An electronic landscape is a virtual environment which, in addition to containing browsable information is a social place in its own right. The second year of the eSCAPE project is focussed on the construction and study of two thematic places - prototype demonstrates of these 'electronic landscapes'. The two prototypes explore two extremes of this concept. The first is that of the 'abstract' information space and is reported elsewhere. The second, reported here, is the 'physical' more literal interpretation of an 'electronic landscape'. The idea of a 'physical' metaphor is that the e-scape is modeled on a literal landscape, that presents itself in a manner familiar to participants in much the same way that real-world city immediately suggests familiar notions of transportation, navigation, interaction, purpose, and more subtle ideas such as the varying degrees of temporal permanence for its structures.

The UCLA Virtual World Data Server (VWDS) is a terabyte data server with a clustered architecture based on commodity PCs. It provides data storage for multiple, simultaneous, high- performance clients running different multimedia and virtual world applications. It serves the data on demand, in real time, with a pre-negotiated data transfer rate. UCLA is currently running two prototype servers and building a large VWDS for campus use.

The multidimensional data orb (mdOrb) is a technique for visualizing sets of discrete data points that possess high dimensionality. The mdOrb partitions a 3D virtual environment into a number of smaller working regions, each representing a lower dimensional projection of the data space. Data points are then projected into and visualized in each of these regions, and the relationships between regions allow for interpretation of the data in many dimensions. The algorithm is based on the distribution of a set of axes form a common central origin to the surface of a sphere. The axes represent the dimensions of the data space. Adjacent axes and the spaces they form can be used for varying visualization strategies, in this paper a number of such strategies are described. The mdOrb is an interactive system in which the user can move axes about the sphere's surface, changing the proximity of axes to each other and thus reconfiguring the set of adjacencies and visualization spaces. The mdOrb is a visualization algorithm that allows easy exploration, manipulation and visualization of complex multidimensional data spaces.

Methods for representing spacetime events as hyperobjects in 4D space have been developed that let one examine events from arbitrary spacetime perspectives. 3D objects that move either rigidly or nonrigidly are extruded to create 4D hyperobjects. The boundaries of the 4D objects are represented using polyhedra, in much the same way that the boundaries of 3D objects may be represented using polygons. The user views one or more static, 4D hyperobjects using software which lets the user control in an interactive, realtime fashion the position and orientation of a 3D cross- section of the hyperobjects. The 3D cross-sections are rendered using standard techniques of 3D computer graphics. Extrusion of nonrigid 3D objects is useful for visualizing events that involve objects with time-varying shape. For instance, one can use nonrigid extrusion to visualize a character animation, in which multiple frames of a walking humanoid character are used to create a single, static hyperobject that represents the entire animation. With these methods, one need no longer be limited to watching a movie in which time is a hidden axis. Rather, one can view and alter events immersively from arbitrary vantage points.

This paper describes research conducted in collaboration with Greater Manchester Police (UK), to evalute the utility of Virtual Environments for scene of crime analysis, forensic investigation, and law enforcement briefing and training. We present an illustrated case study of the construction of a high-fidelity virtual environment, intended to match a particular real-life crime scene as closely as possible. We describe and evaluate the combination of several approaches including: the use of the Manchester Scene Description Language for constructing complex geometrical models; the application of a radiosity rendering algorithm with several novel features based on human perceptual consideration; texture extraction from forensic photography; and experiments with interactive walkthroughs and large-screen stereoscopic display of the virtual environment implemented using the MAVERIK system. We also discuss the potential applications of Virtual Environment techniques in the Law Enforcement and Forensic communities.

WebVis, the Hierarchical Web Home Page Visualizer, is a tool for managing home web pages. The user can access this tool via the WWW and obtain a hierarchical visualization of one's home web pages. WebVis is a real time interactive tool that supports many different queries on the statistics of internal files such as sizes, age, and type. In addition, statistics on embedded information such as VRML files, Java applets, images and sound files can be extracted and queried. Results of these queries are visualized using color, shape and size of different nodes of the hierarchy. The visualization assists the user in a variety of task, such as quickly finding outdated information or locate large files. WebVIs is one solution to the growing web space maintenance problem. Implementation of WebVis is realized with Perl and Java. Perl pattern matching and file handling routines are used to collect and process web space linkage information and web document information. Java utilizes the collected information to produce visualization of the web space. Java also provides WebVis with real time interactivity, while running off the WWW. Some WebVis examples of home web page visualization are presented.

