Figure-ground Segregation Research Articles

Contextual Modulation in Primary Visual Cortex of Macaques

Recent studies have suggested that V1 neurons extract figures from their backgrounds, in that they respond better to interior features of figures than to equivalent features of background stimuli. This is reportedly true even when the figure boundaries are distant from the borders of the classical receptive field (RF). To test the role of V1 neurons in figure-ground segregation, we recorded their responses to texture figures on texture backgrounds, centered on the RF. The texture elements of the figures remained identical across trials, and figure boundaries were defined by orientation differences between the elements in the background texture relative to elements in the figure. For nearly all neurons (98/102), responses to a large texture figure did not differ from the responses to a uniform-texture background. Although many neurons gave enhanced responses to texture boundaries, this occurred only when the boundaries were within or close to the RF borders. Similar effects were found in V2. For neurons in V1, the limited spatial extent of the contextual modulation was not increased either at low stimulus contrast or when the animal was rewarded for detecting an orientation-defined figure. Thus, V1 neurons appear to signal texture boundaries rather than figures per se. Unexpectedly, many V1 neurons gave significant long-latency responses to texture stimuli located entirely outside the classical RF, up to 5 degrees from the RF border in some cases. However, these responses did not depend on the stimulus forming a figure that contained the RF. Although V1 neurons are influenced by stimuli outside the classical RF, they do not appear to segregate figures from ground.

Illusory contours and spatial judgment.

We investigated whether, in the human visual system, the mechanisms responsible for relative location judgments are the same when those judgments are made in the context of illusory contours and in the context of mentally joining two points. We asked subjects to align a dot with the oblique contour of an illusory surface or to align a dot with two markers at an oblique orientation. The systematic errors differed in direction for these two conditions. All the systematic errors were orientation dependent. The errors in aligning a dot with an illusory contour seem to be related to the asymmetrical shape of the single objects, which are able to induce an illusory contour, as well as figure-ground segregation.

Contour decouples gamma activity across texture representation in monkey striate cortex.

Previous work on figure-ground coding in monkey V1 revealed enhanced spike rates within an object's surface representation, synchronization of gamma oscillations (gamma = 35-90 Hz) in object and background regions, but no decrease in signal correlation across the representation of a contour. The latter observation seems to contradict previous statements on the role of gamma-synchronization for scene segmentation. We re-examine these findings by analyzing different coupling measures and frequency ranges of population activities potentially contributing to figure-ground segregation. Multiple unit activity (MUA) and local field potentials (LFPs) were recorded by parallel mu-electrodes in monkey V1 during stimulation by a grating in which an object was defined by a shifted rectangle. In contradiction to the conclusions in previous work, we find strong decoupling of population activity between figure and ground representations compared to the situation in which the object is absent. In particular, coherence of late gamma-LFPs is strongly reduced, while reduction is absent during the early epochs of high-amplitude transients for LFP- and MUA-coherence at all frequencies, and at low frequencies also in the subsequent epochs. Our results of decoupling in late LFP gamma-components among figure and ground representations suggest that these signals may support figure-ground segregation.

Cortical synchronization suggests neural principles of visual feature grouping

Compositions of visual scenes are related here to neural signals in visual cortex and to cortical circuit models to understand neural mechanisms of perceptual feature grouping. Starting from the hypothesis that synchronization and decoupling of cortical γ-activities (35-90 Hz) define the relations among visual objects, we concentrate on synchronization related to (1) static retinal stimulation during ocular fixation, and (2) transient stimulation by sudden shifts in object position. The synchronization hypothesis has been tested by analyzing signal correlations in visual cortex of monkeys with the following results: Static retinal stimuli induce loosely phase-coupled γ-activities among neurons of an object's cortical representation. Patches of γ-synchronization become decoupled across the representation of an object's contour, and therby can code figure-ground segregation. Transient stimuli evoke synchronized volleys of stimulus-locked activities that are typically non-rhythmic and include low frequency components in addition to those in the γ-range. It is argued that stimulus-induced and stimulus-locked synchronizations may play different roles in perceptual feature grouping.

