This paper describes an experimental exploratory information visualization system under development at PNNL for integrating multimedia data and visualizing the contents of large multimedia databases. Our system protot...
详细信息
ISBN:
(纸本)0818672951
This paper describes an experimental exploratory information visualization system under development at PNNL for integrating multimedia data and visualizing the contents of large multimedia databases. Our system prototype, named Starlight, represents multimedia corpora graphically as collections of icons in 3-space. Each icon in the information display represents an individual element of a multimedia database. The proximity of any two icons in the display is an indication of their general similarity, providing fast access to thematically related information. We extend the effectiveness of this approach by coupling it with a linkage display system to enable the simultaneous visualization of database structure and contents at multiple levels of abstraction. The utility of the system is further extended by enabling a variety of graphical and text-based interactions with the information display. The system enables the interactive generation of multiple simultaneous `views' into multimedia databases, providing intelligence analysts with the potential to process large multimedia data volumes quickly and effectively.
Our approach to testing graphical user interfaces involves logging large amounts of data. These logs capture information at the key press and mouse click level about how an application is used. Since the raw data is v...
详细信息
ISBN:
(纸本)081867668X
Our approach to testing graphical user interfaces involves logging large amounts of data. These logs capture information at the key press and mouse click level about how an application is used. Since the raw data is voluminous and nor at a useful level of detail, we use analysis and visualization to find information that is interesting and useful to a usability analyst but was previously buried in the data. We call some of our custom visualizations ''contextual'' meaning they use key elements of the context the data was collected in as an organizing structure. We expect this type of visualization to be easier and faster to understand and more helpful than traditional charts. We hope that our finding a natural geometry for these visualizations will inspire others whose data apparently has no inherent geometry to find natural ways to visualize their data.
An observed image can be considered to be a single sample of a stochastic process under an assumed model. It is often desirable to generate a multitude of scenes which have the same stochastic properties as the origin...
详细信息
An observed image can be considered to be a single sample of a stochastic process under an assumed model. It is often desirable to generate a multitude of scenes which have the same stochastic properties as the original scene as a means of evaluating and validating proposed models. A class of stochastic models has been developed to characterize landscape processes represented in multispectral imagery and then to simulate these processes. In particular, models are derived from remotely sensed imagery and utilized to develop initial conditions for temporal simulations of vegetation in an ecology study. Landscape observed in remotely sensed imagery often exhibits characteristic patch mosaic structures at largescale and class dependent variability within each region at detailed scale. A Markov random field (MRF) model is employed to model the region process as a largescale characteristic and generate a spatially aggregated class map. Boundary variation between adjacent regions is represented using a fuzzy approach implemented within a multiresolution data structure. Class dependent variability and noise are superimposed on the resultant regions.
The problem of determining the eigenvalues of large matrices occurs often in the design and analysis of modern tomography systems. As there is an interest in solving systems containing an ever-increasing number of var...
详细信息
The problem of determining the eigenvalues of large matrices occurs often in the design and analysis of modern tomography systems. As there is an interest in solving systems containing an ever-increasing number of variables, current research effort is being made to create more robust solvers which do not depend on some special feature of the matrix for convergence (e.g. block circulant), and to improve the speed of already known and understood solvers so that solving even larger systems in a reasonable time becomes viable. Our standard techniques for singular value analysis are based on sparse matrix factorization and are not applicable when the input matrices are large because the algorithms cause too much fill. Fill refers to the increase of non-zero elements in the LU decomposition of the original matrix A (the system matrix). So we have developed iterative solutions that are based on sparse direct methods. data motion and preconditioning techniques are critical for performance. This conference paper describes our algorithmic approaches for largescale singular value analysis of spatially variant imaging systems, and in particular of PCR2, a cylindrical three-dimensional PET imager built at the Massachusetts General Hospital (MGH) in Boston. We recommend the desirable features and challenges for the next generation of parallel machines for optimal performance of our solver.
