Search

Our perception of a surface's colour depends in a complex way on the colours of surrounding surfaces; this ‘colour contrast' is linked to the ‘colour constancy' which is important for object recognition, but in ways more complex than we yet understand.

Theories of image segmentation suggest that the human visual system may use two distinct processes to segregate figure from background: a local process that uses local feature contrasts to mark borders of coherent regions and a global process that groups similar features over a larger spatial scale. We performed psychophysical experiments to determine whether and to what extent the global similarity process contributes to image segmentation by motion and color. Our results show that for color, as well as for motion, segmentation occurs first by an integrative process on a coarse spatial scale, demonstrating that for both modalities the global process is faster than one based on local feature contrasts. Segmentation by motion builds up over time, whereas segmentation by color does not, indicating a fundamental difference between the modalities. Our data suggest that segmentation by motion proceeds first via a cooperative linking over space of local motion signals, generating almost immediate perceptual coherence even of physically incoherent signals. This global segmentation process occurs faster than the detection of absolute motion, providing further evidence for the existence of two motion processes with distinct dynamic properties.

Characterizing humans' ability to discriminate changes in illumination provides information about the visual system's representation of the distal stimulus. We have previously shown that humans are able to discriminate illumination changes and that sensitivity to such changes depends on their chromatic direction. Probing illumination discrimination further would be facilitated by the use of computer-graphics simulations, which would, in practice, enable a wider range of stimulus manipulations. There is no a priori guarantee, however, that results obtained with simulated scenes generalize to real illuminated scenes. To investigate this question, we measured illumination discrimination in real and simulated scenes that were well-matched in mean chromaticity and scene geometry. Illumination discrimination thresholds were essentially identical for the two stimulus types. As in our previous work, these thresholds varied with illumination change direction. We exploited the flexibility offered by the use of graphics simulations to investigate whether the differences across direction are preserved when the surfaces in the scene are varied. We show that varying the scene's surface ensemble in a manner that also changes mean scene chromaticity modulates the relative sensitivity to illumination changes along different chromatic directions. Thus, any characterization of sensitivity to changes in illumination must be defined relative to the set of surfaces in the scene.

New visual illusions provide further evidence for the influence of higher-order analyses of the visual scene on the perceived brightnesses of surfaces within that scene.

Adult colour preference has been summarized quantitatively in terms of weights on the two fundamental neural processes that underlie early colour encoding: the S-(L+M) ('blue-yellow') and L-M ('red-green') cone-opponent contrast channels (Ling, Hurlbert & Robinson, 2006; Hurlbert & Ling, 2007). Here, we investigate whether colour preference in 4-5-month-olds may be analysed in the same way. We recorded infants' eye-movements in response to pairwise presentations of eight colour stimuli varying only in hue. Infants looked longest at reddish and shortest at greenish hues. Analyses revealed that the L-M and S-(L+M) contrast between stimulus colour and background explained around half of the variation in infant preference across the hue spectrum. Unlike adult colour preference patterns, there was no evidence for sex differences in the weights on either of the cone-opponent contrast components. The findings provide a quantitative model of infant colour preference that summarizes variation in infant preference across hues.

Summary The long history of color preference studies has been described as "bewildering, confused and contradictory" [1]. Although recent studies [1–3] tend to agree on a universal preference for ‘blue', the variety and lack of control in measurement methods have made it difficult to extract a systematic, quantitative description of preference. Furthermore, despite abundant evidence for sex differences in other visual domains, and specifically in other tasks of color perception [4,5], there is no conclusive evidence for the existence of sex differences in color preference. This fact is perhaps surprising, given the prevalence and longevity of the notion that little girls differ from boys in preferring ‘pink' [6]. Here we report a robust, cross-cultural sex difference in color preference, revealed by a rapid paired-comparison task. Individual color preference patterns are summarized by weights on the two fundamental neural dimensions that underlie color coding in the human visual system. We find a consistent sex difference in these weights, which, we suggest, may be linked to the evolution of sex-specific behavioral uses of trichromacy.

Ten years ago Greenberg and colleagues presented their framework for realistic image synthesis [Greenberg et al. 1997], aiming "to develop physically based lighting models and perceptually based rendering procedures for computer graphics that will produce synthetic images that are visually and measurably indistinguishable from real-world images", paraphrasing Sutherland's 'ultimate display' [Sutherland 1965]. They specifically encouraged vision researchers to use natural, complex and three-dimensional (3D) visual displays to get a better understanding of human vision and to develop more comprehensive visual models for computer graphics that will improve the efficiency of algorithms. In this paper we follow Greenberg et al.'s directive and analyse colour and luminance gradients in a complex 3D scene. The gradients arise from changes in the light source position and orientation of surfaces. Information in image gradients could apprise the visual system about intrinsic surface reflectance properties or extrinsic illumination phenomena, including shading, shadowing and inter-reflections. Colour gradients induced by inter-reflection may play a similar role to that of luminance gradients in shape-from-shading algorithms; it has been shown that 3D shape perception modulates the influence of inter-reflections on surface colour perception [Bloj et al. 1999]. Here we report a psychophysical study in which we tested whether observers were able to discriminate between gradients due to different light source positions and found that observers reliably detected a change in the gradient information when the light source position differed by only 4 deg from the reference scene (Experiment 1). This sensitivity was mainly based on the luminance information in the gradient (Experiment 2 and 3). We conclude that for a realistic impression of a scene a global illumination algorithm should model the luminance component of inter-reflections accurately, whereas it is less critical to accurately represent the spatial variation in chromaticity.

