The relation between the color of the illuminant and the color of the illuminated object

Abstract The eminence of Herbert Eugene Ives has been a cascading influence over color science, and particularly over activities in the CIE, throughout this century. Ives is well known for having been the first to measure the luminosity function, nine years before Guild's measurements became the basis of the first CIE standard. He also represented in explicit equation form the algebra of transforming primaries in trstimulus space and representing the transformation in projective chromaticity coordinates. [Ives was not the first to discuss such transformations, for they had been implicit in Grassmann's laws and in Maxwell's color triangle since the middle of the nineteenth century. However, he rendered the rather abstract geometry into a form readily accessible to engineers.] In addition, Ives was a pioneer in the art and science of simulating daylight byplacing filters in front of tungsten illumination sources. Finally, about thirty years ahead of his time, Ives anticipated the Fourier approach to timedependence in vision. However, a lesser‐known achievement of Ives was his contribution to the technology and assessment of color rendering by light sources. In particular, this work comprised what may be the first computational test of von Kries's theory of color constancy. This early work, which contained a startling number of the in gredients of modern work in color constancy, is reprinted as this month's Classical Article. Ives had neither good standard apparatus for measuring spectral reflectance, nor the convenience of a computer. Therefore, he invented piecewise‐constant reflectance spectra that were plausible in the sense of having values between 0.1 and 1, and three or fewer transitions across the visible spectrum (see Fig. 3). to obtain reflected‐light tristimulus values in the absence of CIE color‐matching functions, he first multiplied Koenig's sprimary‐sensation curves by two illuminant spectral power distributions (see Fig. 1) to get what Worthey called much later “object‐color matching functions.” He then summed the values over wavelength, weighted by the block reflectances, to obtain the tristimulus values. the differences in chromaticity for the same reflectance under the two different illuminants are recorded in Fig. 4 as “colorimetric shifts.” Although these colorimetric shifts are mathematically definite and useful quantities to compute, Ives realized that, in themselves, they do not represent changes in color percepts under change of illumination. Rather, the visual system undergoes an adaptation to the prevailing illumination that transforms the measured chromaticities of reflected lights to other values. This transformation implements Helmholtz's “discounting of the illuminant,” an attempt of the visual system to restore the visual representation of object reflectances to illuminant‐independent positions in the space of perceived colors. As a particular way to discount the illuminant (and hence to achieve what is known as color constancy ), Ives chose to divide each tristimulus value of a reflected light by the corresponding tristimulus value of a white (with a reflectance of 1 across the spectrum). This algorithm, an application of von Kries's model of chromatic adaptation, 6,7 certainly sends the illuminant unerringly to the point (1, 1, 1), independent of the spectrum of the illuminant. However, the algorithm does not perform of satisfactorily on other reflectance spectra. Although the colorimetric shifts are reduced by the von Kries adaptation, there are still residual shifts (see Fig. 5). This kind of analysis is quite revealing of the limitations of color constancy, especially constancy of the von Kries sort. an analysis of the same sort on measured reflectance spectra was not done until the work of Helson, et al. in 1952. Other analyses followed, stimulated by the retinex theory of Edwin Land and including the work of McCann, McKee, and Taylor and Worthey. of course, the question remained unanswered and hotly debated as to how the visual system infers the tristimulus values of the white object that it uses as the denominator in the von Kries adaptation; this has been discussed by Land, McCann, and Brainard and Wandell, among others. Another question is to define the spectral assumptions (for illuminants, reflectances, and color‐matching functions) under which the colorimetric shifts would reduce to zero for nonwhite reflectances. Much has been written in this area, including the following kinds of assumptions: (a) reflectances, color‐matching functions, and illuminant spectra that are piecewise constant over the same bands in the visible spectrum; (b) very narrowband color‐matching functions, but unconstrained illuminants and reflectances; 16,17 (c) reflectance spectra that are designed to avoid a “forbidden subspace” dictated by the (unconstrained) illuminant spectra and color‐matching functions; and finally (d) reflectance and illuminant spectra that must be linear combinations of two and three basis functions (respectively), or the reverse, and the measured color‐matching functions are linearly combined to achieve the von Kries invariance. 19,20 Other simulations have also been done with models different from that of von Kries, starting with Helson, et al., and commencing with more modern theories. 22,23,24 (See Pokorny, et al. for an excellent review of the more modern theories.) Such is the legacy of the article by Ives. The reader of Ives's article will note the claim to priority at the beginning of the second paragraph. to my best knowledge, this claim is true, and can be trusted even though Ives has not referred to any prior investigators by explicit citation (even to the article by Koenig or to the theory of von Kries). the reader will also note that the author refers to his own ongoing work in daylight simulation using filters that was to appear two years later in finished form. However, I have learned from Calvin McCamy of earlier work on daylight design that was not referenced by Ives. 27,28 Perhaps Ives viewed his article as apractical note for engineers rather than as a scholarly tract. A historical question I have not yet completely resolved is whether von Kries viewed his transformation as a vehicle to Helmholtz's discounting of the illuminant, or whether von Kries viewed his transformation merely as offering degrees of freedom that would reveal the particular linear combination of color‐matching functions that are receptor fundamentals. If the latter were true, then Ives may have been the first to use the von Kries transformation in a theory of color constancy. the Englishtranslated von Kries articles with which we are familiar do not discuss color constancy. History seems to have given only rare acknowledgment to Ives's “Color of the illuminant” article. In fact, it was only by reading the scholarly review in Helson, et al. that I became aware of the article at all. the reason for this paucity of acknowledgment can hardly be Ives's lack of eminence in his own time. Rather, it seems that his methodology in this particular article was well ahead of its time, even by the standards of his other early work that presaged the methods of the CIE. Readers who are interested in more complete accounts of the life and accomplishments of Herbert E. Ives may consult the testimonial upon his receipt of the Ives medal (named after his father, and given by the Optical Society of America), and also might find helpful the journal obituary. In particular, Ives was quite critical of Einstein's theory; he published over 30 papers about relativity starting the year he received the Ives medal, which have been reprinted with commentary. His publication list includes subjects as diverse as television, spectroscopy and Special Relativity. Much more could be written in a more generalized context about the contributions of Herbert E. Ives.