Machine learning approach to color constancy.

A number of machine learning (ML) techniques have recently been proposed to solve color constancy problem in computer vision. Neural networks (NNs) and support vector regression (SVR) in particular, have been shown to outperform many traditional color constancy algorithms. However, neither neural networks nor SVR were compared to simpler regression tools in those studies. In this article, we present results obtained with a linear technique known as ridge regression (RR) and show that it performs better than NNs, SVR, and gray world (GW) algorithm on the same dataset. We also perform uncertainty analysis for NNs, SVR, and RR using bootstrapping and show that ridge regression and SVR are more consistent than neural networks. The shorter training time and single parameter optimization of the proposed approach provides a potential scope for real time video tracking application.     

Neural mechanisms for object and color recognition]

We reported double-dissociation between the visual processing of the edges and the surfaces of objects. Patients with lateral occipital damage showed selective impairment in the perception of edges whereas those with medial ventral occipital damage showed selective impairment in the perception of the 3D structure of the surface. Patients with medial ventral occipital damage also exhibited impaired perception of color, which is also a surface property. Those results were consistent with those from neuroimaging studies. Taken together, those studies suggest that objects may be processed in two separate pathways in the ventral occipital cortex: the edges of objects are processed in the lateral pathway and the surface of objects are processed in the medial pathway. Both edges and surfaces play important roles in object recognition, and both types of perception should be evaluated in patients with visual agnosia.     

Colour constancy and conscious perception of changes of illuminant

A sudden change in illuminant (e.g., the outcome of turning on a tungsten light in a room illuminated with dim, natural daylight) causes a "global" change in perceived colour which subjects often recognise as a change of illuminant. In spite of this distinct, global change in the perceptual appearance of the scene caused by significant changes in the wavelength composition of the light reflected from different objects under the new illuminant, the perceived colour of the objects remains largely unchanged and this cornerstone property of human vision is often described as instantaneous colour constancy (ICC). ICC mechanisms are often difficult to study. The generation of appropriate stimuli to isolate ICC mechanisms remains a difficult task since the extraction of colour signals is also confounded in the processing of spatial chromatic context that leads to ICC. The extraction of differences in chromaticity that describe spatial changes in the wavelength composition of the light on the retina is a necessary operation that must precede colour constancy computations. A change of illuminant or changes in the spectral reflectance of the elements that make up the scene under a constant illuminant cause spatial changes in chromatic context and are likely to drive colour constancy mechanisms, but not exclusively. The same stimulus changes also cause differences in local luminance contrast and overall light flux changes, stimulus attributes that can activate different areas of the visual cortex. In order to address this problem we carried out a series of dichoptic experiments designed to investigate how the colour signals from the two eyes are combined in dichoptically viewed Mondrians and the extent to which the processing of chromatic context in monocularly driven neurons contributes to ICC. The psychophysical findings show that normal levels of ICC can be achieved in dichoptic experiments, even when the subject remains unaware of any changes of illuminant. Functional MRI (fMRI) experiments using new stimuli that produce stimulation of colour constancy mechanisms only in one condition with little or no difference in the activity generated in colour processing mechanisms in both test and reference conditions were also carried out. The results show that the processing of ICC signals generates strong activation in V1 and the fusiform colour area (V4, V4A). Significant activation was also observed in areas V2 and V3.     

Physiologic mechanisms underlying the antidepressant discontinuation syndrome

The rate at which serotonin reuptake inhibitor (SRI) treatment is terminated and the duration of treatment appear to be key factors in predicting discontinuation symptoms. The development of animal models to explain the mechanisms of this clinical problem has proved challenging, because less than half of all patients experience any discontinuation symptoms, many of which are subjective in nature. One explanation is that SRI discontinuation symptoms may arise from the rapid decrease in serotonin (5-HT) availability when treatment ends abruptly. Yet, it would appear that discontinuation discomforts may not be mediated exclusively through 5-HT receptors, given the major regulatory role 5-HT exerts on a number of specific chemical receptor systems in the brain. As a result, attempts to explain the determinants of this phenomenon rely on a certain level of speculation. This article examines the possible physiologic bases for the antidepressant discontinuation syndrome and briefly describes these adaptations. It discusses the 3 systems most likely to account for at least part of the symptomatology--the 5-HT, the norepinephrine, and the cholinergic systems--and the possible interactions among them. It also attempts to explain their implications in the therapeutic actions of antidepressants in patients with affective and anxiety disorders.     

Investigation of color constancy with a neural network.