Correlates of figure-ground segregation in fMRI

We investigated which correlates of figure-ground-segregation can be detected by means of functional magnetic resonance imaging (fMRI). Five subjects were scanned with a Siemens Vision 1.5 T system. Motion, colour, and luminance-defined checkerboards were presented with alternating control conditions containing one of the two features of the checkerboard. We find a segregation-specific activation in V1 for all subjects and all stimuli and conclude that neural mechanisms exist as early as in the primary visual cortex that are sensitive to figure-ground segregation.

The patchwork engine: image segmentation from shape symmetries

We propose blind segmentation of images into shape-related ‘patches’ based on pre-calculated local symmetries (Van Tonder, G.J. & Ejima, Y. (1999). (Forthcoming a) Flexible computation of shape symmetries. Submitted for publication) in shape boundary contours. First, lateral weights between all points in the boundary contour map are assigned analogous to Euclidean distance maps in watershed segmentation (Beucher, S. & Lantejoul, C. (1979). Use of watersheds in contour detection. Proceedings of the International Workshop on Image Processing, CCETT, Rennes, France.). Lateral weights are then used to: (1) extract local maxima in symmetries; (2) link maxima within locally enclosed boundary contours; and (3) reconstruct shape contours using symmetry maxima as ‘seeds’. The new model overcomes weaknesses of watershed segmentation. The new model closes gaps in relatively more solid image contours, but it is fundamentally different from methods based on contour interpolation (Grossberg, S., Mingolla, E. & Todorovć, D. (1989). A neural network architecture for pre-attentive vision, IEEE Transactions on Biomedical Engineering 36, 65–84; Heitger, F. & von der Heydt, R. (1993). A computational model of neural contour processing: figure-ground segregation and illusory contours. Proceedings of the Fourth International Conference on Computer Vision, IEEE Computer Society Press, Washington D.C. (pp. 32–40)). Images are segmented into shape-relevant color-by-number-like patches which compare well to related methods (Gauch, J. & Pizer, M. (1993). The intensity axis of symmetry and its application to image segmentation, IEEE Transactions on Pattern Analysis and Machine Intelligence, 15 (8), 753–770; Ilg, W. & Ogniewicz, R. (1995). The application of Voronoi skeletons to perceptual grouping in line images, Proceedings of the 11th International Conference on Pattern Recognition, The Hague, The Netherlands, pp. 382–385; Zhu, S.C. & Yuille, A.L. (1996) FORMS: a flexible object recognition and modeling system, International Journal of Computer Vision, 20 (3), 187–212.). Two primitive operations, comparison and merging of patches, are proposed as drives for exposing more global shape contours from patches. We conclude that symmetry goes beyond abstract shape morphology: it can contribute to figure-ground segmentation in early vision and form part of primitive operations needed to create hypotheses of complex shape.

Contextual influence on orientation discrimination of humans and responses of neurons in V1 of alert monkeys.

We studied the effects of various patterns as contextual stimuli on human orientation discrimination, and on responses of neurons in V1 of alert monkeys. When a target line is presented along with various contextual stimuli (masks), human orientation discrimination is impaired. For most V1 neurons, responses elicited by a line in the receptive field (RF) center are suppressed by these contextual patterns. Orientation discrimination thresholds of human observers are elevated slightly when the target line is surrounded by orthogonal lines. For randomly oriented lines, thresholds are elevated further and even more so for lines parallel to the target. Correspondingly, responses of most V1 neurons to a line are suppressed. Although contextual lines inhibit the amplitude of orientation tuning functions of most V1 neurons, they do not systematically alter the tuning width. Elevation of human orientation discrimination thresholds decreases with increasing curvature of masking lines, so does the inhibition of V1 neuronal responses. A mask made of straight lines yields the strongest interference with human orientation discrimination and produces the strongest suppression of neuronal responses. Elevation of human orientation discrimination thresholds is highest when a mask covers only the immediate vicinity of the target line. Increasing the masking area results in less interference. On the contrary, suppression of neuronal responses in V1 increases with increasing mask size. Our data imply that contextual interference observed in human orientation discrimination is in part directly related to contextual inhibition of neuronal activity in V1. However, the finding that interference with orientation discrimination is weaker for larger masks suggests a figure-ground segregation process that is not located in V1.

