Color constancy and color rendering: Concomitant engineering of illuminants and reflectances

A strategy for cooperative illumination and reflectance design to produce color-stable environments includes reduction of metamerism and paramerism, followed by attempts at color constancy. Given a model of color constancy based on linear basis-function expansions of illuminant and reflectance spectra, all these goals could be served by designing reflectances and illuminants to inhabit certain compatible subspaces of spectral functions. Such linear models suggest two new indices of metamerism, one for designing illuminants and the other for designing reflectances. Implications are noted for designing color atlases. Finally, a physiological model of color constancy is described that is nonlinear and hence may add challenge to the concomitant design.     

Chromatic adaptation and color constancy: A possible dichotomy

Differences between chromatic adaptation and color constancy are discussed, in order to call into question the commonly held view that chromatic adaptation is the mechanism of color constancy. Whereas chromatic adaptation requires many seconds of time and occurs for simple visual scenes, color constancy asserts itself immediately and is most powerful in complex visual scenes. Furthermore, models of chromatic adaptation are not so illuminant invariant as other models of color vision. Therefore, a new operational foundation for color constancy is proposed, and existing non-adaptation models of color constancy are enumerated for future tests.     

Categorical color constancy under RGB‐LED light sources

The light‐emitting diode (LED)‐based light sources have been widely applied across numerous industries and in everyday practical uses. Recently, the LED‐based light source consisting of red, green and blue LEDs with narrow spectral bands (RGB‐LED) has been a more preferred illumination source than the common white phosphor LED and other traditional broadband light sources because the RGB‐LED can create many types of illumination color. The color rendering index of the RGB‐LED, however, is considerably lower compared to the traditional broadband light sources and the multi‐band LED light source (MB‐LED), which is composed of several LEDs and can accurately simulate daylight illuminants. Considering 3 relatively narrow spectral bands of the RGB‐LED light source, the color constancy, which is referred to as the ability of the human visual system to attenuate influences of illumination color change and hold the perception of a surface color constant, may be worse under the RGB‐LED light source than under the traditional broadband light sources or under the MB‐LED. In this study, we investigated categorical color constancy using a color naming method with real Munsell color chips under illumination changes from neutral to red, green, blue, and yellow illuminations. The neutral and 4 chromatic illuminants were produced by the RGB‐LED light source. A modified use of the color constancy index, which describes a centroid shift of each color category, was introduced to evaluate the color constancy performance. The results revealed that categorical color constancy under the 4 chromatic illuminants held relatively well, except for the red, brown, orange, and yellow color categories under the blue illumination and the orange color category under the yellow illumination. Furthermore, the categorical color constancy under red and green illuminations was better than the categorical color constancy under blue and yellow illuminations. The results indicate that a color constancy mechanism in the visual system functions in color categories when the illuminant emits an insufficient spectrum to render the colors of reflecting surfaces accurately. However, it is not recommended to use the RGB‐LED light source to produce blue and yellow illuminations because of the poor color constancy.     

The Perception of a Colored Translucent Sheet on a Background

Metelli's theory of the perception of translucency models the effective reflectance of a translucent sheet on a background as a partitive mixture of the color of the sheet and the color of the background. In the achromatic case, the usable (that is, scale-invariant) rules governing the apparent reflectances are the same as those that would emerge from the Kubelka-Munk theory. For chromatic translucency, the relationships are more complex, but a set of rules still emerge from the partitive-mixture theory that are invariant to tradeoffs between illuminant and reflectances. This column sets forth the rules in the hopes that they will have graphical applications, both on computer VDUs and in hardcopy.     

Chromatic adaptation in a 2D photograph demonstrated by a D‐up viewer

It is said that we cannot have color constancy in a photograph. The concept of recognized visual space of illumination (RVSI) asserts that chromatic adaptation occurs when one perceives the illumination that is filling a space and not the objects in the space. It predicts then that if one perceives a 3D scene in a photograph, then color constancy will occur in the photograph. In this work, a dimension‐up (D‐up) viewer was developed to perceive a 3D scene on a 2D photograph, and the effect of chromatic adaptation was measured by the color appearance of a gray patch placed at the center of the photograph. Subjects saw the patch as a vivid color when they saw a photograph that had been taken under colored illumination, which is a normal experience in a real space observation. When the color appearance was measured by the elementary color naming method, the amount of chromaticness of the patch in percentage and the apparent hue were very similar to those observed in the 2‐room technique, thus confirming the prediction by the RVSI theory.     