An ongoing challenge for visualization researchers is to create tightly coupled dynamic query interfaces that response quickly and smoothly to user interaction, regardless of the amount of data or the complexity of the display. Our solution is to follow a simple philosophy: during user activity, maintain the quality of visual context at the expense of visual content by reducing the impact of expensive visualization operations on interactivity. Our interactive visualization strategy splits finite computational resources by giving the user interface whatever resources it needs to maintain interactivity, and giving the data system whatever remains; user interface optimizations increase resources left for the data system, and data system optimizations make the best use of leftover resources. In this paper we describe three techniques that implement this strategy. First, we use a family of techniques called throttling to moderate the execution of queries and renders. Second, we improve refresh speed by elimination several unnecessary, graphically-intensive operations. Third, we preprocess and render data into bitmaps using an asynchronous imaging engine. Using these techniques, we have increased the speed and smoothness of interface navigation in our visualization framework even during expensive query and render operations.

To date, many web visualization applications have shown the usefulness of a hyperbolic tree. However, we have discovered that strict hierarchical tree structures are too limited. For many practical applications, we need to generalize a hyperbolic tree to a hyperbolic space. This approach results in massive cross-links in a highly connected graph that clutter the display. To resolve this problem, an invisible link technique is introduced. In this paper. we describe the navigation in a large hyperbolic space using invisible links in some detail. We have applied this invisible link method to three data mining visualization applications: e-business web navigation for URL visits, customer call center for question-answer service, and web site index creation.

This paper describes a set of visualization techniques that will support the training of environmental economic input- output life-cycle assessment (EIO-LCA) users. EIO-LCA is a software tool used to estimate the environmental effects of any product throughout the course of its life. The process of estimating the environmental effects of a product using the EIO-LCA software prototype consists of the following steps: problem formulation, computations, and results analysis. Empirical software evaluation has shown that user solutions to EIO-LCA problems yield large variability; therefore, users need to have a clear understanding of the model, and they need to be trained. Instead of using only long tutorial documents and tables to train the user to use the data sources, we have designed a software tool that uses visualization tools to train users in the problem formation and data analysis tasks. For the problem-formulation task, the coloring matrices will allow users to better understand the economic data used. Also, treemaps are included so users can evalute the relationships between different levels of aggregation used in the problem formulation. Treemaps were used to allow users to understand both the direct and indirect effects in the result analysis.

The latest results in the development of the software tool 'Procession' is presented. The research underlying Procession delivers a conceptual 3D framework for the interpretation of non-physical construction industry processes. Procession is the implementation of the proposed 3D framework, as an information visualization software tool. The conceptual transformation of construction clients' informational needs into 3D visual structures is documented. Also discussed is the development of an 'intelligent' software process to calculate the relevance of individual project elements. This is used to determine the representation of project elements within a 3D surface. Construction is not short of technologies for visualizing physical building models. However, it would seem that little or no consideration has been given to improving the intelligibility of non-physical construction processes. This type of information is usually known as Project Planning data and is concerned with the individual tasks that make up construction projects. While, there are software applications that allow access to this data for the professional members of the project team, clients are currently without a suitable tool. Procession's data surface is an abstract representation of three selected project dimensions. Its 3D progress reports provide construction clients with an 'at-a-glance' indication of project 'health'.

This paper examines the issues on visualizing an individual careers, considering the problem as a particular example of an event history. The even history is displayed as a pencil- like multi-faceted object in 2D or 3D space, with changes of states being represented by changes in color texture or height. The extension of these ideas to viewing many criminal histories is then described. The Lexis diagram provides a suitable paradigm and is used to position the pencils in space. The resulting graphical representation is closely related to the ideas of Bertin.

Recent trends in the software industry have shown that the number of software development projects and their increasing complexity will continue to increase the demand for skilled programmers. However, debugging technology has not kept pace and valuable programmer resources are being wasted in debugging cycles. Correction this trend requires innovative approaches to assisting programmers during the debugging cycle to make them more productive both time-wise and with respect to how thoroughly the application is tested and debugged. We have investigated techniques which allow programmers to visually steer an application. Through these new mechanisms, programmers can identify and locate problems without the need for revisiting the code. Through steering the programmer can change the application environment to more thoroughly understand and test the algorithm.