Visual Form Agnosia: Neural Mechanisms and Anatomical Foundations

Abstract Visual form agnosia is a severe disorder of visual recognition resulting from extrastriate lesions in occipital and temporal cortex. Despite preserved visual acuity and minor visual field defects, patients with visual form agnosia are severely impaired in shape and form discrimination. The perception of colour, motion and stereoscopic depth is relatively unimpaired. This review presents four case reports of visual form agnosia attributable to a common aetiology (carbon monoxide poisoning) showing remarkable homogeneity. The basic deficit found in these patients is a failure to group single elements of a composite visual scene into a ‘Gestalt’, and to segregate figure from ground in static visual displays. Evidence from single-cell and lesion studies in non-human primates shows that extrastriate cortex is essential for these tasks. Correlations between anatomy and physiology further suggest that mechanisms of figure-ground segregation based on static occlusion cues are located in regions of area V2, which contribute to the form processing pathway that leads to area V4 and to the inferotemporal cortex. These findings lead to the hypothesis that visual form agnosia, as described in the present review, can be caused by selective damage to the visual form pathway.

Subjective contours, amodal completion, and transparency in animals

Perception of 'subjective contours', 'amodal completion', and 'transparency' is one of the controversial problems in human perception studies of the day. This paper attempts to outline the studies focused on perception of these phenomena in a variety of animals. Dozens of results have indicated that animals are capable of perceiving some of these phenomena except pigeons in amodal completion. It is also reported that perception of tested animals is comparable to that of humans'. Moreover, we reviewed some neurophysiological studies that examined some brain areas related to the perception. Our review of the studies suggest that the process for these perceptual phenomena is common among animals, including humans, and that process is involved in figure-ground segregation and surface perception. However, clarification of the process awaits further studies that directly compare humans and animals.

Orientation topography of layer 4 lateral networks revealed by optical imaging in cat visual cortex (area 18).

The functional specificity of corticocortical connections with respect to the topography of orientation selectivity was studied by optical imaging of intrinsic signals and bulk injections of fluorescent latex beads (green and red) and biocytin into layer 4. The distributions of retrogradely labelled cells and anterogradely labelled axon terminals were histologically reconstructed from all cortical laminae, and the resulting anatomical maps compared with the optically imaged functional maps. Layer 4 injections produced extensive horizontal labelling up to 2-3 mm from the injection centres albeit without the clear patchy pattern described after layer 2/3 injections (Gilbert & Wiesel 1989, J. Neurosci., 9, 2432-2442; Kisvárday et al. 1997, Cerebral Cortex, 7, 605-618). The functional (orientation) distribution of the labelled projections was analysed according to laminar location and lateral spread. With regard to the former, no major difference in the orientation topography between supragranular- (upper tier), granular- (middle tier) and infragranular (lower tier) layers was seen. Laterally, proximal and distal projections were distinguished and further dissected into three orientation categories, iso- (+/- 30 degrees ), oblique- (+/- 30-60 degrees ) and cross-orientations (+/- 60-90 degrees ) with respect to the orientation preference at the injection sites. The majority of distal connections (retrograde and anterograde) was equally distributed across orientations (35.4% iso-, 33.7% oblique-, and 30.9% cross-orientations) that are equivalent with a preponderance to dissimilar orientations (oblique- and cross-orientations, 64.6%). In one case, distal excitatory and inhibitory connections could be morphologically distinguished. For both categories, a marked bias to dissimilar orientations was found (excitatory, 63.7%; inhibitory, 86.6%). Taken together, these results suggest that the long-range layer 4 circuitry has a different functional role from that of the iso-orientation biased (52.9%, Kisvárday et al. 1997, Cerebral Cortex, 7, 605-618) layer 2/3 circuitry, and is perhaps involved in feature difference-based mechanisms, e.g. figure ground segregation.

Perceptual causality in the amodal completion of kinetic structures.

