Interactive Display of Complex Datasets

Principal Investigators: Dinesh Manocha and Ming Lin 
Funding Agencies: Lawrence Livermore National Laboratories / University of California 
Agency Number: B504968

Abstract
We propose to develop better algorithms and systems for interactive visualization of complex datasets. These include very large CAD/CAM models and volumetric datasets with multiple scalar values. The set of research goals include:

• Multi-resolution approaches: Over the last few years we have developed a number of algorithms to simplify geometric models. We will like to extend them to geometric models with multi-variate data (e.g. scalar fields) and volumetric datasets. We will also develop efficient representations for such datasets.
• Image-based rendering (IBR) for rendering acceleration: Our research has demonstrated that IBR techniques can be used to guarantee a fixed frame rate while rendering complex geometry. We will expand this work to include our new IBR techniques, such as Relief Textures, for both polygonal and volume data sets.
• Database management: The entire dataset will not fit in the main memory. Our goal is to develop dynamic paging schemes that load parts of the dataset that the user wants to see. We will organize the large dataset using a viewpoint emphasis function and use it for pre-fetching.
• Parallel rendering algorithms: We will devise algorithms for interactive display that take into account the nature of future graphics architectures: multiple graphics pipelines and multiple general-purpose processors. Thus, we can build systems that achieve higher speeds by automatically using resources not only for traditional geometry rendering, but also for alternative approaches, such as replacement of geometry with images, 3D image warping, geometric levels-of-detail, visibility culling, and multi-pass rendering.
• Haptic rendering: Develop techniques for interacting with large datasets using haptic interfaces. The goal is to develop a visualization environment where we can directly perceive and manipulate these objects. We propose to develop efficient algorithms for haptic display of large datasets using 3-DOF and 6-DOF haptic devices.