Visual Primitives Research Articles

Integrating active perception with an autonomous robot architecture

Today’s robotics applications require complex, real-time, high-bandwidth sensor systems. Although many such systems have been developed, integrating them into an autonomous agent architecture remains an area of active research. We have integrated an active stereo vision system with an autonomous agent architecture using a system of perceptual memory. Perceptual memory is an important class of memory because it is designed for the ‘behavior-based’ portion of the agent’s architecture, and not the deliberative portion. This memory maintains current and recent task-dependent perceptual information, as well as expectations about the agent’s immediate environment. Our system of perceptual memory is composed of visual primitives from our stereo system, called proximity spaces. Each proximity space represents a virtual fovea or locus of the agent’s attention. As an application, we present a robot that uses our system of perceptual memory and proximity spaces to ‘attend to’ multiple humans in a complex and unstructured indoor environment.

VISIONARY: a Viewpoint-based Visual Language for Querying Relational Databases

The adoption of a visual interface can simplify the query formulation process in DBMSs by enabling naive users to interact with a friendly environment. In this work, we propose a visual query language based on a diagrammatic paradigm, used for both data and query representation. The external data model is called vision and is made up of the visual primitives of concept and association. The external query model is based on the definition of a viewpoint, which is a perspective for accessing data defined dynamically by selecting a concept of primary interest. Internally, the data model is relational and the query language is SQL. An intermediate graph-based model ensures consistent mapping between the visual and the relational worlds. Our language has been implemented within a tool which can be mounted on top of any relational DBMS supporting ODBC. The system has been tested with naive users; the results of the experiment are reported and compared with those obtained with other visual languages.

Eliciting visual primitives for detection of elongated shapes

This paper deals with the problem of eliciting visual primitives for visual search with the aim of detecting 2D objects characterized, primarily, by an elongated shape. The paper proposes a new visual primitive obtained by combining in a suitable correlation, a basic set of standard local features. This primitive is able to synthesize the information associated with local features and, as a more effective ensemble of proprieties of the considered model, enhance detection. The paper discusses the approach, presents the new primitive and evaluates its robustness in the case of non-ideal and noisy images. Finally an application to the context of visual inspection is presented.

Segmentation of Video by Clustering and Graph Analysis

Many video programs have story structures that can be recognized through the clustering of video contents based on low-level visual primitives and the analysis of high-level structures imposed by temporal arrangement of composing elements. In this paper we propose techniques and formulations to match and cluster video shots of similar visual contents, taking into account the visual characteristics and temporal dynamics of video. In addition, we extend theScene Transition Graphrepresentation for the analysis of temporal structures extracted from video. The analyses lead to automatic segmentation of scenes and story units that cannot be achieved with existing shot boundary detection schemes and the building of a compact representation of video contents. Furthermore, the segmentation can be performed on a much reduced data set extracted from compressed video and works well on a wide variety of video programming types. Hence, we are able to decompose video into meaningful hierarchies and compact representations that reflect the flow of the story. This offers a mean for the efficient browsing and organization of video.

DEVise (demo abstract)

DEVise is a data exploration system that allows users to easily develop, browse, and share visual presentations of large tabular datasets (possibly containing or referencing multimedia objects) from several sources. The DEVise framework, implemented in a tool that has been already successfully applied to a variety of real applications by a number of user groups, makes several contributions. In particular, it combines support for extended relational queries with powerful data visualization features. Datasets much larger than available main memory can be handled—DEVise is currently being used to visualize datasets well in excess of 100MB—and data can be interactively examined at several levels of detail: all the way from meta-data summarizing the entire dataset, to large subsets of the actual data, to individual data records. Combining querying (in general, data processing) with visualizations gives us a very versatile tool, and presents several novel challenges. Our emphasis is on developing an intuitive yet powerful set of querying and visualization primitives that can be easily combined to develop a rich set of visual presentations that integrate data from a wide range of application domains. In this demo, we will present a number of examples of the use of the DEVise tool for visualizing and interactively exploring very large datasets, and report on our experience in applying it to several real applications.

DEVise

DEVise is a data exploration system that allows users to easily develop, browse, and share visual presentation of large tabular datasets (possibly containing or referencing multimedia objects) from several sources. The DEVise framework is being implemented in a tool that has been already successfully applied to a variety of real applications by a number of user groups. Our emphasis is on developing an intuitive yet powerful set of querying and visualization primitives that can be easily combined to develop a rich set of visual presentations that integrate data from a wide range of application domains. While DEVise is a powerful visualization tool, its greatest strengths are the ability to interactively explore a visual presentation of the data at any level of detail (including retrieving individual data records), and the ability to seamlessly query and combine data from a variety of local and remote sources. In this paper, we present the DEVise framework, describe the current tool, and report on our experience in applying it to several real applications.

