We study image segmentation on the basis of locally excitatory globally inhibitory oscillator networks (LEGION), whereby the phases of oscillators encode the binding of pixels. We introduce a potential for each oscillator so that only those oscillators with strong connections from their neighborhood can develop high potentials. Based on the concept of potential, a solution to remove noisy regions in an image is proposed for LEGION, so that it suppresses the oscillators corresponding to noisy regions, without affecting those corresponding to major regions. We show analytically that the resulting oscillator network separates an image into several major regions, plus a background consisting of all noisy regions, and illustrate network properties by computer simulation. The network exhibits a natural capacity in segmenting images. The oscillatory dynamics leads to a computer algorithm, which is applied successfully to segmenting real graylevel images. A number of issues regarding biological plausibility and perceptual organization are discussed. We argue that LEGION provides a novel and effective framework for image segmentation and figure-ground segregation. DeLiang Wang and David Terman Image Segmentation 1.

Advances in neuroscience implicate reentrant signaling as the predominant form of communication between brain areas. This principle was used in a series of masking experiments that defy explanation by feed-forward theories. The masking occurs when a brief display of target plus mask is continued with the mask alone. Two masking processes were found: an early process affected by physical factors such as adapting luminance and a later process affected by attentional factors such as set size. This later process is called masking by object substitution, because it occurs whenever there is a mismatch between the reentrant visual representation and the ongoing lower level activity. Iterative reentrant processing was formalized in a computational model that provides an excellent fit to the data. The model provides a more comprehensive account of all forms of visual masking than do the long-held feed-forward views based on inhibitory contour interactions. From the time a stimulus first enters the eye to the time a percept emerges into consciousness, the initial stimulus has been coded at several levels in the visual system. One of the main goals in studying visual information processing is to specify the representations at each level and the temporal sequence between

This article presents a theory that inte~ates space-based and object-based approaches to visual attention. The theory puts together M. P. van Oeffelen and P. G. Vos's ( 1982, 1983) COntour DEtector (CODE) theory of perceptual grouping by proximity with C. Bundesen's (1990) theory of visual attention (TVA). CODE provides input to TVA, accounting for spatially based between-object selection, and TVA converts the input to output, accounting for feature- and category-based withinobject selection. CODE clusters nearby items into perceptual groups that are both perceptual objects and regions of space, thereby integrating object-based and space-based approaches to attention. The combined theory provides a quantitative account of the effects of grouping by proximity and dis~nce between items on reaction time and accuracy data in 7 empirical situations that shaped the current literature on visual spatial attention. For the last decade the attention literature has been embroiled in a debate over the nature of visual spatial attention that focuses on the "thing " that attention selects (e.g., Baylis &

Five experiments explored the question of whether new perceptual units can be developed if they are diagnostic for a category learning task, and if so, what are the constraints on this unitization process? During category learning, participants were required to attend either a single component or a conjunction of five components to correctly categorize an object. Evidence consistent with unitization was found in that the conjunctive task became much easier with practice, and this improvement was not found for the single component task, or for conjunctive tasks in which the components could not be unitized. Influences of

Experiments are presented showing that visual search for Mueller-Lyer (ML) stimuli is based on complete configurations, rather than component segments. Segments easily detected in isolation were difficult to detect when embedded in a configuration, indicating preemption by low-level groups. This preemption—which caused stimulus components to become inaccessible to rapid search—was an all-ornothing effect, and so could serve as a powerful test of grouping. It is shown that these effects are unlikely to be due to blurring by simple spatial filters at early visual levels. It is proposed instead that they are due to more sophisticated processes that rapidly bind contour fragments into spatially-extended assemblies. These results support the view that rapid visual search cannot access the primitives formed at the earliest stages of visual processing; rather, it can access only higher-level, more ecologically-relevant structures. The processes that underlie human vision are often divided into two fundamentally different classes: operations that are carried out in parallel over space, and operations that are not (e.g., Neisser, 1967; von Helmholtz, 1867/1962). For the most part, parallel processes are rapid (i.e., they occur within a few hundred milliseconds), effortless, and automatic (i.e., they cannot be affected by immediate changes in higher-level goals), whereas nonparallel processes are slower, more effortful, and nonautomatic. In its current embodiment, this dichotomy divides vision into an early preattentive and a subsequent attentive stage (e.g.,