Improving color constancy of object colors

In a study of improving the color constancy of object colors, the spectral reflectances of the eight CIE color-rendering test samples (Munsell painted papers) were chosen as reference reflectance distributions. Many other distributions, more highly structured than those of the reference set, were synthesized by computer so as to be rendered by illuminant D₆₅ at the chromaticity at which one or another of the CIE-Munsell samples is rendered by D₆₅. The chromaticities, at which each of the synthesized reflectances is rendered by each of 30 additional illuminants, define both dominant wavelength and chroma vector for the resulting 50,000 illuminant–sample combinations. For most natural illuminants, and for present commercial lamplights, color constancy is maximized by synthesizing each sample reflectance from three relatively narrow components, 50–60 nm at half height, peaking at wavelengths near 450 nm, 530 nm, and 610 nm.     

Methods for generating spectral reflectance functions leading to color-constant properties

The Munsell color order system was rigorously defined for illuminant, observer, and surround. Using Nayatani's nonlinear model of chromatic adaptation, approximately colorconstant 1931 CIE tristimulus values for the notations of the Munsell Book of Color were calculated for a variety of continuous-spectrum illuminants between CIE A and 7600 K daylight. Several linear-programming models were devised for generating spectral reflectance functions that integrate to these tristimulus values. The most successful of these was a model based on an approximate-hue vector in tristimulus space, in which movement off and along this vector was restricted. Restrictions were also applied to the rate of change of reflectance with wavelength, following Ohta, and the model led to relatively smooth curves, comparable to those of real colorants. Indices of color constancy were devised to estimate the accuracy of the predictions. Comparisons with actual reflectance functions from physical samples revealed, in most cases, an improvement in color constancy and hue constancy.     

Surface color and color constancy

Color constancy is often treated as the tendency of surfaces to stay the same perceived color under changing illumination or context (removing/adding/replacing surrounding objects). But these types of color constancies are not basic ones and there is another kind of color constancy that is fundamental for the explanation of all color constancy phenomena. We experience it when looking at a curved uniformly colored surface or when changing the shape of the surface. A new concept of surface color is developed and the variety of all perceived colors is suggested to be described as a nine-dimensional set of 3 X 3 matrices corresponding to different surface colors. Examples of color matrices calculated for some colored surfaces being viewed by the standard viewer are presented and arguments supporting the concept are discussed. It is shown that the set of color matrices represents all perceived colors quite adequately.     

Resolving the color-image irradiance equation

The case of orthogonal projection of a smooth and piece-wise uniformly colored lambertian surface with complex color illumination is considered. The algorithm presented here is based on two equations: n·n = 1 (the norm constancy condition) and Φ n·n = 0 (the integrability condition, where Φ is a linear differential operator). Both equations hold over the whole image. Starting with the color-image irradiance equation (CHE) for the trichromatic visual system, we infer an algebraic formulae for direct computation of the normal vector field n up to some rotation U , the same one for all the points in the region of color and illuminant constancy. The transition from the image to the normal field is performed in this stage of the algorithm with a symmetric and nonnegative matrix B , which is constant in the region. This property is used for labeling segments of the color image. The second step of the algorithm computes the rotation U mentioned above. This computation is based on CHE and the integrability condition, but the way of finding a solution is quite different. An error-function dependent upon the rotation parameters is developed and some effective optimization algorithm is used for estimating the parameters. As a result, the algorithm can compute: (1) division of the image into regions of constant color and illumination; (2) the matrices U·B (in each segment) of linear transformation to recover the shape n of the surface. Results of computational experiments are given. © 1996 John Wiley & Sons, Inc.     

A FORTRAN program for predicting the effects of chromatic adaptation on color appearance based on current CIE recommendations

The CIE has recently recommended for field trial a method for predicting corresponding colors with a change in chromatic adaptation. To aid the CIE in collecting results illustrating the accuracy of these predictions, a FORTRAN program currently used by the Munsell Color Science Laboratory is listed along with test data. This chromatic-adaptation transform can be used to calculate CIELAB or CIELUV metrics for non-daylight illuminants, indices of illuminant metamerism, and indices of color constancy. Including this transform in these calculations, it is hoped, will result in metrics with better correlation to visual assessments. It is hoped that readers will implement this method, compare the effectivness of this chromatic-adaptation transform to visual evaluations, and report their results to CIE Technical Committee 1–06.     

Defining a practical method of ascertaining textile color acceptability

The relations between supplier and customer are today more important than they have ever been. However, conflicts do sometimes arise between them, deriving from differences in the judgment of color matchings. Colorimetry's role is precisely to avoid such conflicts through instrument measurements. A study was made on the pass/fail problems, based on 1,830 measurements and observations made in industrial textile firms, followed by 350 new tests. Human judgments are as liable to errors as instrument measurements, because the surface effects are often misleading for the observer. This study proposes a sorting method that combines the differences deriving from measurements by colorimetric instruments and by visual judgment. The Color Measurement Committee (CMC) equation, widely used in the textile field, has given excellent practical results. The CIE94 equation, which uses the same principle of ellipsoid tolerance, offers a mathematical simplification as well as further information on the sample observation conditions in order to determine color differences. Nevertheless, these two equations are different, and the CIE94 indexes must not be interpreted with the same tolerances as those of the CMC. Pending the CIE recommendations concerning textile samples, new acceptability tolerances should be redetermined for the CIE94. This article presents an innovative way of calculating metameric indexes that, when coupled with acceptability equations, allow the agreement rate between visual judgment and automatic selection to be increased.     