The perceptual stability of an object's color under different illuminants is called color constancy. We created a neural network to investigate this phenomenon. The net consisted of one input channel for the background and one for the test object. Each channel had a set of three (L, M, and S) receptors that were transmitting to three opponent neurons. The signals from the opponent neurons were passed to hidden neurons, which were connected to the output neurons. The output signal was generated from the three components of a color vector. The neural net was trained to identify the color of Munsell samples under various illuminants using the back-propagation algorithm. Our study investigated the properties of a successfully trained neural network. Based on the cross-neuron weight analysis, we report that the successfully trained neural net calculates color differences between the test object and the background. By comparing the human visual system to the neural net, we conclude that to satisfy the color constancy phenomenon, the human visual system has to contain two separate components: one to approximate the background color and the other to estimate the color difference between the object and the background.     

Towards representation of a perceptual color manifold using associative memory for color constancy

In this paper, we propose the concept of a manifold of color perception through empirical observation that the center-surround properties of images in a perceptually similar environment define a manifold in the high dimensional space. Such a manifold representation can be learned using a novel recurrent neural network based learning algorithm. Unlike the conventional recurrent neural network model in which the memory is stored in an attractive fixed point at discrete locations in the state space, the dynamics of the proposed learning algorithm represent memory as a nonlinear line of attraction. The region of convergence around the nonlinear line is defined by the statistical characteristics of the training data. This learned manifold can then be used as a basis for color correction of the images having different color perception to the learned color perception. Experimental results show that the proposed recurrent neural network learning algorithm is capable of color balance the lighting variations in images captured in different environments successfully.     

Chromaticity of color perception and object color knowledge

Sensorimotor theories of semantic memory require overlap between conceptual and perceptual representations. One source of evidence for such overlap comes from neuroimaging reports of co-activation during memory retrieval and perception; for example, regions involved in color perception (i.e., regions that respond more to colored than grayscale stimuli) are activated by retrieval of object color. One unanswered question from these studies is whether distinctions that are observed during perception are likewise observed during memory retrieval. That is, are regions defined by a chromaticity effect in perception similarly modulated by the chromaticity of remembered objects (e.g., lemons more than coal)? Subjects performed color perception and color retrieval tasks while undergoing fMRI. We observed increased activation during both perception and memory retrieval of chromatic compared to achromatic stimuli in overlapping areas of the left lingual gyrus, but not in dorsal or anterior regions activated during color perception. These results support sensorimotor theories but suggest important distinctions within the conceptual system.     

Selective color constancy deficits after circumscribed unilateral brain lesions.

The color of an object, when part of a complex scene, is determined not only by its spectral reflectance but also by the colors of all other objects in the scene (von Helmholtz, 1886; Ives, 1912; Land, 1959). By taking global color information into account, the visual system is able to maintain constancy of the color appearance of the object, despite large variations in the light incident on the retina arising from changes in the spectral content of the illuminating light (Hurlbert, 1998; Maloney, 1999). The neural basis of this color constancy is, however, poorly understood. Although there seems to be a prominent role for retinal, cone-specific adaptation mechanisms (von Kries, 1902; P?ppel, 1986; Foster and Nascimento, 1994), the contribution of cortical mechanisms to color constancy is still unclear (Land et al., 1983; D'Zmura and Lennie, 1986). We examined the color perception of 27 patients with defined unilateral lesions mainly located in the parieto-temporo-occipital and fronto-parieto-temporal cortex. With a battery of clinical and specially designed color vision tests we tried to detect and differentiate between possible deficits in central color processing. Our results show that color constancy can be selectively impaired after circumscribed unilateral lesions in parieto-temporal cortex of the left or right hemisphere. Five of 27 patients exhibited significant deficits in a color constancy task, but all of the 5 performed well in color discrimination or higher-level visual tasks, such as the association of colors with familiar objects. These results indicate that the computations underlying color constancy are mediated by specialized cortical circuitry, which is independent of the neural substrate for color discrimination and for assigning colors to objects.     

Investigating human color preferences in the perception of complex three-dimensional structures

This article describes preliminary results of a investigation that addresses the problem of human color perception in the visual analysis of complex three-dimensional shapes. We built 3D visualization models of synthetic neural cells, and designed two experiments to identify user preferences for color when observing and executing tasks on these 3D models. Though preliminary, the results obtained from these experiments are consistent and indicate that some trends exist that deserve further investigation.     

Role of neurons of the thalamus and striopallidar complex of the human brain in ensuring the constancy of visual perception

The neuronal impulse activity during visual testing for orientation sensitivity in different head tilts was recorded in parkinsonian++ patients diagnosed and treated with long-term intracerebral electrodes implanted into different nuclei of the thalamus and striopallidar complex. It was revealed that some neuronal populations responded only to the definite orientation of the stimuli. The "preferable" orientations were not changed for the most orientation-selective neurons during left and right head positions differing by 90 degrees.     

