Algorithmic Development of Visualization Under Foveated Geometries

点击翻译按钮获取中文摘要

This project concerns the use of foveated geometries for scientific visualization applications. Two main motivations for such geometries are (1) the analogy with biological vision systems, and (2) their remarkably reduced pixel density that is logarithmic in the radius of the visual field. To compensate for the loss of peripheral vision in foveated geometries, the approach of active visualization is to allow the user interactive control of several parameters of vision, including rapid saccades and adjustable resolution. These are new capabilities, to be added to the more conventional navigational control. Foveated geometries will be a key to making current visualization systems more robust, scalable and extendible. The project will investigate three main areas: - Algorithms and data structures for rendering and manipulating foveated images. - Wavelet theory and algorithms for a corresponding concept of foveated multiresolution. - Application of active visualization in "thin-wire models" of computation such as the internet. In a typical "thin-wire" visualization scenario, imagine an image server with powerful computational resources, a client displaying images with the the computational power of stock PC's, and between them, a thin-wire. In current visualization applications, the bandwidth usage from client to server is negligible but the bandwidth in the other direction is a severe bottleneck. In exchange for a modest amount of bandwidth going from client to image server, active visualization is able to remove this bottleneck. Such a development can have major impact on visualization. ***