We present a visualization technique that allows a user to identify and detect patterns and structures within a multivariate data set. Our research builds on previous efforts to represent multivariate data in a 2D information display through the use of icon plots. Although the icon plot work done by Pickett and Brinstein is similar to our approach, we improve on their efforts in several ways. Our technique allows analysis of a time series without using animation; promotes visual differentiation of information clusters based on measures of variance; and facilitates exploration through direct manipulation of geometry based on scales of variance. Our goal is to provide a visualization that implicitly conveys the degree to which an elements ordered collection of attributes varies from the prevailing pattern of attributes for other elements in the collection. We apply this technique to multivariate abstract data nd use it to locate exceptional elements in a data set and divisions among clusters.

Todays car body models used for crash-worthiness simulations consist of around half a million finite elements. The interactive visualization of these large scale time- dependent geometries with constant topology on workstations platforms requires a variety of modeling an rendering optimizations. We present a memory efficient scene graph design, an algorithm to concatenate the mainly four-sided elements into optical quadrilateral strips, and a simplifier which generates an approximating triangle mesh by using the one-sided Hausdorff distance as an error measure. Furthermore we describe a technique to explore scalar data mapped onto complex scenes by hiding geometry with values outside an interactively specified range of interest. These optimizations allow for the first time interactive visualization of a full car crash on medium range graphics workstations. They are embedded in an application which is based on Cosmo3D/OpenGl Optimizer. It was developed in close cooperation with the BMW Group and it is in productive use.

In this paper we describe our efforts towards the parallel visualization of large data sets. We describe the fully threaded tree (FTT) structure developed at NRL to tackle the problem of massive parallel calculations using adaptive mesh refinement methods. All operations with FTT are performed in parallel and require only a small memory overhead. The FTT can be viewed as a data compression scheme that dramatically improves the performance of standard finite difference algorithms by performing calculations on the compressed data in situ. Because of the tremendous benefits of this type of data structure, it is of great interest to develop visualization algorithms that are native to the FTT. Using the FTT library, we convert the FTT data structure to an unstructured data set and discuss applications to both scatter dot visualization and parallel ray-tracing. The latter technique gives a good indication of the performance and scalability of the FTT algorithm for ray-tracking. We then discuss conversion of the FTT data structure for virtual reality visualization in an immersive room. Our results are presented using an example of a numerical calculation of a detonation in a rectangular cavity using from 1 to 2 million cells.

Many situations exist where the plotting of large data sets with categorical attributes is desired in a 3D coordinate system. For example, a marketing company may conduct a survey involving one million subjects and then plot peoples favorite car type against their weight, height and annual income. Scatter point plotting, in which each point is individually plotted at its correspond cartesian location using a defined primitive, is usually used to render a plot of this type. If the dependent variable is continuous, we can discretize the 3D space into bins or voxels and retain the average value of all records falling within each voxel. Previous work employed volume rendering techniques, in particular, splatting, to represent this aggregated data, by mapping each average value to a representative color.

This paper describes a method to reduce multi-texture 3D meshes using a multi-resolution wavelet analysis. Large and dense multi-modal meshes require new methods for efficient display. In this paper, we present a mesh simplification process, that inherently deals with multi-dimensional data set, controlled in a feature space composed of geometry, curvature, and the textures themselves. The result of the wavelet analysis using a multi-resolution analysis (MRA) based on the 2D quincunx-wavelet transform is considered as texture map called the 'detail relevance'. Virtual range and texture images are captured from selected viewpoints located around the object. The detail extraction is achieved using a multi-resolution approach based on the wavelet cascade analysis. The MRA process extracts detail information at various resolutions and produces a texture image that shows the relevance information attached to each vertex of the mesh. The user has input in this process to select what resolutions are more relevant than others. This approach is useful for filtering noise, preserving discontinuities, mining for surface details, reducing data, and many other applications. We present simplification results of digital elevation maps and 3D objects.

