A new look at the Bezold–Brücke hue shift in the peripheral retina

Contrast-dependent red-green hue shift

On bright surrounds, red-green-balanced yellow targets become greenish brown with decreased target luminance, and red-green-balanced brown targets become reddish yellow with increased target luminance. These effects imply luminance- and/or contrast-dependent weighting of M- and L-cone signals in post-receptoral pathways. We show psychophysically that luminance contrast between the surround and the target is the primary determinant of the magnitude of red-green hue shift, requiring surround luminance at least twice the target luminance and increasing with further increases of surround/target contrast. There is a much smaller effect of absolute stimulus luminance, with dimmer stimuli showing slightly larger hue shifts. To evaluate a possible retinal origin of the changes in cone-signal weightings underlying the hue shift, we recorded spike responses from both ON- and OFF-center midget ganglion cells in peripheral primate retina. We found no evidence that the relative strength of L- and M-cone post-receptoral responses changed systematically with change of surround irradiance. Nor was there any systematic difference between ON- and OFF-subtypes. This suggests that the change in cone signal weighting occurs later in the visual system.

All effects of psychophysical variables on color attributes: a classification system

This paper reports the research and structuring of a classification system for the effects of psychophysical variables on the color attributes. A basic role of color science is to psychophysically specify color appearance. An early stage is to specify the effects of the psychophysical variables (as singles, pairs, etc) on the color attributes (as singles, pairs, etc), for example to model color appearance. Current data on effects are often scarce or conflicting. Few effects are well understood, and the practice of naming effects after their discoverer(s) is inadequate and can be confusing. The number and types of possible effects have never been systematically analyzed and categorized. We propose a simple and rigorous system of classification including nomenclature. The total range of effects is computed from the possible combinations of three psychophysical variables (luminance, dominant wavelength, purity) and six color attributes (lightness, brightness, hue, chroma, colorfulness, saturation) in all modes of appearance. Omitting those effects that are normally impossible to perceive at any one time (such as four- or five-dimensional colors), the total number perceivable is 161 types of effects for all modes of appearance. The type of effect is named after the psychophysical stimulus (or stimuli) and the relevant color attribute(s), e.g., Luminance-on-hue effect (traditionally known as Bezold-Brucke effect). Each type of effect may include slightly different effects with infinite variations depending on experimental parameters.

Rod hue biases produced on CRT displays

Studies of rod hue biases using monochromatic stimuli have shown that rod stimulation can shift the balance of hues at mesopic light levels. We found that the CRT display produced all three previously identified rod hue biases, which shifted the loci of all four unique hues at low mesopic light levels. Rod hue biases occurred at 2.6 cd/m(2) for some observers but not at 26 cd/m(2). At optimal light levels below 0.5 cd/m(2), rod hue biases varied among observers but generally (1) enhanced green versus red at unique yellow and sometimes at unique blue, (2) enhanced blue versus yellow at both unique green and unique red, and (3) enhanced red versus green at unique blue. Rod hue biases persisted for some observers even for smaller foveal stimuli.

Variant and invariant color perception in the near peripheral retina

Perceived shifts in hue that occur with increasing retinal eccentricity were measured by using an asymmetric color matching paradigm for a range of chromatic stimuli. Across nine observers a consistent pattern of hue shift was found; certain hues underwent large perceived shifts in appearance with increasing eccentricity, while for others little or no perceived shift was measured. In separate color naming experiments, red, blue, and yellow unique hues were found to be correlated with those hues that exhibited little or no perceptual shift with retinal eccentricity. Unique green, however, did not exhibit such a strong correlation. Hues that exhibited the largest perceptual shifts in the peripheral retina were found to correlate with intermediate hues that were equally likely to be identified by adjacent color naming mechanisms. However, once again the correlation was found to be weakest for the green mechanism. These data raise the possibility that perceptually unique hues are linked to color signals that represent the most reliable (minimally variant) chromatic information coming from the retina.

Bezold-Brücke effect in normal trichromats and protanopes

Luminance-dependent change in color appearance--the Bezold-Brücke effect--was investigated in protanopes and related to that in normal trichromats. Spectral lights were presented at six luminance levels covering mesopic, low, and high photopic vision-across three log steps from 0.76 to 760 Td. To judge color appearance, a variant of the color-naming method was used with four primary basic color terms and a "White" response. This modification enabled us to examine apparent saturation changes along with the Bezold-Brücke hue shift. Color-naming frequency functions were acquired across ten presentations of each stimulus. Since protanopes name colors idiosyncratically, changes in color appearance cannot be quantified directly from the color-naming functions. To circumvent the difficulty, these functions were transformed into color similarity measures for analysis with multidimensional scaling purported to reconstruct individual color spaces. In these, luminance-dependent shifts in color appearance were represented by means of geometric displacements. We found that for normal trichromats, shifts measured in this way agreed with those derived in our study directly, and with the hue shifts reported in earlier studies. For protanopes, contrary to some models of dichromatic vision, changes in color appearance are significant and indicate superimposed shifts in hue and saturation. The results obtained for normal trichromats, especially for protanopes, imply that nonlinearity in the yellow-blue opponent system is insufficient to explain the Bezold-Brücke effect, given the nature of the saturation shift and the demonstrated divergence between unique hues and invariant hues.

Components of bottom-up gaze allocation in natural images

Recent research [Parkhurst, D., Law, K., & Niebur, E., 2002. Modeling the role of salience in the allocation of overt visual attention. Vision Research 42 (1) (2002) 107-123] showed that a model of bottom-up visual attention can account in part for the spatial locations fixated by humans while free-viewing complex natural and artificial scenes. That study used a definition of salience based on local detectors with coarse global surround inhibition. Here, we use a similar framework to investigate the roles of several types of non-linear interactions known to exist in visual cortex, and of eccentricity-dependent processing. For each of these, we added a component to the salience model, including richer interactions among orientation-tuned units, both at spatial short range (for clutter reduction) and long range (for contour facilitation), and a detailed model of eccentricity-dependent changes in visual processing. Subjects free-viewed naturalistic and artificial images while their eye movements were recorded, and the resulting fixation locations were compared with the models' predicted salience maps. We found that the proposed interactions indeed play a significant role in the spatiotemporal deployment of attention in natural scenes; about half of the observed inter-subject variance can be explained by these different models. This suggests that attentional guidance does not depend solely on local visual features, but must also include the effects of interactions among features. As models of these interactions become more accurate in predicting behaviorally-relevant salient locations, they become useful to a range of applications in computer vision and human-machine interface design.