarily fine examples of two-color gravure. In the main however most two-color printing has been to give the effect of an added color rather than a multi-color picture. The failure to utilize to the full the multi-color possibilities to be realized through the use of two colors must not be entirely charged against the photographic and photo-mechanical fields. A great deal of education is needed in commercial art schools, as much by the faculties as by the students, and in the art departments of advertising agencies, if the fullest potential of the two-color process is to be placed at the service of consumers of photography and printed reproduction."
[Hunt, p137-8]
"If all the colours of a scene could be matched by a mixture of only two dyes, colorimetrically correct two-color subtractive reproductions could be obtained. The colours in most scenes, howver, are not confined even approximately to the above type of restrictions, but two-colour (usually cyan and orange) reproductions can sometimes be surprisingly realistic, and in cinematography have been used commercially (Cornwell-Clyne, 1951, p343)."
"The acceptability of two-colour reproductions is markedly dependant on the subject matter. Indoor scenes are often very realistic, probably because light sources very deficient in blue content, such as candles and yellowish tungsten lamps, are commonly experienced, and the low level of the blue signal tends to reduce vision to nearly two variables. Outdoor scenes, on the other hand, are generally less acceptable, and the inability to render the hue difference between blue sky and green foliage is a serious drawback."
[Yule, p303-304]
"Until recently, the use of additional colors was common practice, especially in lithography. For example, a pink and a light cyan has often been used in addition to a deep red or magenta and a deep blue or cyan. This improves the purity of light blues, pinks, and flesh tones, which are usually degraded because of proportionality failure (Chapter 9) associated with the halftone structure."
"A number of special sets of four-color or five-color inks have been proposed. Zander (1924) patented a four-color system based on the use of green, magenta, blue, and yellow inks. Jacobs (1924) proposed the use of a fourth ink similar in color to Prussian Blue. Murray (1934) proposed the addition of a purple or lavender ink. With these systems, there is considerable overlapping of the absorption bands."
"Friedman (1944, p32) suggested splitting the spectrum in to four instead of three equal bands, and Ball (1950a) adjusted the bands to take advantage of what is known about color vision and color mixture. Work with Ball's system has been described by Leekley, Cox, and Jordan (1953). Ball suggested several sets of inks, each set consisting of a yellow, a pink, a purple, and a cyan. His compromise set is intended to split the spectrum at 485, 545, and 600 mu. The purple ink absorbs what he calls the 'chlor' band, a greenish yellow band extending from 545 to 600 mu. Some overlapping of absorption bands is of course unavoidable with available pigments. A masking method was used to compensate for this, but a scanner could probably do this more effectively. Ball (1950b) also suggested a five-color set which covered an even larger gamut of colors."
{Sipley, p133]
"Although the previous description has been entirely devoted to six-color photolithography, it must not be assumed that all photolithography is done by the six-color method. The use of four colors in offset printing was successfully employed by Huebner as early as 1910 (see black-and-white reproduction on page 50) and today some of the finest color printing in the country is being done by four-color photolithography."
[Sipley, p129]
"Modern color presses are built in the form of two-color and four-color units, although the latter are extremely expensive and consequently few in number. Two-color presses prove to be very flexible for four-color and six-color work whether in letterpress (although more than four colors is a rarity) or in photolithography. Large printing establishments fitted with two-color presses find it possible to run a four-color job complete at one time by two sets of plates on one press and the other two sets of plates on the second press and feeding the sheets from the first press right through the second press."
"In this kind of setup it is practical to run a four-color job through two presses, feeding from one press directly to the other."
"A battery of two-color presses similar to that shown on page 135 makes it possible to run a six-color series though in the same manner in which two two-color letterpress units handle a four-color job."
[Billmeyer & Saltzman, p106]
"Always Remember That Nobody Accepts or Rejects for Color Because of Numbers: It Is the Way It Looks That Counts."
[Sipley, p84]
[Quoting Ives, writing about his 'Polychrome' process in 1933]
"In consequence some of the most beautiful color photographs have been made by the Polychrome process by careful balancing of exposures and densities without recourse to reprinting, redyeing, or retouching. But it is also true that equally beautiful results have been obtained when miscalculated exposures and densities have necessitated such compensations; and the facility with which they can be made in the Polychrome process is a revelation to anyone who first sees such work done by an expert."