Perceptual causality as an amodal completion of kinetic structures was demonstrated. The experimental situation was as follows: a square appeared on the left, moved from left to right till it touched one side of a stationary rectangle (a perceptual screen), and disappeared; after a while, another square appeared at the other side of the rectangle and moved right. The first square moved faster than the second one. Sometimes, two moving objects were perceived, with an amodal launching effect: the first moving object seemed to collide with the second one behind the screen. When both the screen and the occluding interval were small, despite the discrepancy of velocities, a single moving object was perceived, but the object seemed to abruptly change its speed behind the screen. The perception of an amodal completion of movement raises some theoretical issues: the discrepancy between amodal completion of kinetic structures and that of static ones, the problem of perceptual identity, and figure-ground segregation in dynamic scene. The dilemma of the naive subject will also be discussed.

The role of convexity in perceptual completion: beyond good continuation

Since the seminal work of the Gestalt psychologists, there has been great interest in understanding what factors determine the perceptual organization of images. While the Gestaltists demonstrated the significance of grouping cues such as similarity, proximity and good continuation, it has not been well understood whether their catalog of grouping cues is complete — in part due to the paucity of effective methodologies for examining the significance of various grouping cues. We describe a novel, objective method to study perceptual grouping of planar regions separated by an occluder. We demonstrate that the stronger the grouping between two such regions, the harder it will be to resolve their relative stereoscopic depth. We use this new method to call into question many existing theories of perceptual completion (Ullman, S. (1976). Biological Cybernetics, 25, 1–6; Shashua, A., & Ullman, S. (1988). 2nd International Conference on Computer Vision (pp. 321–327); Parent, P., & Zucker, S. (1989). IEEE Transactions on Pattern Analysis and Machine Intelligence, 11, 823–839; Kellman, P. J., & Shipley, T. F. (1991). Cognitive psychology, Liveright, New York; Heitger, R., & von der Heydt, R. (1993). A computational model of neural contour processing, figure-ground segregation and illusory contours. In Internal Conference Computer Vision (pp. 32–40); Mumford, D. (1994). Algebraic geometry and its applications, Springer, New York; Williams, L. R., & Jacobs, D .W. (1997). Neural Computation, 9, 837–858) that are based on Gestalt grouping cues by demonstrating that convexity plays a strong role in perceptual completion. In some cases convexity dominates the effects of the well known Gestalt cue of good continuation. While convexity has been known to play a role in figure/ground segmentation ( Rubin, 1927; Kanizsa & Gerbino, 1976), this is the first demonstration of its importance in perceptual completion.

The temporal correlation hypothesis of visual feature integration: still alive and well.

The temporal correlation hypothesis of visual feature integration: still alive and well.

Central neural mechanisms for detecting second-order motion

Single-unit neurophysiology and human psychophysics have begun to reveal distinct neural mechanisms for processing visual stimuli defined by differences in contrast or texture (second-order motion) rather than by luminance (first-order motion). This processing begins in early visual cortical areas, with subsequent extrastriate specialization, and may provide a basis for form-cue invariant analyses of image structure, such as figure-ground segregation and detection of illusory contours.

Discrepancy detection and developmental changes in attentional engagement in infancy

The processing of discrepancies in visual arrays is fundamental to basic visual processes such as figure-ground segregation and object recognition. In six experiments, we examined this function in 3- and 5.5-month-olds. In Experiment 1, 5.5-month-olds detected a textural discrepancy induced by changes in individual color and shape features but not one induced by changes in relations among these features. These results suggest that, in infancy, as in adulthood, there are differences in the processes that detect featural discrepancies versus those that detect discrepancies in relations among features. Experiments 2, 3A, and 3B suggested that, unlike in the case of 3-month-olds in prior studies, textural and singleton discrepancies in arrays that 5.5-month-olds detect do not hold their attention in the presence of other attention-seeking cues. A comparison of the performance of 3- and 5.5-month-olds in Experiments 4A and 4B confirmed the presence of this developmental change. Altogether, these results indicate that infants’ detection of color and shape textural discrepancies is consistent with models of adult visual processing that posit a preattentive system for processing features and a resource-demanding attentional system for processing relations among features. They also suggest that the ability to disengage attention from a discrepancy and deploy it at another location develops between 3 and 5.5 months of age.