Color constancy--by which objects tend to appear the same color under changes in illumination--is most likely achieved by several mechanisms, operating at different levels in the visual system. One powerful contributory mechanism is simultaneous spatial color contrast. Under changes in natural illumination the spatial ratios of within-type cone excitations between natural surfaces tend to be preserved (Foster and Nascimento, 1994); therefore, the neural encoding of colors as spatial contrasts tends to achieve constancy. Several factors are known to influence the strength of chromatic contrast induction between surfaces, including their relative luminance, spatial scale, spatial configuration and context (Ware and Cowan, 1982; Zaidi et al., 1991). Here we test the hypothesis that color contrast is weakened by differences between surfaces which indicate that they may be under distinct illuminants. We summarize psychophysical measurements of the effects of relative motion, relative depth and texture differences on chromatic contrast induction. Of these factors, only texture differences between surfaces weaken chromatic contrast induction. We also consider neurophysiological and neuropsychological evidence and conclude that the mechanisms which mediate local chromatic contrast effects are sited at low levels in the visual system, in primary visual cortex (V1) or below, prior to image segmentation mechanisms which require computation of relative depth or motion. V1 and lower areas may therefore play a larger role in color constancy than previously thought.

Color constancy--by which objects tend to appear the same color under changes in illumination--is most likely achieved by several mechanisms, operating at different levels in the visual system. One powerful contributory mechanism is simultaneous spatial color contrast. Under changes in natural illumination the spatial ratios of within-type cone excitations between natural surfaces tend to be preserved (Foster and Nascimento, 1994); therefore, the neural encoding of colors as spatial contrasts tends to achieve constancy. Several factors are known to influence the strength of chromatic contrast induction between surfaces, including their relative luminance, spatial scale, spatial configuration and context (Ware and Cowan, 1982; Zaidi et al., 1991). Here we test the hypothesis that color contrast is weakened by differences between surfaces which indicate that they may be under distinct illuminants. We summarize psychophysical measurements of the effects of relative motion, relative depth and texture differences on chromatic contrast induction. Of these factors, only texture differences between surfaces weaken chromatic contrast induction. We also consider neurophysiological and neuropsychological evidence and conclude that the mechanisms which mediate local chromatic contrast effects are sited at low levels in the visual system, in primary visual cortex (V1) or below, prior to image segmentation mechanisms which require computation of relative depth or motion. V1 and lower areas may therefore play a larger role in color constancy than previously thought.

New results have revealed that neurons in visual area V1 are influenced by chromatic context, in a way consistent with colour constancy. Other studies have mapped the internal cone–input structure of V1 receptive fields. Put together, these findings suggest important dual roles for V1 in colour perception.

Recent psychophysical experiments demonstrate that for simple configurations, colour appearance is largely determined by the ratios of within-type cone excitations (cone contrasts) between a target surface and its immediate background. Other experiments demonstrate that both the mean and variance of the cone excitations from remote surfaces may influence the colour of a target surface. The relative contribution of local and remote surfaces to the colour appearance of a centrally viewed target also depends on adaptational state and, therefore, on stimulus duration. Cone-contrast models of colour appearance that include the influence of cone excitations from local and global surfaces may be viewed as modern-day successors of the Retinex model for colour constancy. Here we describe psychophysical experiments of colour matching under simulated illumination changes, and examine the effects of the size and configuration of local and remote chromatic elements in a complex background on the colour appearance of a central target. We compare the observed colour matches with predictions from a standard Retinex model and from a modified Retinex-like model with weighting factors on the distance-order of chromatic edges.© (2002) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

The traditional achromatic Mach card effect is an example of lightness inconstancy and a demonstration of how shape and lightness perception interact. We present a quantitative study of this phenomenon and explore the conditions under which it occurs. The results demonstrate that observers show lightness constancy only when sufficient information is available about the light-source position, and the perceptual task required of them is surface identification rather than direct colour-appearance matching. An analysis and comparison of these results with the chromatic Mach card effect (Bloj et al 1999 Nature402 877–879) demonstrate that the luminance effects of mutual illumination do not account for the change in lightness perception in the traditional Mach card.

One proposed mechanism for underpinning colour constancy is computation of the relative activity of cones within one class – cone ratios, or cone contrasts – between surfaces in a fixed scene undergoing a change in illuminant. Although there is evidence that cone ratios do determine colour appearance under many conditions, the site or sites of their computation is unknown. Here, we report that a cerebrally achromatopsic observer, MS, displayed evidence of colour constancy in asymmetric colour matching tasks and was able to discriminate changes in cone ratios for simple, but not complex scenes. We hypothesise that the site of local cone-ratio computation is therefore early in the visual system, probably retinal.

