Colour perception

Chromatic adaptation concomitantly displaces a subjective category boundary and a locus of enhanced discrimination

John Mollon and Marina Danilova

A fundamental boundary in colour space is that between reddish and greenish hues. Under neutral conditions of adaptation (to a field metameric to Illuminant D65), chromatic discrimination is optimal near this category boundary (Danilova and Mollon, 2010, Journal of Vision, 10(8): 4, 1-9). However, it has long been established that adaptation to a non-neutral field will displace the subjective boundary in chromaticity space (e.g. Wei and Shevell, 1995, Journal of the Optical Society of America A, 12, 36-46). We ask whether there will there be a concomitant shift in the locus of optimal discrimination. Using a two-alternative spatial forced choice, we measured discrimination thresholds for foveal fields subtending 2°. Brief (150 ms), bipartite targets were presented on neutral or coloured fields, and thresholds were measured along lines approximately orthogonal to the red-green boundary. In separate, interleaved runs, the subjective hue boundary was measured empirically for each condition and for each observer. The subjective hue boundary and the locus of optimal discrimination were shifted concomitantly by chromatic adaptation, a result suggesting they share a neural basis. Support: RFBR 12-04-01797-a, The Royal Society International Exchanges IE110252.

Effects of contour frequency and amplitude on the strength of the Watercolor Effect

Peggy Gerardin, Michel Dojat, Frédéric Devinck and Kenneth Knoblauch

The Watercolor Effect is a long-range, filling-in phenomenon induced by a pair of distant, wavy contours of complementary chromaticities. Here, we measured the influence of the contour frequency and amplitude and the luminance of the interior contour on the strength of the effect. Contour pairs (8 min width for each), each enclosing a circular region (4 deg diameter), were presented with two of the dimensions varying at once (luminance/frequency; luminance/amplitude; frequency/amplitude) in a conjoint measurement paradigm (Ho et al., 2009). On each trial, observers judged which of the stimuli evoked the most salient fill-in color. Control stimuli were identical except that the contours were braided and generated little filling-in. Perceptual scales were estimated by a maximum likelihood method. As shown previously using a difference scaling paradigm (Devinck & Knoblauch, 2012), the strength increases with luminance of the interior contour. The strength of the phenomenon was independent of the amplitude of the undulation but increased with the frequency up to an asymptotic level. An additive model accounted for the joint contributions of luminance and frequency. The strength of the luminance effect was comparable to that observed using difference scaling, suggesting that these distinct paradigms are mediated by a common neural process. Devinck F, Knoblauch K. (2012) J Vis 12 (3). Ho, Landy & Maloney (2009) Psych Sci. 19 (2).

Color information processing in early visual analysis

Maria Michela Del Viva, Noemi Tarallo, Daniele Benedetti, Giovanni Punzi and Steven Shevell

The visual system extracts rapidly the most important elements of the external world from a large flux of information, using early and intensive data reduction. If and how color information is used at this stage is still unknown. A recent model of early visual processing predicts - the features that human observers exploit for discrimination of compressed representations (sketches). [Del Viva et al, 2010. Journal of Vision, 10(7), 1360]; - that color does not provide a significant improvement in the information content of such sketches, compared to luminance alone [Punzi et al, 2010, Journal of Vision, 10(7), 432]. Here we measured human visual discrimination of briefly presented sketches (20 ms) containing either color or luminance information or both. Results show that performance obtained with equiluminant sketches (1-bit of color) is significantly lower (chance level) than that with dark/light grey sketches (1-bit of luminance). Adding an extra bit of color to the luminance bit does not increase performance, which is much lower than that obtained with 2-bits of luminance information. This suggests that early visual representations may not use color. Instead, color may be more suitable for a separate level of processing, following a rapid, initial luminance-based analysis.