[Billmeyer & Saltzman, p166-7]
"First, if the system is going to work, the entire coloring process must be under complete control to a high degree of accuracy. Before even considering trying a system out, much less purchasing one, the prospective user should demonstrate, using instruments, that he can in fact go through the entire coloring process from incoming colorant testing to final product out the door with a reproducability, in terms of color differences, that is satisfactorily small. Believe us, this is not easy, and many prospective users of computer color matching have (or should have) stopped right there."
[Hunt, p34]
"It will be realized that these three expedients cannot correct for the fundamental limitations of the process, which spring from the nature of the colour machanism of the eye and the shape of the spectral absorption curves of the best available cyan, magenta, and yellow dyes. What is claimed for modern subtractive processes is that they produce pleasing colour pictures, and that the inevitable inaccuracies are balanced in such a way as to be least noticable."
[Billmeyer & Saltzman, p66]
"Unfortunately, there appear to be strong prejudices associated with industrial and national preferences which have prevented agreement on exactly what the 'perfect' white is, and which directions of departure from it (toward blue, yellow, red) are preferred or avoided. As a result no single formula for whiteness is now widely accepted."
[Billmeyer & Saltzman, p176-7]
"Nimeroff (1965) suggested a means of calculating a general index of metamerism from differences in the spectral curves of two samples. It is well known (for example, Wright 1969, Thornton 1978b) that these curves must cross at least three times if the samples are metameric. One gets the general feeling that the more crossings there are, and the closer the curves approximate one another, the less must be the degree of metamerism."
[Billmeyer & Saltzman, p177]
"Just as the color-rendering index rates sources used for the critical examination of colors, so one could devise a color-preference index for those used for the appreciative viewing of colors. Judd (1967) proposed such an index, in which the base from which the color differences are calculated is the set of colors people would prefer to see, rather than the actual colors of the test samples-- bluer skies, greener grass, ruddier Caucasion flesh tones, and so on."
[Billmeyer & Saltzman, p178]
"The CIE color-rendering index was designed in studies of, and works best for, sources with continuous spectral power distributions. However, a number of modern energy-efficient flourescent sources, known as prime-color lamps (Haft 1972) and sold under various trade names such as Westinghouse's Ultralume, have discontinuous distributions such as the one shown in page 174. The various indices described in this section, for both metamerism amd color rendering, do not appear to work well for sources with these unusual spectral power distributions.
[Billmeyer & Saltzman, p53]
"In the figures on page 54 we consider only the change in illuminant, for convenience, but we wish to emphasize that the change in observer to another, *all with 'normal' color vision*, is equally important (Billmeyer, 1980a), a fact all to often overlooked!"
[Billmeyer & Saltzman, p72-3]
"Differences among observers, all with normal color vision by the usual tests, can lead to at least as great a variation in the judgement of what constitutes a match as can the use of different light sources (Brown 1957, Nimeroff 1962, Smith 1963, Billmeyer 1980a, Kaiser 1980). Unless observer differences are recognized and taken accound of, for example, by prior agreement, no amount of standardization of light sources can lead to satisfactory results. Again, the importance of this often unrecognized variable cannot be overemphasized."
[Billmeyer & Saltzman, p174-5]
"What is more important, but not yet widely taken into account, is the large spread in observer characteristics within the range considered to have normal color vision. Our warning of the seriousness of this spread in the first edition of this book was largely ignored or disbelieved. We have recently documented it (Billmeyer 1980a) using a Color Rule: When observers of a wide range of ages (roughly 20-60) are considered, the spread in their settings on the Color Rule is as wide as the difference between 6500 K daylight ant "horizon sunlight" in a standard color-matching booth! If the reader has any doubts that this is an important variable in colorimetry, let him observe the difference in appearance of a metameric match using these two sources!"
"It is well known that a major part of this spread among normal observers results from the gradual yellowing of the lens of the eye with increasing age. Nardi (1980) has followed up our research to show that in a group of college-age observers (17-29 years old) the spread is about a quarter as great as that indicated above."