Central neural mechanisms contributing to the perception of tactile roughness

This paper summarizes recent work showing that tactile roughness appreciation increases in a nearly linear fashion as tactile element spacing or spatial period (SP, distance centre-to-centre between raised dots in these experiments) is increased from 1.5 to 8.5 mm. Although a previous study had reported a U-shaped psychophysical function peaking at a nominal SP of 3.2 mm, differences in the surfaces (including changing SP in only one dimension as compared with two and higher dot heights that minimized contact with the smooth floor) likely contributed to the difference in the results. Roughness estimates were also unaffected by a 2-fold change in scanning speed (50 vs. 95 mm/s). Parallel recordings from neurones in primary somatosensory cortex (SI) during a texture discrimination task indicate that the discharge frequency of many SI cells shows a monotonic relation with SP (up to 5 mm tested). For some cells, the texture signals were ambiguous because discharge frequency co-varied with both texture and the scanning speed, as has also been reported for the peripheral mechanoreceptors that are activated by textured surfaces. Yet other SI cells showed a speed-invariant response to surface texture, consistent with perceptual constancy for roughness over a range of scanning speeds. We suggest that such a discharge pattern could be based on a simple intensive, or mean rate, code: an invariant central representation of surface texture could be obtained by subtracting a speed-varying signal from the ambiguous signals that co-vary with roughness and speed.     

Neuropsychological mechanisms of visual face and body perception

Human faces and bodies provide important social cues, which contribute to the identification of other people, their age and gender as well as their intentions and affective states. The underlying neuropsychological mechanisms of face processing have been studied extensively and recent interest has also focused on the study of body shape perception. The present article aims to summarize and to critically evaluate the evidence for and against the specificity of body shape processing. Cognitive mechanisms, neurocognitive models and neuronal correlates of body processing will be compared with corresponding evidence related to human face processing. Clinical phenomena related to body shape perception will also be addressed. The available data base documents a range of similarities and differences between face and body perception with respect to the cognitive mechanisms, neuronal correlates and neuropsychological impairment patterns. The lack of a selective deficit in body perception is the most important difference between both categories. The sparse data base for human body shape perception does not yet allow any firm conclusions with respect to its underlying neuropsychological mechanisms.     

Conscious perception of emotional stimuli: brain mechanisms

Emotional stimuli are thought to gain rapid and privileged access to processing resources in the brain. The structures involved in this enhanced access are thought to support subconscious, reflexive processes. Whether these pathways contribute to the phenomenological experience of emotional visual awareness (i.e., conscious perception) is unclear. In this review, it is argued that subcortical networks associated with the rapid detection of emotionally salient stimuli also play a key role in shaping awareness. This proposal is based on the idea that awareness of visual stimuli should be considered along a continuum, having intermediate levels, rather than as an all-or-none construct. It is also argued that awareness of emotional stimuli requires less input from frontoparietal structures that are often considered crucial for visual awareness. Evidence is also presented that implicates a region of the medial prefrontal cortex, involved in emotion regulation, in modulating amygdala output to determine awareness of emotional visual stimuli; when emotional stimuli are present, the conscious perception of alternative stimuli requires greater regulatory influences from cortical structures. Thus, emotional stimuli are privileged not only for neuronal representation and impact on subconscious processes, but also for awareness, allowing humans to deal flexibly rather than merely reflexively to biologically significant stimuli.     

From constancy and polyfunctionality in perception to the transcendental psychology approach: historical overview of a novel psychological paradigm

This paper presents an overview of a novel approach called Transcendental Psychology Methodology (TPM), largely inspired by Bach-y-Rita's work and developed by A.I. Mirakyan (1929-1995) and his associates. Beginning with the perceptual constancy problem, in which stimuli do not change perceptually despite variations in their physical characteristics, Mirakyan recognized that contemporary accounts of constancy included both theoretical contradictions and empirical discrepancies. This led him to propose TPM as an alternative to the traditional Product Basis Paradigm (PBP). Whilst PBP focuses upon perceptual phenomena, TPM focuses upon the underlying processes and upon the principles that support the flexibility needed to create complex, coherent representations under different stimulus conditions. The central idea that generative perceptual processes may be universal is grounded in Bach-y-Rita's famous experiments and has important parallels with present day conceptions of the construction of meaning in neurophysiological process. TPM inspired studies of a wide range of perceptual phenomena have so far suggested basic principles that can be applied to all perceptual processes, regardless of their modality. Its new conception of the perception of spatial extent can contribute to our understanding of the visual effects that Bach-y-Rita's blind subjects experience and it may provide a useful general tool for uncovering the psychological principles behind Bach-y-Rita's findings. Because of its axiomatic approach and its focus upon universal, generative processes, TPM may also be useful in other disciplines, e.g., in providing a line between neurophysiological and psychological levels of investigation. It may ultimately serve as a general theory of how sensory experience is created.