This paper describes an experimental exploratory information visualization system under development at PNNL for integrating multimedia data and visualizing the contents of large multimedia databases. Our system protot...
详细信息
This paper describes an experimental exploratory information visualization system under development at PNNL for integrating multimedia data and visualizing the contents of large multimedia databases. Our system prototype, named Starlight, represents multimedia corpora graphically as collections of icons in 3-space. Each icon in the information display represents an individual element of a multimedia database. The proximity of any two icons in the display is an indication of their general similarity, providing fast access to thematically related information. We extend the effectiveness of this approach by coupling it with a linkage display system to enable the simultaneous visualization of database structure and contents at multiple levels of abstraction. The utility of the system is further extended by enabling a variety of graphical and text-based interactions with the information display. The system enables the interactive generation of multiple simultaneous "views" into multimedia databases, providing intelligence analysts with the potential to process large multimedia data volumes quickly and effectively.
Our approach to testing graphical user interfaces involves logging large amounts of data. These logs capture information at the key press and mouse click level about how an application is used. Since the raw data is v...
详细信息
Our approach to testing graphical user interfaces involves logging large amounts of data. These logs capture information at the key press and mouse click level about how an application is used. Since the raw data is voluminous and not at a useful level of detail, we use analysis and visualization to find information that is interesting and useful to a usability analyst but was previously buried in the data. We call some of our custom visualizations "contextual" meaning they use key elements of the context the data was collected in as an organizing structure. We expect this type of visualization to be easier and faster to understand and more helpful than traditional charts. We hope that our finding a natural geometry for these visualizations will inspire others whose data apparently has no inherent geometry to find natural ways to visualize their data.
Interactive techniques are powerful tools for manipulating visualizations to analyze, communicate and acquire information. This is especially true for largedata sets or complex 3D visualizations. Although many new ty...
详细信息
Interactive techniques are powerful tools for manipulating visualizations to analyze, communicate and acquire information. This is especially true for largedata sets or complex 3D visualizations. Although many new types of interaction have been introduced recently, very little work has been done on understanding what their components are, how they are related and how they can be combined. This paper begins to address these issues with a framework for classifying interactive visualizations. Our goal is a framework that will enable us to develop toolkits for assembling visualization interfaces both interactively and automatically.
The analysis of large amounts of data to extract generalizations, exceptions, trends, and hidden relationships is a common activity in the business and scientific communities. While some kinds of "knowledge"...
详细信息
The analysis of large amounts of data to extract generalizations, exceptions, trends, and hidden relationships is a common activity in the business and scientific communities. While some kinds of "knowledge" can be extracted automatically with preselected algorithms or data mining techniques, others require an experienced human, often an expert in analysis, the business or scientific context, or both. We have found that such humans combine exploration, the search for a relevant subset or view of the data, with analysis, statistical or other techniques for measurement. We designed and implemented a visual language, IDEA, to assist the data analyst in these two intertwined tasks. The language is a convenient representation for dataanalysis and provides environmental support for keeping track of sequences of operations, reuse of the dataanalysis itself, and enforced semantics between operations and data.
We present a computational steering approach for controlling, analyzing, and visualizing very largescale molecular dynamics simulations involving tens to hundreds of millions of atoms. Our approach relies on extensib...
详细信息
We present a computational steering approach for controlling, analyzing, and visualizing very largescale molecular dynamics simulations involving tens to hundreds of millions of atoms. Our approach relies on extensible scripting languages and an easy to use tool for building extensions and modules. The system is easy to modify, works with existing C code, is memory efficient, and can be used from inexpensive workstations over standard Internet connections. We demonstrate how we have been able to explore data from production MD simulations involving as many as 104 million atoms running on the CM-5 and Cray T3D. We also show how this approach can be used to integrate common scripting languages (including Python, Tcl/Tk, and Perl), simulation code, user extensions, and commercial dataanalysis packages.
暂无评论