[Hunt, p35-36]
"In spite of the inabilty of simple trichromatic methods to render all colours colormetrically correct, it is often the exception, rather than the rule, for the result to *look* incorrect. Moreover, measurements tell us that the main defect is that colours are rendered insufficiently vivid (because of the unwanted eye-responses, see Fig. 2.3) and, in the case of subtractive reproductions, too dark (because of the unwanted dye absorptions, see Fig. 4.1); and yet the user often feels that, far from the colors being too pale and too dark, there is rather a tendency for them to appear, if anything, too vivid and too bright, and the process is accused of exaggerating the colours. In short, measurement defines she shortcomings of the process, but when the photographer exposes a colour film or the viewer looks at his tube, the defects often seem to have disappeared."
"To explain this apparent anomoly it has to be remembered that, although colours may be conveniently defined in terms of physical quantities such as spectral transmission and reflection curves, they are perceived as sensations in the mind. We must therefore consider the psychological as well as the physical side of the story."
[Hunt, p36]
"It is only on rather special occasions that the reproduction and the original are seen side by side; more usually the reproduction is seen at a different place or time, and the time interval may vary from a few hours to several weeks or even months or years. The human memory therefore plays an important part. It might be thought, then, that the process involved in appraising colours in a reproduction consists of making mental comparisons between the sensation produced in the mind by the reproduction, and a recollection from the memory of the colour sensation produced by the original object at the time when the picture was taken, It is, however, a fact that the average person generally feels competent to appraise the colours in pictures taken by people other than himself, of objects which he has never seen, at times when he was not present. This implies that colours in a reproduction are not generally appraised by comparing them either with the original object, nor even with some mental recollection thereof. By what means are then then judged?"
[Hunt, p37]
"In short, our standard of comparison, a recollection of the usual colour of green grass, is an extremely vague one, and, consequently, provided that the reproduction of the green grass is included somewhere in the range of colour sensations produced by actual samples of green grass, we are satisfied."
[Hunt, p42]
"Thus a pale red tomato, for example, is more acceptable in a reproduction than an orange or a magenta one. Correctness of hue would, therefore, seem to be more important than correctness of saturation. Moreover, the variations in saturation which occur in natural colours are generally similar to those produced by adding white light uniformly over the whole field of view. Hence, if, in a reproduction, all colours are desaturated by about the same amount, one would expect the result to look more natural than if colours of different hues and saturations were desaturated to different extents, so that some colours shone out like signal lights, while others were grossly desaturated."
"The above considerations would seem to suggest that, as far as colour is concerned, the requirements for a successful colour reproduction are, in order of importance:
(1) Correctness of hue,
(2) Correctness of tone reproduction,
(3) Approximately equal saturation of colours of all hues,
(4) Approximately proportional saturation of colours of all saturations."
[Hunt, p43]
"We can now summarize the situation. For fundamental and unavoidable reasons, simple trichromatic methods cannot result in colorimetrically correct colour reproduction, and the errors inherent in most systems are considerable when measured physically. But when the colour of an object in a colour picture is appraised by an observer, it will generally look acceptable, provided it falls somewhere within the range of colours which that object customarily exhibits in everyday life. Practically all colours met with in everyday experience are subject to wide variations in hue, lightness and saturation, and this means that no precise colour standards of familiar objects can be carried in the memory. In particular, the variations in saturation are very great and this obscures the unavoidable tendency of all processes to produce losses in saturation."
[Hunt, p165]
"The ultimate test of any colour reproduction is the opinion of the person who views it. But opinions differ, and, in cases where dissatisfaction is felt, the viewer often finds great difficulty in saying exactly why he does not like the sensations which he experiences when looking at the picture. Trained observers may feel more competent to name the faults in a reproduction, but training often makes an observer especially sensitive to certain faults which have been prevelant in his experience, while other faults, equally bad to a naive (but less articulate) observer, he may overlook. A scientific approach to the problem, though difficult, has therefore to be attempted."
[Yule, p256-7]
"These will probably be flesh, green foliage, blue sky, and a white cloud... These particular objects are not usually encountered in graphic arts problems where a picture is to be reproduced, but they are included here because of their importance in color photography."
[The above sounded very strange to me when I first read it. I think what he means is that we're not reproducing the color of the object, we're reproducing the color of a *photograph* of the object, so we're only dealing with colors that the film can reproduce. For digital cameras, replace "film" with "CCD".]