We discuss work in progress towards a comprehensive software system for interactive volume rendering and the techniques we use to achieve high frame rates. The system was designed for Pentium-III CPUs and is currently running under Windows NT. It accepts three voxel formats and provides perspective views, on-the-fly classification and gradient shading. For high rendering speed, the entire recasting algorithm has been implemented and optimized in assembler, taking advantage of the MMX instructions and the Streaming SIMD Extensions of Pentium-III CPUs. These instructions are used to speed up frequently performed volume rendering operations such as tri-linear interpolations and ray-volume intersection test. The algorithm run is in parallel on two CPUs using multi-threading, and fast access to the frame buffer is established using DirectDraw. For further speed- up, early ray termination has been implemented as well. However, the major obstacle towards high performance is the limited memory bandwidth, even more so because volume rendering requires 3D access to the data set, and frequency access to tables. Thus, a memory layout was developed which maximizes the processor cache hit rate and keeps frequently used tables in the L1 cache, although at the cost of an increased memory consumption.

Visualization is about discovery and understanding; the user wishes to gain a correct insight into the underlying information, to explore and analyze how different parts are related. Thus, presentation, exploration and explanation tools are used with manipulation and investigative techniques to display, discover and gain a 'correct dissemination' of the information. Moreover, by displaying the information simultaneously in multiple ways the user is aided in their investigation. Such multiform techniques may be generated through various algorithms; we organize these methods according to how they apply to the individual stages of the dataflow paradigm. These multiforms may be displayed in separate windows. Multiple views are useful (1) to overcome misinterpretations and provide additional insight, (2) for scientific exploration tasks of relating, coupling and to aid the 'drilling down' of information, and (3) to provide alter-native viewpoints by expressing different user-interpretations of the same information. Finally, to use multiple views effectively they should be, among other things, easily created, automatically coupled to other views and dynamically manipulated.

An effective visualization system depends on a user's ability to interpret a visual representation and made valid inferences. This paper first summarizes the role of domain knowledge when interpreting a visualization. Once the visual perception system has interpreted the visual representation, the user transforms the data into information by the introduction of domain knowledge; these are the rules or items of knowledge that are relevant tot this visual representation allowing the user to make meaningful inferences. In the remainder of the paper we concentrate on a visualization architecture that encapsulates domain knowledge to improve user interpretation of a visual representation. We use an agent-based paradigm to provide a distributed model of computation which moves away from a heavyweight constrained based algorithm towards a lightweight distributed system that empower individual data items. Finally, we present DIME, an implementation based on this approach. DIME is an ongoing research project that tightly integrates data storage, knowledge capture, and information visualization in a 'visual environment'.

We discuss the evolution of the NIST Information Retrieval Visualization Engine. This prototype employs modern interactive visualization techniques to provide easier access to a set of documents resulting from a query to a search engine. The motivation and evaluation of several design features, such as keywork to concept mapping, explicit clustering, the use of 3D vs. 2D, and the relationship of visualization to logical structure are described. In particular, the result of an extensive usability experiment show how visualization may lead to either increased or decreased cognitive load.

Cluster analysis is a common exploratory multivariate data analysis method which groups similar objects together. The rapid growth of data size in cases and dimensions leads cluster analysis to receive more attention. Data visualization and case reduction are two important issues in cluster analysis. The visualization of data helps us to detect clusters which are difficult to detect with other clustering algorithms by using human pattern perception ability. The time and memory requirements for clustering are often problems especially for large data sets. This makes it difficult for promising but computationally heavy clustering methods to be run for large data. In this paper we will address these two issues by introducing our visualization software and the algorithm for case reduction.

Software visualization is a graphical representation of software characteristics and behavior. Certain modes of software visualization can be useful in isolating problems and identifying unanticipated behavior. In this paper we present a new approach to aid understanding of object- oriented software through 3D visualization of software metrics that can be extracted from the design phase of software development. The focus of the paper is a metric extraction method and a new collection of glyphs for multi- dimensional metric visualization. Our approach utilize the extensibility interface of a popular CASE tool to access and automatically extract the metrics from Unified Modeling Language class diagrams. Following the extraction of the design metrics, 3D visualization of these metrics are generated for each class in the design, utilizing intuitively meaningful 3D glyphs that are representative of the ensemble of metrics. Extraction and visualization of design metrics can aid software developers in the early study and understanding of design complexity.

