Lightness constancy and illumination discounting

Abstract

Contrary to the implication of the term "lightness constancy", asymmetric lightness matching has never been found to be perfect unless the scene is highly articulated (i.e., contains a number of different reflectances). Also, lightness constancy has been found to vary for different observers, and an effect of instruction (lightness vs. brightness) has been reported. The elusiveness of lightness constancy presents a great challenge to visual science; we revisit these issues in the following experiment, which involved 44 observers in total. The stimuli consisted of a large sheet of black paper with a rectangular spotlight projected onto the lower half and 40 squares of various shades of grey printed on the upper half. The luminance ratio at the edge of the spotlight was 25, while that of the squares varied from 2 to 16. Three different instructions were given to observers: They were asked to find a square in the upper half that (i) looked as if it was made of the same paper as that on which the spotlight fell (lightness match), (ii) had the same luminance contrast as the spotlight edge (contrast match), or (iii) had the same brightness as the spotlight (brightness match). Observers made 10 matches of each of the three types. Great interindividual variability was found for all three types of matches. In particular, the individual Brunswik ratios were found to vary over a broad range (from .47 to .85). That is, lightness matches were found to be far from veridical. Contrast matches were also found to be inaccurate, being on average, underestimated by a factor of 3.4. Articulation was found to essentially affect not only lightness, but contrast and brightness matches as well. No difference was found between the lightness and luminance contrast matches. While the brightness matches significantly differed from the other matches, the difference was small. Furthermore, the brightness matches were found to be subject to the same interindividual variability and the same effect of articulation. This leads to the conclusion that inexperienced observers are unable to estimate both the brightness and the luminance contrast of the light reflected from real objects lit by real lights. None of our observers perceived illumination edges purely as illumination edges: A partial Gelb effect ("partial illumination discounting") always took place. The lightness inconstancy in our experiment resulted from this partial illumination discounting. We propose an account of our results based on the two-dimensionality of achromatic colour. We argue that large interindividual variations and the effect of articulation are caused by the large ambiguity of luminance ratios in the stimulus displays used in laboratory conditions.

A unified account of gloss and lightness perception in terms of gamut relativity

“…Gamut relativity is-to cite another example-consistent with recent psychophysical and modeling work supporting the idea that achromatic colors are specified in terms of two perceptual coordinates [61,62]. According to gamut relativity, these coordinates are whiteness and blackness, whereas alternative proposals either assert that these coordinates correspond to brightness and lightness (without specifying the mapping from luminance) [61] or leave the coordinates unspecified altogether [62]. The advantage of gamut relativity, then, is that the theory specifies perceptual coordinates in terms of a specific mapping from luminance values.…”

“…The difference is illustrated in the manner in which gamut relativity explains a wide range of phenomenology in a unified way. The theory explains, for instance, why surface regions with the same diffuse reflectance, seen under different perceived illumination levels, cannot be perceptually matched [61,62]; namely, a given surface lightness layer contains points whose whiteness coordinates differ but nonetheless can be perceptually grouped into layered representations due to their relative proximity in blacknesswhiteness space [54]. The theory also explains how observers can perceive discontinuous whitish highlight and blackish shading layers in Figs.…”

“…Due to the limitations of the heuristics embodied in statistical approaches [28][29][30][31][32][33]36], however, the log-normal model should be considered a computational expediency employed to showcase the representation of lightness and gloss in gamut relativity, rather than a realistic model of how the visual system discriminates matte and specular surface reflections. Although the model is only applicable to a limited class of images, the broader theoretical framework in which the model is caste nonetheless clarifies, unifies, and generalizes a wide body of previous experimental and theoretical work aimed at identifying how the brain parses the retinal image into layered representations of surface and illumination properties [4,7,8,[12][13][14][16][17][18][28][29][30][31][32][33][34][41][42][43][44][45][46][47][48][49][50][51][52][60][61][62][63][64].…”

The nature of instructional effects in color constancy.

“…These studies were all conducted using naive subjects and experimental methods other than adjustment-based asymmetric matching (such as achromatic adjustment or palette matching) and, predominantly, a between-subjects design (Logvinenko & Tokunaga 2011 study was within-subjects). In addition, these studies employed more naturalistic stimulus configurations than the studies reviewed above that reported large instructional effects.…”

Perceptual dimensions underlying lightness perception in homogeneous center-surround displays

“…It is often reported that lightness judgments are closer to reflectance matches whereas brightness judgments are closer to luminance matches (Arend & Goldstein, 1987;Arend & Reeves, 1986;Arend, Reeves, Schirillo, & Goldstein, 1991;see also Bauml, 1999;Cornelissen & Brenner, 1995;Troost & de Weert, 1991). However, subsequent work has shown that the size of these instructional effects (or whether there is an effect of instructions) depends on the class of stimuli (simple versus complex/ natural scenes), the task (asymmetric matching, achromatic adjustment, palette matching, or color selection task; Delahunt & Brainard, 2004;Logvinenko & Tokunaga, 2011;Madigan & Brainard, 2014;Ripamonti et al, 2004), and whether a within-or betweensubjects design was implemented (for further discussion, see Radonjić & Brainard, 2016). In our study, all stimuli were rendered under the same illumination, so reflectance and intensity matches would have been identical.…”