[Hunt, p44]
"In fact, some workers have reported that otimum reproduction of some well-known colours, such as skin, is achieved when a definite difference exists between the original and reproduction colours (MacAdam, 1951; Bartleson and Bray, 1962)."
[Hunt, color plate following p170]
"Good reproduction of Caucasion skin colours requires careful control of hue: although real Caucasion skin varies towards yellowish hues with sun-tan and towards more magenta hues with flushed areas, the subtle changes of hue in these directions from one area to another are important, hence the tolerances may not be large; in the green and purple directions the tolerances are usually quite small (see Section 5.6)."
[Hunt, p180]
"There is a considerable body of evidence that for Caucasian skin colour the above concepts must be supplemented to allow for the fact that a sun-tanned appearance is generally preferred to average skin colour (MacAdam, 1951; Bartleson and Bray, 1962). There may also be other colours where similar considerations apply; for instance, blue sky and blue water are usually preferred in real life to grey sky and grey water; most colour films are sensitive to ultraviolet light and hence tend to increase the blueness of sky and water relative to the saturation of the other reproduced colours, but such a tendency, if not overdone, may well be preferred to a more consistent reproduction. It may also be desirable to introduce other distortions of colour rendering to create mood or atmosphere in a picture."
[Hunt, p181]
"It is clear from Fig. 11.7 that, for reflection prints, the preferred skin colour lies, as expected, on the yellowish (sun-tanned) side of typical average real skin (Thomas, 1973), but the difference is small; and the preferred grass colour lies on the yellowish side of typical average real grass (Thomas, 1973), but again, the difference is small. The chromaticities for real skin and grass lie within the area of acceptable colour reproduction, and this suggests that, for these colours, colorimetric and preferred color reproduction are similar (although the relative luminances are rather different for grass). But, for the blue sky colour, although the dominant wavelength of the preferred and real (Hendley and Hecht, 1949) colours are closely similar, the preferred colour has an appreciably higher purity."
[Hunt, p263]
"Another useful feature often used in masking was first suggested by Yule (Yule, 1944). Exact registration of the mask when bound up with the original is obviously difficult, and if not perfectly achieved results in halos appearing around any well-defined edges. Yule suggested that the masks should be made deliberately *unsharp* by printing them with a thin spacer between the transparency and the mask material. This not only helps to obscure slight lack of registration of the masks, but also improves the reproduction of fine detail. A negative mask reduces contrast, but fine detail is seen more clearly if reproduced at high contrast; by having the mask unsharp the fine detail is not resolved by the mask and hence, when it is bound up with the original transparency, it does not reduce the contrast of fine detail, but only of large areas."
[Yule, p74]
"Color-correcting masks should be made somewhat unsharp; for example by the use of a diffusion sheet (Spiegler and Juris, 1931 and 1933; Yule, 1944). This makes the registration of the masks less critical, and it improves the sharpness of the detail in the reproduction."
[Yule, p343]
"It is often said that the yellow printer does not usually cause noticable patterns because its color is so light. Although there is some truth in this, Pollak (1958, 1959) has shown that it is more a question of overlapping absorption bands. Each wavelength of light produces moire patterns independantly, and what we see is the sum of all of these. If only one halftone print absorbed light at each wavelength, there would be no pattern. A pattern is produced because of the overlapping absorption bands of the inks. The absorption band of the black ink, which covers the whole spectrum, is bound to overlap with all three of the other inks. At wavelengths where the magenta and cyan both absorb light (mainly in the green region of the spectrum), a moire pattern will therefore be produced between the magenta, cyan, and black if the 30 or 60 degree angles are incorrect. This pattern will appear most pronounced if observed though a green filter."
[Yule, p343-4]
"When the yellow ink is at 45 degrees to the black, these two images will not form a noticeable moiré pattern. The yellow is then at an angle of 15 degrees to the other two colors, and it will normally form a small moire pattern with either one if the absorption bands overlap, and a somewhat larger second-order pattern with both of them together if all three absorption bands overlap. The yellow ink absorbs light only in the blue region of the spectrum. With a reddish magenta and a greenish cyan, both of which absorb blue light, a more noticeable pattern will be produced by interaction with the yellow ink than when bluish magenta and cyan inks are used. With a clean bluish cyan, in fact, the moiré pattern formed between the cyan and yellow may be so weak that they can both be printed at the same angle without intorducing much variation from sheet to sheet on account of register changes."