The importance of information as a resource for economic growth and education is steadily increasing. Due to technological advances in computer industry and the explosive growth of the INternet much valuable information will be available in digital libraries. This paper introduces a system that aims to support a user's browsing activities in document sets retrieved from a digital library. Latent Semantic Analysis is applied to extract salient semantic structures and citation patterns of documents stored in a digital library in a comptutationally expensive batch job. At retrieval time, cluster techniques are used to organize retrieved documents into clusters according to the previously extracted semantic similarities. A modified Boltzman algorithm is employed to spatially organize the resulting clusters and their documents in the form of a 3D information landscape or 'i-scape'. THe i-scape is then displayed for interactive exploration via a multi- modal, virtual reality CAVE interface. Users' browsing activities are recorded and user models are extracted to give newcomers online help based on previous navigation activity as well as to enable experienced users to recognize and exploit past user traces. In this way, the system provides interactive services to assist users in the spatial navigation, interpretation, and detailed exploration of potentially large document sets matching a query.

In this paper, we present our work with the VisionDome, a fully immersive, single-projection, 3D display device. As contrasted to other immersive systems, the VisionDome does not require any equipment for the user to wear, is relatively inexpensive and portable, and can support multi- user collaboration. In this paper we describe the technical specifications of the VisionDome, as well as the problems encountered when viewing inside this immersive device. As a case study, we show the results of incorporating the VisionDome within GeoViz, which is a visualization software package for geophysical and geological data. We show how users of GeoViz are successfully finding new locations for oil more rapidly based on the immersive display of seismic and other data that they are viewing from inside the VisionDome.

This paper describes the development of applications for the interactive investigation of 3D geological and geophysical data, using an Interactive Workbench. The Interactive Workbench is essentially a large light table, on which stereo images are back-projected. The position of the user's head and hands are determined by a magnetic tracking system. The user interacts via a 3D mouse with geoscientific data objects that appear to be placed on or above the table. A high-end Silicon Graphics Workstation renders the stereo images 30 times a second for the users current eye position. We have implemented a pen-driven user interface as well as virtual tools for positioning, picking and cutting in order to visualize and, in some instances, edit geological objects like horizons, faults, seismic data, well bores, etc. This approach enables geoscientists to more fully exploit 3D geoscientific data by providing them with a truly 3D display that can be shared with others. In addition, the 3D interactivity that this technology provides can support more productive modeling of geoscientific data and more efficient verification of the resulting interpretations.

Modern computational fluid dynamics simulations are capable of the detailed simulation of fluid flow. The output data sets of these simulations are very large and information rich. The importance of data visualization is clearly recognized for the presentation of these data set. For gaining new insight in the nature of flow, interactive visualization methods are essential. The goal of our work is to develop an environment which allows fluid dynamics experts to analyze very large flow fields. This paper motivates the need for an interactive visualization environment. The design of the environment is centered around two fundamental beliefs: first, the environment should integrate modeling and visualization. Second, interactive visualization is essential so that exploration is stimulated. We discuss two interactive visualization methods and their application to a flow field resluting from a direct numerical simulation. We believe that the working methods discussed in this paper will be typical of future visualization environments.

Computers have enabled the generation and acquisition of large vector and tensor data-sets. Reynolds numbers, and in data acquisition producing large data-set such as weather and ocean current databases. However, techniques for automated searching and comparison of large vector and tensor data-sets have been lacking. Many of the techniques are based on visual comparisons which do not have the quantitative rigor required for automated comparisons. Other methods which use data based comparisons require a common domain, grid alignment, and interpolation often making data- sets acquired from different sources difficult to compare. This paper provides comparison methods that are based on the topology of the vector field. These methods can be used for both 2D and 3D fields, and depending upon the complexity of the field, different levels of comparisons can be utilized to reduce search times. Topology based methods have the added benefits of describing a field that is independent of the grid, and that is based solely on topological components which have a significant compression over the original field.

The success of using streamline technique for visualizing a vector field usually depends largely on the choosing of adequate seed points. This paper propose a new technique for automatically placing seed points to create evenly spaced streamlines on 3D parametric surfaces found in curvilinear grids. The new technique extends Jobard and Lefer's distance-based single pass approach for placing streamlines in the 2D computational space of the surface. Experimental result show that the new technique produces streamline images of competitive quality at much lower computational expense image-guided progressive refinement approach. A method for compensating the visual streamline density distortion caused by projection is also presented.

