The proximity structure of achromatic surface colors and the impossibility of asymmetric lightness matching

Abstract

When scene lighting changes in intensity, objects that appear white tend to remain white in appearance, black objects remain black, and other achromatic surfaces stay roughly the same shades of gray. Although the intensity of light reflected from an object is proportional to both the intensity of incident light and the object's reflectivity (albedo), the resulting perceptual estimates of surface appearance depend primarily on surface albedo. Explaining how the human visual system achieves this lightness constancy is a major challenge for visual science (Brainard, 2003;Gilchrist et al., 1999;Hurlbert, 1998;Maloney, 1999).The principal method used to study lightness and lightness constancy is asymmetric matching. In an asymmetric matching task, the observer is instructed to match the lightness of two objects under different illumination by, in effect, changing the albedo of one of them. If the observer consistently chooses identical albedos under different illuminants, he or she is judged to be lightness constant.Although asymmetric matching is widely employed to study both lightness and color perception, the method has a major shortcoming. When observers finally set a match, they often report that the chosen chip is not, in fact, identical in appearance to the target, as the following comment taken from an asymmetric color matching study indicates:At this match point, however, the test and the match surfaces looked different, and the observers felt as if further adjustments of the match surface should produce a better correspondence. Yet turning any of the knobs or combinations of knobs only increased the perceptual difference. (Brainard, Brunt, & Speigle, 1997, p. 2098 In a recent article, Foster (2003) highlights this problem with color matching: Alleged color matches are not always perceptually identical. Our own observers confirm the observations of Foster and of Brainard and colleagues, but in the context of lightness matching. They report that a lightness match is not, in general, achievable when two surfaces are illuminated by different illuminants. Intuitively, a white surface in shadow often does not have the same appearance as a white surface under bright illumination, and no adjustment of the albedo of either surface can produce a perceptual match. So far as we can determine, Katz (1935, pp. 79ff) was the first to report that, when observers make a match in a lightness or color constancy experiment, there is usually a residual difference.In this article, we recast the asymmetric matching paradigm in terms of the apparent dissimilarity of surfaces under different illuminants. We asked observers not to set matches between pairs of surfaces under different illumiThis work was supported by BBSRC Grant 81/S13175, NIH/NEY Grant EY08266, and HFSP Grant RG0109/1999-B. The authors thank Deborah Ross for her assistance in conducting the experiments and Alan Gilchrist for comments on an earlier draft. The data reported here are available online at www.gcal.ac.uk/sls/Vision/research/staff/Logvinenko. html. Corresp...

Lightness constancy and illumination discounting

“…That the human visual system produces two-dimensional perceptual 1 Notably, complete lightness constancy has been found only in studies in which well-articulated stimulus displays were employed (e.g., Burzlaff, 1931). In the present context, articulation means the number of surfaces with different albedos (Gilchrist & Vidal, 2002). output in response to the one-dimensional input (reflected light intensity) has been supported by a study that confirmed that achromatic colours composed a two-dimensional manifold (Logvinenko & Maloney, 2006). The achromatic colour dimensions that emerged in this study from multidimensional analysis were termed lightness and surface brightness.…”

“…3 In other words, the achromatic colour manifold cannot be specified by lightness and surface brightness as a rectangle is by its width and height. More specifically, it was found that the lightness continuum shrank when the light intensity decreased (Logvinenko & Maloney, 2006). Particularly, the subjective distance between black and white decreases under dimmer light.…”

“…Also, it has been found that luminance contrast produced by a sharp illumination edge is underestimated by a factor of 2.2, as compared to the same luminance contrast produced by a reflectance edge (Logvinenko, 2005b). Even larger underestimation of illumination was revealed by Logvinenko and Maloney (2006). They discovered that the dissimilarity between two different grey papers equally illuminated was judged to be 16 times as much as that between two identical grey papers illuminated by two different lights, despite the fact that the luminance difference between the two papers was the same in both cases (Logvinenko & Maloney, 2006).…”

“…Even larger underestimation of illumination was revealed by Logvinenko and Maloney (2006). They discovered that the dissimilarity between two different grey papers equally illuminated was judged to be 16 times as much as that between two identical grey papers illuminated by two different lights, despite the fact that the luminance difference between the two papers was the same in both cases (Logvinenko & Maloney, 2006). These results can be understood if one assumes that the lightness-versus-reflectedlight-intensity response function is steeper than the surfacebrightness-versus-reflected-light-intensity response function.…”

“…Particularly, the subjective distance between black and white decreases under dimmer light. Therefore, a more appropriate geometrical representation for the achromatic colour manifold is a circular sector, where the radii represent surface brightness, and the arcs, lightness (Logvinenko & Maloney, 2006). Thus, complete darkness is represented by the circle centre, where the lightness continuum collapses into a point.…”