Information appliances and tools in Visage

A workspace for information analysis must support two types of analysis: routine information reporting and exploratory analysis. We have been developing a workspace called Visage to address both types of analysis tasks. Our goal is for expert users to move easily between routine tasks and exploratory data analysis as situations demand. Furthermore, recognizing the inherent complexity of data analysis, we want a system that is accessible to occasional users yet simultaneously provides power and flexibility for expert users. Ideally, such a system should support a gradual shift to increasing expertise so that occasional users of routine interfaces can perform more detailed analyses if needed. Visage addresses these needs by providing a consistent interface to a wide variety of data manipulation and visualization primitives. Plus, it lets users combine these operations to create appliances-special-purpose applications capturing multiple steps of a routine data analysis process-in a single, simple interface. Analysts thus benefit from a common set of operations and ways to integrate information across applications. Visage provides both basic tools and specialized information appliances to help them.

Gradients as visual primitives.

Following J. J. Gibson (1950), it is implicitly assumed in the literature that texture gradients are directly available as perceptual primitives. Yet, the depth response to compression gradients is poor compared with gradients of linear perspective. This may indicate that mechanisms for directly detecting the differential structure that constitutes a compression gradient do not exist. We tested this hypothesis outside the context of depth perception by measuring the speed with which participants could detect a gradient anomaly as a function of the number of elements in the gradient. Only in the case of linear perspective did anomalies "pop out." This was attributable to the emergent feature of alignment of the ends of the elements forming the gradient rather than the direct detection of its differential structure. It is argued that gradients are not perceptual primitives and that the poor depth response to compression in a variety of context (motion parallax, stereo, and perspective) therefore is not surprising.

Gradients as visual primitives.

Gradients as visual primitives.

Segregation of Motion Structure from Random Visual Noise

A potential source of information about spatial layout, surface slant, and self-motion is provided by transformations of the optic flow field. Theoretical analyses have demonstrated that such flow fields can be decomposed into translation, rotation, expansion, and deformation components. The objective in this study was to investigate the effectiveness of these motions as visual primitives for preattentive scene segregation. The stimuli consisted of two-frame patterns containing 998 dots. A variable number of these dots (5 to 50) were defined as target dots, representing a specific motion structure, with the residual dots defined as a noise field. The observers' task was to identify the location and type of motion structure. On the basis of the number of target dots needed to detect and correctly identify the target structure, it was determined that translation was the most easily detected structure, followed by rotation, expansion, and compression. Deformation was found to have a detection threshold as high as compression. The results are related to other studies of these motion components.

Théorie des GEONS et interprétation quantitative dans une tâche d'identification d'objets

Summary : Recognition-by-component theory and quantitative interpretation in visual object recognition. In Biederman's Recognition-by-components theory, the objet recognition process involves an internal componential and structural description. Components corresponding to geometrical non-accidental properties of objects represent visual primitives. In a first experiment, ten drawn objects were presented. Object contours were deleted at vertex or mid segments but we applied an algorithm to equalize the information quantity presented for two of the three conditions. Results indicate that the identification rate was partially predicted by the amount of information calculated by energy and three dimensional functions. In a second experiment, symmetrical objects were presented with half of the contour or with the top part of the contour only. The amount of information was also equalized between the two conditions. The results contradict the idea of an economical object internal representation avoiding redundancy and automatic duplication of symmetry along the vertical axis. In the discussion, we propose to explain these results by spatial frequency time integration, the closure process and the three-dimensionality of the presented junctions. Key words : object recognition, componential, redundance, information theory, symmetry, perceptive closure.

Perceptual learning in visual search: Fast, enduring, but non-specific

Visual search has been suggested as a tool for isolating visual primitives. Elementary “features” were proposed to involve parallel search, while serial search is necessary for items without a “feature” status, or, in some cases, for conjunctions of “features”. In this study, we investigated the role of practice in visual search tasks. We found that, under some circumstances, initially serial tasks can become parallel after a few hundred trials. Learning in visual search is far less specific than learning of visual discriminations and hyperacuity, suggesting that it takes place at another level in the central visual pathway, involving different neural circuits.

Spatio-temporal stages in face and word processing. 1. Depth recorded potentials in the human occipital and parietal lobes