No measured effect of a familiar contextual object on color constancy

Some familiar objects have a typical color, such as the yellow of a banana. The presence of such objects in a scene is a potential cue to the scene illumination, since the light reflected from them should on average be consistent with their typical surface reflectance. Although there are many studies on how the identity of an object affects how its color is perceived, little is known about whether the presence of a familiar object in a scene helps the visual system stabilize the color appearance of other objects with respect to changes in illumination. We used a successive color matching procedure in three experiments designed to address this question. Across the experiments we studied a total of six subjects (two in Experiment 1, three in Experiment 2, and four in Experiment 3) with partial overlap of subjects between experiments. We compared measured color constancy across conditions in which a familiar object cue to the illuminant was available with conditions in which such a cue was not present. Overall, our results do not reveal a reliable improvement in color constancy with the addition of a familiar object to a scene. An analysis of the experimental power of our data suggests that if there is such an effect, it is small: less than approximately a change of 0.09 in a constancy index where an absence of constancy corresponds to an index value of 0 and perfect constancy corresponds to an index value of 1. © 2013 Wiley Periodicals, Inc. Col Res Appl, 39, 347–359, 2014     

Practical demonstration of the CIEDE2000 corrections to CIELAB using a small set of sample pairs

A set of 10 color pairs was proposed and produced in 2002 to show the advantages of the CIEDE2000 color‐difference formula with respect to CIELAB. These 10 color pairs illustrated each of the five corrections to CIELAB proposed by CIEDE2000. The 10 color pairs were visually assessed, under reference conditions close to those proposed by CIEDE2000, by two groups of 31 and 21 inexperienced observers, using two different gray scales. Average visual results in these experiments fitted CIEDE2000 predictions much better than CIELAB, as shown by a decrease of Standardized Residual Sum of Squares values of about 20 units. Current visual results showed only the improvement of CIEDE2000 upon CIELAB in predictions of perceived color differences, but they are not recommended for testing new advanced color‐difference formulas. © 2012 Wiley Periodicals, Inc. Col Res Appl, 38, 429–436, 2013.     

Physics‐based spectral sharpening through filter‐chart calibration

The spectral overlap of color‐sampling filters increases errors when using a diagonal matrix transform, for color correction and reduces color distinction. Spectral sharpening is a transformation of colors that was introduced to reduce color‐constancy errors when the colors are collected through spectrally overlapping filters. The earlier color‐constancy methods improved color precision when the illuminant color is changed, but they overlooked the color distinction. In this article, we introduce a new spectral sharpening technique that has a good compromise of color precision and distinction, based on real physical constraints. The spectral overlap is measured through observing a gray reference chart with a set of real and spectrally disjoint filters selected by the user. The new sharpening method enables to sharpen colors obtained by a sensor without knowing the camera response functions. Experiments with real images showed that the colors sharpened by the new method have good levels of color precision and distinction as well. The color‐constancy performance is compared with the data‐based sharpening method in terms of both precision and distinction. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 564–576, 2015     

Appearance variations of textile materials due to different near gray backgrounds

The preliminary experiments were carried out for investigating the effect of lightness of achromatic background (white, grays, and black) on color appearance attributes (lightness and colorfulness) of some colored fabrics. A series of achromatic backgrounds at eight levels of lightness was provided. Each test color was viewed against these backgrounds and the perceived lightness and colorfulness were evaluated by using the pair comparison method. The results indicated that the lightness of achromatic background affects perceived lightness and colorfulness. The lightness of tests colors increases when surrounded by dark backgrounds while a consistent trend was not noticed for perceived colorfulness. The results from visual assessment were compared with the predicted results by the CIECAM97s model. The predictions of the model were in agreement with the perceived lightness by visual assessment but there was not any correlation between the results of model and visual assessment for colorfulness. In the second part of the experiments, a colorimetric match of samples on different backgrounds was carried out by visual judgments as well as implementation of color appearance models in reverse mode. Results from visual trials were significantly different from those predicted by the models. © 2006 Wiley Periodicals, Inc. Col Res Appl, 31, 133–141, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.20190     

Nonvisual color perception: A critical review

Since the last century, there have been a number of claims that people can identify and discriminate color with their fingers. The most completely documented evidence for these claims, begining in 1963, is critically reviewed. It was the hope of the early investigators of this phenomenon that this ability would some day be useful for the blind. The best evidence suggests that people are capable of making certain temperature discriminations and that objects of different reflectances can be discriminated on this basis. The temperature exchange stems from the infrared energy emitted by the skin. Research that has presented evidence for finger color perception was often shown to have methodological problems, including learning to discriminate stimuli by texture cues and failure to adequately blindfold observers and prevent use of visual information. The present review concludes that the phenomenon of using the fingers to “see” color does not and will not, in the forseeable future, provide any additional hope for the blind.     