Visual segmentation by contextual influences via intra-cortical interactions in the primary visual cortex

Stimuli outside classical receptive fields have been shown to exert a significant influence over the activities of neurons in the primary visual cortex. We propose that contextual influences are used for pre-attentive visual segmentation. The difference between contextual influences near and far from region boundaries makes neural activities near region boundaries higher than elsewhere, making boundaries more salient for perceptual pop-out. The cortex thus computes global region boundaries by detecting the breakdown of homogeneity or translation invariance in the input, using local intra-cortical interactions mediated by the horizontal connections. This proposal is implemented in a biologically based model of V1, and demonstrated using examples of texture segmentation and figure-ground segregation. The model is also the first that performs texture or region segmentation in exactly the same neural circuit that solves the dual problem of the enhancement of contours, as is suggested by experimental observations. The computational framework in this model is simpler than previous approaches, making it implementable by V1 mechanisms, though higher-level visual mechanisms are needed to refine its output. However, it easily handles a class of segmentation problems that are known to be tricky. Its behaviour is compared with psycho-physical and physiological data on segmentation, contour enhancement, contextual influences and other phenomena such as asymmetry in visual search.

Figure ground segregation modulates perceived direction of ambiguous moving gratings and plaids.

A translating oriented grating viewed through a circular aperture with an occluding area in the middle appeared to move alternately in an oblique or in a vertical direction depending on the foreground/background assignment on the central occluding area. The effect occurred even when the central area was simply removed from the display, thus giving rise to a 'subjective' occluder. Parametric studies revealed that the probability of seeing oblique or vertical motion was affected by the size of the central area but not by its contrast relationships with the grating. Similar phenomena of ambiguous motion direction were observed using changes in colour along a translating grating that produced neon colour spreading effects, or using oriented edge discontinuities that collapsed into subjective plaids composed of two one-dimensional gratings. These results are discussed with respect to the hypothesis that surface segmentation mechanisms play a crucial part in the interpretation of motion signals.

Perceptual filling-in: a parametric study

We studied perceptual filling-in during maintained peripheral viewing of a uniform gray or red figure presented on a large textured background. Changes in the figure's size, shape, and eccentricity caused variations in the time required for filling-in that could be predicted from the size of its cortical projection within early visual areas. The data suggest that the time which elapsed before the figure was filled-in by its background reflects the time required for figure-ground segregation to fail, rather than a slow spread of the background across the figure. Our findings reveal interactions between surface segregation and filling-in which may be at the basis of normal surface perception.

Real-Time Tracking of Moving Objects with an Active Camera

This article is concerned with the design and implementation of a system for real-time monocular tracking of a moving object using the two degrees of freedom of a camera platform. Figure-ground segregation is based on motion without making anya prioriassumptions about the object form. Using only the first spatiotemporal image derivatives, substraction of the normal optical flow induced by camera motion yields the object image motion. Closed-loop control is achieved by combining a stationary Kalman estimator with an optimal Linear Quadratic Regulator. The implementation on a pipeline architecture enables a servo rate of 25 Hz. We study the effects of time-recursive filtering and fixed-point arithmetic in image processing and we test the performance of the control algorithm on controlled motion of objects.

Dynamic binding of visual features by neuronal/stimulus synchrony.

When people see a visual scene, certain parts of the visual scene are treated as belonging together and we regard them as a perceptual unit, which is called a "figure". People focus on figures, and the remaining parts of the scene are disregarded as "ground". In Gestalt psychology this process is called "figure-ground segregation". According to current perceptual psychology, a figure is formed by binding various visual features in a scene, and developments in neuroscience have revealed that there are many feature-encoding neurons, which respond to such features specifically. It is not known, however, how the brain binds different features of an object into a coherent visual object representation. Recently, the theory of binding by neuronal synchrony, which argues that feature binding is dynamically mediated by neuronal synchrony of feature-encoding neurons, has been proposed. This review article portrays the problem of figure-ground segregation and features binding, summarizes neurophysiological and psychophysical experiments and theory relevant to feature binding by neuronal/stimulus synchrony, and suggests possible directions for future research on this topic.