Modeling the receptive structure of midget ganglion cells

Barry Lee and Dingcai Cao

It is postulated that midget ganglion cells near fovea receive input from a single (M or L) cone in the receptive field center while having selective or mixed cone opponent input in the surround, although physiological evidence suggests much more selectivity (Lee et al., 2012, JOSA A,29,223-232). Beyond 10 deg eccentricity, convergence of midget bipolar onto midget ganglion cells suggests cone selectivity to the center should be lost, and recent measurements (Crook et al., 2011, J. Neurosci.,31,1762-1772; Lee et al., 2012, JOSA A,29,223-232) support this view. However, ganglion cells at these eccentricities frequently show strong opponency. We report here measurements of receptive field structure using luminance, chromatic and cone isolating gratings, and attempt to model the spatial distribution of cone inputs. Near the fovea, a simple DOG model is often inadequate to describe response profiles. At higher eccentricity, although input to the center may be random, cone selective input to the surround is required to generate the opponency observed. We suggest that, although development of midget ganglion cell receptive fields may start from random wiring, mechanisms exist to generate opponent receptive fields with approximately equal cone weights, perhaps to compensate for the patchiness of cone distributions in the cone mosaic.

Colour and luminance inputs into mid- and high-level vision

Ben J. Jennings and Jasna Martinovic

In order to create a coherent representation of our visual environment multiple parallel pathways process incoming information, segmenting the image through a series of rapid hierarchically organized stages. We investigated the interdependence of activity within the luminance (L+M) and opponent chromatic (L-M and S-(L+M)) post-receptoral pathways in mid- and high-level vision. Mid-level processes extract contours and perform figure-background organisation, whereas high-level processes depend on additional semantic input such as object knowledge. We collected mid-level (contour/non-contour) and high-level (object/non-object) discrimination threshold data over a range of conditions that isolate pathways or simultaneously stimulate them. Contrast-dependent interactions between the luminance and chromatic pathways were found, with a linear relation of mid and high-level thresholds for luminance inputs. The L-M pathway drove discrimination in the presence of low luminance inputs, but an interaction between the L-M and luminance pathway was observed when the luminance input was higher. When S-cone inputs were relatively high, they slightly facilitated luminance processing. These results are consistent with previous findings on low-level interactions and demonstrate that interdependence between the geniculate pathways extends throughout the visual hierarchy.

Heterozygotes for colour vision deficiency reveal themselves in a test of spatial resolution

Marina Danilova, Chloe Chan and John Mollon

It has been proposed that female carriers of colour vision deficiency have reduced numbers of the type of cones that are lacking in their sons. A new spatial-resolution task was used to estimate the relative number of L- and M-cones in protan and deutan heterozygous carriers and in control women. Acuity was measured in the parafovea where resolution is limited by the sampling density of photoreceptors. A Landolt C target that isolated only one cone type was presented 5° to the left or right from fixation. The contrast was fixed and target diameter was varied. Relative cone numbers were estimated from the ratio of the smallest sizes recognized by M- or L-cones. An ANOVA showed a highly significant effect of phenotype (F(2)=9.47, p<0.001). In addition, there was a strong correlation (Spearman’s r=0.73, p<0.001) between these spatial estimates of L:M ratio and estimates from a temporal task using counterphase modulation photometry (the OSCAR test; Estévez et al, 1983, American Journal of Optometry and Physiological Optics, 60, 892-901). A significant correlation was also found between OSCAR test settings by carriers and by their sons (Spearman’s r=0.785, p<0.001), replicating Jordan and Mollon (Documenta Ophthalmologica Proceedings, 1997, 59, 385-392). Support: RFBR 12-04-01797-a, The Royal Society International Exchanges IE110252.

Multiple S-cone signals in BCM and ESCS patients

Caterina Ripamonti, G. Bruce Henning and Andrew Stockman

In the presence of a longer-wavelength adapting background, we have reported evidence for a delayed and inverted S-cone signal that shares an achromatic pathway with L- and M-cone signals. Without the background, the S-cone signal disappears, which suggests that it may be an indirect, inhibitory signal that acts only on extant L- and M-cone signals. Here, we measure S-cone temporal-frequency responses in observers with only S-cones (Blue-cone-monochromats or BCM), or with an excess of S-cones (enhanced-S-cone-syndrome patients, or ESCS). Our data suggest the existence of multiple S-cone signals, which we model by supposing that the S-cones interact with L- and M-cones and with each other via a network of lateral connections, and that each connection inverts and delays the S-cone signal. The converging S-cone signals destructively or constructively interfere to produce characteristic patterns in the S-cone frequency response. The data for the BCM observer are consistent with a direct S-cone signal with a substantial lateral connection also from S-cones. The data for the ESCS observers, like those for the normal, are consistent with an indirect S-cone signal acting on L- and M-cones via one or two lateral connections. We suggest that the underlying network is provided by horizontal cells.