For the multi variable volumetric tensor field visualization, an efficient direct rendering technique without using geometrical primitive is proposed. The bi- directional reflectance shading model is used to map the anisotropy stress shear tensor components in direct volume rendering. We model the sub-pixel-sized microfacet at tensor sampling points. The nine component of 3D tensor field are mapped onto grid deformation, opacity mapping, color specification, and normal directions of these microfacets. The ray integration is executed though these irregular infinitesimal microfacets distribution. This direct tensor rendering was applied for at-a-glance tensor visualization of earthquake simulation. That realized a view of deformed structure, stress distribution, local shear discontinuity and the shock front, integrated in a single image. The characteristic P- and S-wave modes are distinguished in the rendered earthquake simulations. Compared with the glyph representation of tensor features, the direct tensor rendering gives the general and total image of tensor field even for the low resolution pixel planes, because the sampling object is assumed as infinitesimally small. the computational cost of direct tensor rendering is not so high than that of scalar volume rendering because the modifications are only ins hading calculation but not in the ray integration.

In this paper, we present further improvement to the DISCUS software that can be used to record and analyze the flow and constants of business process simulation session discussion. The tool was initially introduced in 'visual data exploration and analysis IV' conference. The initial features of the tool enabled the visualization of discussion flow in business process simulation sessions and the creation of SOM analyses. The improvements of the tool consists of additional visualization possibilities that enable quick on-line analyses and improved graphical statistics. We have also created the very first interface to audio data and implemented two ways to visualize it. We also outline additional possibilities to use the tool in other application areas: these include usability testing and the possibility to use the tool for capturing design rationale in a product development process. The data gathered with DISCUS may be used in other applications, and further work may be done with data ming techniques.

In this paper we present the CyberNet research project. The aim of this project is to study how 3D visualization may help the user in the process of monitoring large amounts of dynamic information. The project focuses on a specific application domain, network management, but it is designed so that it can be applied to other domains. The paper present how information is collected and structured in order to define services that the user wants to monitor. We also present the impact of the dynamic nature of information on the system. Each 3D world represent a service, that is a model that groups and structures all the information related to a specific aspect of the network. Each modification of the service has a real time impact on the virtual world. This representation is done within the context of a 3D metaphor that defines the rules used to map services onto graphical elements. It also defines how information is mapped onto the visual parameters of these graphical elements. The metaphor also handles suitable navigation and interaction mechanisms to allow the user to explore and manage the virtual world.

The paper presents specialist algorithms of morphologic analysis of shapes of selected organs of abdominal cavity proposed in order to diagnose disease symptoms occurring in the main pancreatic ducts and upper segments of ureters. Analysis of the correct morphology of these structures has been conducted with the use of syntactic methods of pattern recognition. Its main objective is computer-aided support to early diagnosis of neoplastic lesions and pancreatitis based on images taken in the course of examination with the endoscopic retrograde cholangiopancreatography (ERCP) method and a diagnosis of morphological lesions in ureter based on kidney radiogram analysis. In the analysis of ERCP images, the main objective is to recognize morphological lesions in pancreas ducts characteristic for carcinoma and chronic pancreatitis. In the case of kidney radiogram analysis the aim is to diagnose local irregularity of ureter lumen. Diagnosing the above mentioned lesion has been conducted with the use of syntactic methods of pattern recognition, in particular the languages of shape features description and context-free attributed grammars. These methods allow to recognize and describe in a very efficient way the aforementioned lesions on images obtained as a result of initial image processing into diagrams of widths of the examined structures.

Surgery simulation is a growing field of research comprising the efforts of various disciplines including Computer Graphics, Computer Vision, Medicine, Mechanics, Robotics and even Animation. We try to combine, adapt and extended different solutions to this problem and re-assemble them using mid-range 3D graphics hardware, modern object-oriented methods and free visualization toolkits. Reconstructions of the physical based realistic 3D models are achieved form CT scans. In particular, we focus on the generic model concept, an evolving methods to encapsulate generic information in a generic 3D mesh. This model is subsequently deformed using multivariate scattered data interpolation technique show that it matches the reconstructed model of the patient being studied, under the control of common landmark points. We describe a wy to build a topological relationship between these non-related geometrical and topological models, thus opening a framework to transfer different kinds of generic information, ranging from algorithmic simplification and computation hints to anatomical features or learning material. Starting from several uncalibrated photographs, we also show how 2D features points may be used to recover the camera parameters and employ them as control points to deformed our model in a similar fashion, so that the texture information is retrieved after projecting the mode into the photographic pose.

