Awhite point (often referred to asreference white ortarget white in technical documents) is a set oftristimulus values orchromaticity coordinates that serve to define the color "white" in image capture, encoding, or reproduction.[1] Depending on the application, differentdefinitions of white are needed to give acceptableresults. For example, photographs taken indoors may be lit byincandescent lights, which are relatively orange compared todaylight. Defining "white" as daylight will give unacceptable results when attempting tocolor-correct a photograph taken with incandescent lighting.
An illuminant is characterized by its relativespectral power distribution (SPD). The white point of an illuminant is thechromaticity of awhite object under the illuminant, and can be specified by chromaticity coordinates, such as thex,y coordinates on theCIE 1931 chromaticity diagram (hence the use of the relative SPD and not the absolute SPD, because the white point is only related to color and unaffected by intensity).[2]
Illuminant and white point are separate concepts. For a given illuminant, its white point is uniquely defined. A given white point, on the other hand, generally does not uniquely correspond to only one illuminant. From the commonly usedCIE 1931 chromaticity diagram, it can be seen that almost allnon-spectral colors (all except those on theline of purples), including colors described as white, can be produced byinfinitely manycombinations of spectral colors, and therefore by infinitely many different illuminant spectra.
Although there is generally no one-to-one correspondence between illuminants and white points, in the case of theCIE D-series standard illuminants, the spectral power distributions are mathematically derivable from the chromaticity coordinates of the corresponding white points.[3]
Knowing the illuminant's spectral power distribution, thereflectance spectrum of the specified white object (often taken as unity), and the numerical definition of the observer allows the coordinates of the white point in anycolor space to be defined. For example, one of the simplest illuminants is the "E" or "Equal Energy" spectrum. Its spectral power distribution is flat, giving the same power per unit wavelength at any wavelength. In terms of both the 1931 and 1964CIE XYZ color spaces, its color coordinates are [k,k,k], wherek is a constant, and its chromaticity coordinates are [x,y] = [⅓, ⅓].
If the color of an object is recorded under one illuminant, then it is possible to estimate the color of that object under another illuminant, given only the white points of the two illuminants. If the image is "uncalibrated" (the illuminant's white point unknown), the white point has to be estimated. However, if one merely wants towhite balance (make neutral objects appear neutral in the recording), this may not be necessary.
Expressing color as tristimulus coordinates in theLMS color space, one can "translate" the object's color according to theVon Kries transform simply by scaling the LMS coordinates by the ratio of the maximum of the tristimulus values at both white points. This provides a simple, but rough estimate. Another method that is sometimes preferred uses aBradford transform or anotherchromatic adaptation transform; in general, these work by transforming into an intermediate space, scaling the amounts of the primaries in that space, and converting back by the inverse transform.
To truly calculate the color of an object under another illuminant, not merely how it will be perceived, it is necessary to recordmulti-spectral orhyper-spectral color information.