Evoked potentials (EPs) were used to help identify the timing, location, and intensity of the information-processing stages applied to faces and words in humans. EP generators were localized using intracranial recordings in 33 patients with depth electrodes implanted in order to direct surgical treatment of drug-resistant epilepsy. While awaiting spontaneous seizure onset, the patients gave their fully informed consent to perform cognitive tasks. Depth recordings were obtained from 1198 sites in the occipital, temporal and parietal cortices, and in the limbic system (amygdala, hippocampal formation and posterior cingulate gyrus). Twenty-three patients received a declarative memory recognition task in which faces of previously unfamiliar young adults without verbalizable distinguishing features were exposed for 300 ms every 3 s; 25 patients received an analogous task using words. For component identification, some patients also received simple auditory (21 patients) or visual (12 patients) discrimination tasks. Eight successive EP stages preceding the behavioral response (at about 600 ms) could be distinguished by latency, and each of 14 anatomical structures was found to participate in 2-8 of these stages. The earliest response, an N75-P105, focal in the most medial and posterior of the leads implanted in the occipital lobe (lingual g), was probably generated in visual cortical areas 17 and 18. These components were not visible in response to words, presumably because words were presented foveally. A focal evoked alpha rhythm to both words and faces was also noted in the lingual g. This was followed by an N130-P180-N240 focal and polarity-inverting in the basal occipitotemporal cortex (fusiform g, probably areas 19 and 37). In most cases, the P180 was evoked only by faces, and not by words, letters or symbols. Although largest in the fusiform g this sequence of potentials (especially the N240) was also observed in the supramarginal g, posterior superior and middle temporal g, posterior cingulate g, and posterior hippocampal formation. The N130, but not later components of this complex, was observed in the anterior hippocampus and amygdala. Faces only also evoked longer-latency potentials up to 600 ms in the right fusiform g. Words only evoked a series of potentials beginning at 190 ms and extending to 600 ms in the fusiform g and near the angular g (especially left). Both words and faces evoked a N150-P200-PN260 in the lingual g, and posterior inferior and middle temporal g.(ABSTRACT TRUNCATED AT 400 WORDS)

The Change in Critical Technologies for Computational Physics

It is noted that the types of technology required for computational physics are changing as the field matures. Emphasis has shifted from computer technology to algorithm technology and, finally, to visual analysis technology as areas of critical research for this field. High-performance graphical workstations tied to a supercommunicator with high-speed communications along with the development of especially tailored visualization software has enabled analysis of highly complex fluid-dynamics simulations. Particular reference is made here to the development of visual analysis tools at NASA's Numerical Aerodynamics Simulation Facility. The next technology which this field requires is one that would eliminate visual clutter by extracting key features of simulations of physics and technology in order to create displays that clearly portray these key features. Research in the tuning of visual displays to human cognitive abilities is proposed. The immediate transfer of technology to all levels of computers, specifically the inclusion of visualization primitives in basic software developments for all work stations and PCs, is recommended.

Hierarchical construction of orientation and velocity selective filters

As a step towards the early measurement of visual primitives, the authors outline design criteria for the extraction of orientation and velocity information, and present a variety of tools useful in the construction of simple linear filters. A hierarchical parallel-processing scheme is used in which nodes compute a weighted sum of inputs from within a small spatio-temporal neighborhood. The resulting scheme is easily analyzed and provides mechanisms sensitive to narrow ranges of both image velocity and orientation. The hierarchical approach in combination with separability in the first levels yields an efficient implementation. >

Has feature integration theory come unglued? A reply to Tsal.

Tsal (1989) presented a critical review of feature integration theory (FIT) and of a number of studies that have purported to show a direct relationship between focal attention and the perception of illusory conjunctions. Conceptual difficulties with the theory are highlighted, and Tsal concluded that the reviewed studies have not provided sufficient empirical support. We argue that neither Tsal's conceptual criticism nor his empirical ones are convincing enough to warrant rejection of FIT and that he has not suggested an obvious and unique theoretical alternative that inspires an appropriate empirical test. A problem for FIT that Tsal did not raise is discussed. That is, the visual primitives assumed by the theory are not clearly specified or related to known neurophysiological or psychophysical evidence concerning early vision. Nevertheless, we conclude that HT remains a viable theoretical framework. Treisman's feature integration theory (FIT) proposes that illusory conjunctions can occur if attention is diverted from a display of several figures (Treisman & Schmidt, 1982). Converging evidence from a wide variety of paradigms has supported the theory (e.g., Treisman, 1985; Treisman & Gelade, 1980; Treisman & Schmidt, 1982). Tsal (1989) critiqued four studies that purport to show a direct relationship between attention and illusory conjunctions and argued that the lack of empirical support for FIT, combined with conceptual difficulties the model faces, renders its major claims unsupported. In this article we hope to demonstrate that none of his attacks on FIT can be sustained and that despite some ambiguity FIT remains a viable framework.

Measurement of the alpha-cell-migration in the entorhinal region: A marker for the developmental disturbances in schizophrenia? : P. Falkai∗, B. Bogerts∗, G.W. Roberts∗∗ and T.J. Crow∗∗ ∗Department of Psychiatry, University of Düsseldorf, Bergische Landstrasse 2, 4000 Düsseldorf 12, FRG; and ∗∗Division of Psychiatry, Clinical Research Centre, Watford Road, Harrow, Middlesex HA1 3UJ, UK

Measurement of the alpha-cell-migration in the entorhinal region: A marker for the developmental disturbances in schizophrenia? : P. Falkai∗, B. Bogerts∗, G.W. Roberts∗∗ and T.J. Crow∗∗ ∗Department of Psychiatry, University of Düsseldorf, Bergische Landstrasse 2, 4000 Düsseldorf 12, FRG; and ∗∗Division of Psychiatry, Clinical Research Centre, Watford Road, Harrow, Middlesex HA1 3UJ, UK