[Yule, p335]
"It might be thought that the best way to avoid moiré patterns would be to print all the halftones at exactly the same angle. This would not be satisfactory because the color produced would depend on whether the dots fell on top of each other or side by side. With the dots on top of each other, a much lighter tone would be produced, because there would be a much larger area of unprinted white paper. The color produced would therefore be extremely sensitive to minute variations in register."
[Yule, p339]
[Describing screens not quite 60 degrees apart]
"In some parts, rosettes of dots are formed with a single dot in the center (composed of three superimposed dots). Elsewhere, double rings of dots with a clear center are produced. In the latter case, the centers of the clear openings are superimposed. Between these two extremes, irregular patterns of dots are formed."
"If the screen angles in Fig. 13.07B had been exactly 60 degrees, the moire pattern would have become infinitely large. The dot arrangement would be similar over the whole sheet. It might be dot-centered or clear-centered, or it might have the irregular intermediate pattern. A shift in register of only half a screen period would shift it from one to the other. With the clear-centered pattern, there would be no point at which three dots were exactly superimposed; and with the dot-centered pattern, the centers of the clear openings would never coincide. The relative areas of the eight colored components of the pattern would change with this shift in register, and consequently the color would change slightly. This always occurs in four-color printing, especially when there is much black in the middletones, but it is usually not distinguished from other causes of color balance shifts. It becomes very noticable when an error in screen angle brings the different patterns close to one another in the form of moiré."
[Yule, p341-2]
"It is, however, if practical importance that shifting one of the screens by a distance of only half a screen period (usually about 0.004 inch) will convert the pattern from a dot-centered one to a clear-centered one, which produces a color shift if the absorption bands of all three inks overlap. The differences in color are quite small (corresponding to a difference in green-filter density of about 0.04 in the middletones for typical good quality magenta, cyan, and black inks), but they are very noticable when they occur in a moiré pattern."
[Yule, p342]
"One interesting consequence of this is that printing a black halftone over a neutral three-color gray (composed of cyan, magenta, and yellow) can make it slightly greenish or pinkish depending on the register of the dots. Another consequence is that gray-balance measurements (with yellow, magenta, and cyan inks) will be unreliable if 30 and 60 degree angles are used in the test object. In this case, the blue-filter density is the one which is variable, resulting in slight blue-yellow color shifts."
"The dot-centered and clear-centered patterns are complimentary to each other in that dot-centered highlights will be accompanied by a configuration in the shadows like the clear-centered pattern. The small openings in each of the halftones are seen as dots in the shadows, and these are arranged in a ring as in the clear-centered pattern."
"The formula for the period of the moiré pattern (Eq 13.01) shows that with a 150-line-per-inch screen, an error of half a degree will produce a pattern with a period of 0.75 inch, which is very noticable. The angles should be accurate within +/- 0.1 degree, which will give a pattern with a period of nearly 4 inches. This will not be visible except in large uniform areas. The pattern is most noticable in a medium gray with a dot area of about 50% in the cyan, magenta, and black printers."
"Even if the angles are exactly correct, this pattern is still troublesome, since a slight variation in registration of the three colors will cause a change on color from sheet to sheet (Pollak, 1955a). This must occur consistently in four-color printing, but it appears unavoidable and is usually unnoticed. It provides an additional argument against carrying black in the middletones of the picture. With a skeleton black plate, the tones containing black are so dark that the pattern is not noticable."
[Yule, p344]
"In conventional gravure printing, there is very little halftone pattern and all colors may be printed at the same angle. This has been suggested even for the halftone gravure process, such as the Dultgen process, which have a much more pronounced dot pattern. It should be remembered, however, that when all the angles are the same, the moire pattern is of infinite size, and what occurs is not a visible moire pattern but a change of color from sheet to sheet as the register changes. Large, irregular blotches of varying hue, due to local distortions, are also seen. To determine how serious this is, a proof should be made with two of the halftones at slightly incorrect angles. The intensity of the moiré pattern thus produced will be a measure of the sheet-to-sheet variation to be expected."
Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.