Matt Ridley, the science writer, is married to Anya Hurlbert, reader in visual neuroscience at the University of Newcastle. She studied physics at Princeton University as an undergraduate, then went to Cambridge University as a Marshall Scholar, where she took a Part III diploma in theoretical physics and an MA in physiology. She returned to the States to do the HST MD-PhD program at Harvard-MIT, studying with Tomaso Poggio for her PhD. Matt trained as a zoologist at Oxford, and after completing his PhD on the mating habits of the pheasant, turned to journalism. They met when, as science editor for The Economist, Matt interviewed Anya, and they married while he was the magazine's Washington correspondent. They moved to England, she to Oxford for postdoctoral research, he to become American editor of The Economist. After writing as a columnist for the Daily Telegraph and other papers, Matt turned to writing books: his titles include The Red Queen, The Origins of Virtue, Genome and Nature via Nurture. They now live in rural Northumberland with their two children.

Colour constancy is typically measured with techniques involving asymmetric matching by adjustment, in which the observer views two scenes under different illuminants and adjusts the colour of a reference patch in one to match a test patch in the other. This technique involves an unnatural task, requiring the observer to predict and adjust colour appearance under an illumination shift. Natural colour constancy is more a simple matter of determining whether a colour is the same as or different from that seen under different illumination conditions. There are also technical disadvantages to the method of matching by adjustment, particularly when used to measure colour constancy in complex scenes. Therefore, we have developed and tested a two-dimensional method of constant-stimuli, forced-choice matching paradigm for measuring colour constancy. Observers view test and reference scenes haploscopically and simultaneously, each eye maintaining separate adaptation throughout a session. On each trial, a pair of test and reference patches against multicoloured backgrounds are presented, the reference patch colours being selected from a two-dimensional grid of displayable colours around the point of perfect colour constancy. The observer's task is to respond “same” or “different”. Fitting a two-dimensional Gaussian to the percentage of “different” responses yields (1) the subjective colour-constancy point, (2) the discrimination ellipse centred on this point, and (3) a map of changes in sensitivity to chromatic differences induced by the illuminant shift. The subjective colour-constancy point measured in this way shows smaller deviations from perfect colour constancy—under conditions of monocular adaptation—than previously reported; discrimination ellipses are several times larger than standard MacAdam ellipses; and chromatic sensitivity is independent of the direction of the illuminant shift, for broad distributions of background colours.

This report describes the main features of a view-based model of object recognition. The model does not attempt to account for specific cortical structures; it tries to capture general properties to be expected in a biological architecture for object recognition. The basic module is a regularization network (RBF-like; see Poggio and Girosi, 1989; Poggio, 1990) in which each of the hidden units is broadly tuned to a specific view of the object to be recognized. The network output, which may be largely view independent, is first described in terms of some simple simulations. The following refinements and details of the basic module are then discussed: (1) some of the units may represent only components of views of the object—the optimal stimulus for the unit its “center,” is effectively a complex feature; (2) the units' properties are consistent with the usual description of cortical neurons as tuned to multidimensional optimal stimuli and may be realized in terms of plausible biophysical mechanisms; (3) in learning to recognize new objects, preexisting centers may be used and modified, but also new centers may be created incrementally so as to provide maximal view invariance; (4) modules are part of a hierarchical structure—the output of a network may be used as one of the inputs to another, in this way synthesizing increasingly complex features and templates; (5) in several recognition tasks, in particular at the basic level, a single center using view-invariant features may be sufficient. Modules of this type can deal with recognition of specific objects, for instance, a specific face under various transformations such as those due to viewpoint and illumination, provided that a sufficient number of example views of the specific object are available. An architecture for 3D object recognition, however, must cope- to some extent—even when only a single model view is given. The main contribution of this report is an outline of a recognition architecture that deals with objects of a nice class undergoing a broad spectrum of transformations—due to illumination, pose, expression, and so on- by exploiting prototypical examples. A nice class of objects is a set of objects with sufficiently similar transformation properties under specific transformations, such as viewpoint transformations. For nice object classes, we discuss two possibilities: (1) class-specific transformations are to be applied to a single model image to generate additional virtual example views, thus allowing some degree of generalization beyond what a single model view could otherwise provide; (2) class-specific, view-invariant features are learned from examples of the class and used with the novel model image, without an explicit generation of virtual examples.

This paper describes the main features of a view-based model of object recognition. The model tries to capture general properties to be expected in a biological architecture for object recognition. The basic module is a regularization network in which each of the hidden units is broadly tuned to a specific view of the object to be recognized.

Publisher Summary This chapter discusses two approaches to the implementation of color constancy for vision machines, both of which demonstrate the links between neural network and classical techniques of estimation and image reconstruction. These are (1) lightness algorithms synthesized from examples using optimal linear estimation, optimal polynomial estimation and backpropagation on a two-layer artificial neural network and (2) image segmentation algorithms derived from Markov Random Field models.