Research on the perception of texture gradients has relied heavily on the subjective reports of observers engaged in free-viewing. We asked whether these findings generalized to speeded performance. Experiment 1 showed that an important aspect of subjective perception—sizeconstancy scaling with perceived distance—also predicted the speed of pop-out visual search for cylinders viewed against a texture gradient. Experiment 2 showed that this finding could not be attributed to the local contrast between search items and the background texture. Experiment 3 assessed the relative contributions of 2 separable dimensions of texture gradients—perspective (radial spreading) and compression (foreshortening)—finding them to be independent in the more rapid search conditions (long target among shorter distractors) but combined in their influence in the slower conditions (short target among longer distractors). When observers view the texture gradient shown in Figure 1A they usually report seeing a flat surface recede into the distance, despite the fact that a two-dimensional (2-D) image alone cannot specify the three-dimensional (3-D) surface that gave rise to the projection. This study asked whether the factors influencing the perceived slant of such texture gradients also influences rapid visual search for objects placed on their surface. Although a large number of previous studies have examined the perception of slant in texture gradients (e.g., Flock,

One of the classical topics in neural networks is winner-take-all (WTA), which has been widely used in unsupervised (competitive) learning, cortical processing, and attentional control. Because of global connectivity, WTA networks, however, do not encode spatial relations in the input, and thus cannot support sensory and perceptual processing where spatial relations are important. We propose a new architecture that maintains spatial relations between input features. This selection network builds on LEGION (Locally Excitatory Globally Inhibitory Oscillator Networks) dynamics and slow inhibition. In an input scene with many objects (patterns), the network selects the largest object. This system can be easily adjusted to select several largest objects, which then alternate in time. We further show that a two-stage selection network gains efficiency by combining selection with parallel removal of noisy regions. The network is applied to select the most salient object in real images. As a s...

Traditional bottom-up models of visual processing assume that figure-ground organization precedes object recognition. This assumption seems logically necessary: How can object recognition occur before a region is labeled as figure? However, some behavioral studies find that familiar regions are more likely to be labeled figure than less familiar regions, a-problematic finding for bottom-up models. An interactive account is proposed in which figure-ground processes receive top-down input from object representations in a hierarchical system. A graded, interactive computational model is presented that accounts for behavioral results in which familiarity effects are found. The interactive model offers an alternative conception of visual processing to bottom-up models. In a typical visual scene multiple objects partially occlude one another, which makes object recognition a computation-ally complex task. Traditional information-processing theo-ries of visual perception have suggested that prior to object representation and recognition, an earlier stage of perceptual organization occurs to determine which features, locations, or surfaces most likely belong together (for examples, see

Although a number of theories and principles have been developed to guide the creation of visualizations, it is not always apparent how to apply the knowledge in these principles. We describe the application of perceptual and cognitive theories for the analysis of uncertainty visualizations. General principles from Bertin, Tufte, and Ware are outlined and then applied to the analysis of eight different uncertainty visualizations. The theories provided a useful framework for analysis of the methods, and provided insights into the strengths and weaknesses of various aspects of the visualizations.

The likelihood principle states that the visual system prefers the most likely interpretation of a stimulus, whereas the simplicity principle states that it prefers the most simple interpretation. This study investi-gates how close these seemingly very different principles are by combining findings from classical, algorithmic, and structural information theory. It is argued that, in visual perception, the two principles are perhaps very different with respect to the viewpoint-independent aspects of perception but probably very close with respect to the viewpoint-dependent aspects which, moreover, seem decisive in everyday perception. This implies that either principle may have guided the evolution of visual systems and that the simplicity paradigm may provide perception models with the necessary quantitative specifications of the often plausible but also intuitive ideas provided by the likelihood paradigm. In visual perception research, an ongoing debate concerns the question of whether the likelihood principle (Von Helmholtz, 1909/1962) or the simplicity principle (Hochberg & McAlister, 1953) provides the best explanation of the human interpretation of visual stimuli. The phenomenon to be explained is, more specifi-cally, that human subjects usually show a clear preference for only