Recovering spectral data from natural scenes with an RGB digital camera and colored filters

Many spectral‐recovery methods using RGB digital cameras assume the underlying smoothness of illuminant and reflectance spectra, and apply low‐dimensional linear models. The aim of the present work was to test whether a direct‐mapping method could be used instead of a linear‐models approach to recover spectral radiances and reflectances from natural scenes with an RGB digital camera and colored filters. In computer simulations, a conventional RGB digital camera with up to three colored filters was used to image scenes drawn from a hyperspectral image database. Three measures were used to evaluate recovery with the direct‐mapping method: goodness‐of‐fit, root‐mean‐square error, and a color‐difference metric. It was found that with two and three filters both spectral radiances and reflectances could be recovered sufficiently accurately for many practical applications. With little increase in computational complexity, an RGB camera and a few colored filters can provide significantly better recovery of natural scenes than an RGB camera alone. © 2007 Wiley Periodicals, Inc. Col Res Appl, 32, 352–360, 2007     

Changes in color appearance with variations in chromatic adaptation

Chromatic adaptation has been studied by applying methods of direct scaling to color appearances of invariant stimuli seen under different conditions of adaptation. The influence on color appearance of correlated color temperature of illumination, sample luminance factor, illuminance, and surround induction was studied. Perceived hue [expressed as proportions of unitary hues] varies with color temperature of illumination but not significantly with luminance factor or illuminance for the conditions of these experiments. Colorfulness varies with color temperature and also with luminance factor and illuminance, although relative colorfulness does not change significantly with illuminance. Lightness varies with luminance factor but is essentially independent of color temperature and illuminance over the ranges investigated here. Achromatic and chromatic lightnesses for samples of equal luminance differ in systematic ways that depend upon dominant wavelength and excitation purity. Color appearance data for daylight adaptation are highly correlated with Munsell Renotation specifications. The results may be used to determine corresponding colors for the adaptation conditions studied [equivalent to CIE Illuminants D₆₅, D₅₀, A, and dark adaptation]. They may also be used to determine color appearances under those conditions throughout a color solid. It is anticipated that they will be used as the basis for developing mathematical expressions for predictions of corresponding colors under other illumination conditions as well.     

Near perfect visual compensation for atmospheric color distortions

Light is scattered and absorbed in the atmosphere producing visual effects that increase with viewing distance. Contrast and chromatic diversity decrease with distance, thereby modifying the way objects and scenes are perceived. Although some perceptual compensation to produce color constancy of individual surfaces has been reported, it is unclear to what extent unknown original images can be visually inferred from their distorted versions, that is, how much these effects can be discounted by the visual system. We investigated this issue with a paradigm akin to the paper-matching paradigm used in color constancy studies but with complex natural images. Hyperspectral data from 11 natural scenes were used to simulate their colors for distances up to 2000 m using a precise physical model of the effects of atmosphere. In each trial of the experiment, observers viewed the range of simulated images derived from one scene displayed on a calibrated monitor and selected the one perceived as more natural, without color distortion. Pooling data across scenes and observers showed that the image selected as more natural was very close to the undistorted one, corresponding to a simulated distance of only 2 m. These results suggest that observers are sensitive to changes in the naturalness of colors due to atmospheric effects and, crucially, can retrieve the original chromatic content with good accuracy revealing an efficient form of color constancy.     

Numerical methods for smoothest reflectance reconstruction

Three numerical methods are presented for finding the smoothest reflectance curve associated with a given triplet of tristimulus values. The methods differ in how “smooth” is defined, and also differ in the domain of colors over which they are applicable. The first method is very quick and applies to any tristimulus values, but sometimes can yield reflectance curves with portions that fall outside the range 0 to 1. The second method applies to colors within the spectral locus (real colors) and guarantees that the reflectances produced are positive. The third method applies to colors within the object color solid (object colors) and guarantees that the reflectances fall within the range 0 to 1. The methods are shown to create reflectances that closely resemble those of real colors (natural and synthetic). Focus is given to implementing the numerical methods in very short MATLAB/Octave functions and to understanding the numerical behavior of the methods near the limits of their respective domains of applicability in terms of matrix conditioning and discretization artifacts.