Mobile robots commonly employ a variety of sensor in order to gather information about their immediate environment. Whereas some types of sensor produce a wealth of environmental information in a human-understandable for, it is often low-level data such as range-finder measurements which are of greatest interest to the operator. This paper describes an approach to the visualization of data form acoustic sensor which use echo-location to determine object distances. Acoustic distance measurement is problematic in that echoes may fail to be returned to the source, range assessments may be inaccurate, and ghost readings may also appear in the data. The visualization method described attempts to overcome the idiosyncrasies of acoustic sensing to present a relatively consistent representation of the robot's spatial environment to the operator. It maintains a geometric model of the perceived environment, allowing ambiguities in the current sensor data to be identified. The model is updated using polygon set operations to incorporate new sensor readings. Two features are recorded: the extent of free space identified within the environment, and the location of any obstacles which have been detected. This visualization approach is also tolerant to inaccuracies in odometry measurement. An experimental application of the method is reported.

We present a multiresolution technique for interactive texture-based volume visualization. This method uses an adaptive scheme that renders the volume in a region-of- interest at a high resolution and the volume further away from this region at progressively lower resolutions. We use indexed texture maps, which allow for interactive modification of the opacity transfer function. Our algorithm is based on the segmentation of texture space into an octree, where the leaves of the tree define the original data and the internal nodes define lower-resolution approximations. Rendering is done adaptively by selecting high-resolution cells close to a center of attention and low-resolution cells away from this area. We limit the artifacts introduced by this method by modifying the transfer functions in the lower-resolution data sets and utilizing spherical shells as a proxy geometry. It is possible to use this technique for viewpoint-dependent renderings.

DEVise is a general-purpose data visualization and exploration system capable of handling datasets from a variety of sources. Users create visual presentations, which they, or others, can dynamically explore to search for patterns in their data. The JavaScreen is a Java applet we have recently developed, which functions as a DEVise client. With it, users can explore visual presentation via the web, without having to actually install the DEVise software. Our paper covers three main areas: DEVise itself, including the fundamental DEVise model and adaptations required to support the JavaScreen; the JavaScreen and related software; and the use of DEVise and the JavaScreen in several applications.

This research applies recent advances in 3D isosurface reconstruction to images of test spheres and plant cells growing in suspension culture. Isosurfaces that represent object boundaries are constructed with a Marching Cubes algorithm applied to simple data sets, i.e., fluorescent test beads, and complex data sets, i.e., fluorescent plant cells, acquired with a Zeiss Confocal Laser Scanning Microscope (LSM). The marching cubes algorithm treats each pixel or voxel of the image as a separate entity when performing computations. To test the spatial accuracy of the reconstruction, control data representing the volume of a 25 micrometer test shaper was obtained with the LSM. This volume was then judged on the basis of uniformity and smoothness. Using polygon decimation and smoothing algorithms available through the visualization toolkit, 'voxellated' test spheres and cells were smoothed using several different smoothing algorithms after unessential polygons were eliminated. With these improvements, the shape of subcellular organelles could be modeled at various levels of accuracy. However, in order to accurately reconstruct these complex structures of interest to us, the subcellular organelles of the endosomal system or the endoplasmic reticulum of plant cells, measurements of the accuracy of connectedness of structures need to be developed.

We present methods for film visualization and browsing based on algorithms from computer vision, information visualization and on a hierarchical model for content semantics. We introduce OM-Images for the visualization of temporal changes in a moving image sequence. Together with interactive browsing technique the visualization methods can be used for the exploration of a film and video at different levels of abstraction. The proposed levels of abstraction are the physical, image, object or discourse level. Visualization methods are used to generate 1) static descriptions, which printed on paper yield a film book and 2) interactive documents, e.g. web pages or special video browsers. Finally, we give examples of a film book of a feature length film.