VISUALIZATION: Out-of-Core Simplification and Multiresolution Visualization of Large Volume Data Exploring Topological Features

\大规模数据的核心外简化和多分辨率可视化探索拓扑结构Yi-Jen Chiang，Polytechnic University of纽约近年来，随着大规模数据的规模呈指数级增长，科学可视化面临着新的挑战。数据的庞大规模往往使交互式探索的任务变得不可能，科学家们通过对原始数据的蛮力试错过程来提出所需的可视化参数不再可行;需要额外的特征分析信息来指导他们完成数据探索过程。 解决大规模可视化问题有两个重要且有前途的研究方向：多分辨率技术和核外方法。此外，标量场数据集最重要的特征之一是嵌入在体数据中的所有等值面的拓扑结构，这在理解标量场的行为中起着核心作用。本研究涉及开发核外简化算法，该算法平滑地简化体网格的几何形状以及嵌入的等值面的拓扑结构，以构建多分辨率体层次结构，以及核外细节层次（LOD）可视化算法，包括等值面提取和直接体绘制，它们使用多分辨率体层次作为统一的基础结构，支持满足拓扑和几何误差界限的LOD可视化。正在开发的技术还提供了所有等值面拓扑特征的路线图，以指导用户完成数据探索过程。研究人员对稳态（即，单个时间步长）和时变数据，并且通常考虑表示为四面体网格的不规则网格体积数据集的类别。该项目正在探索理论与实践之间的丰富相互作用，将莫尔斯理论和计算拓扑学的方法应用于可视化算法的设计。我们的目标是开发一系列新颖的、核心外的可视化算法，这些算法探索拓扑特征，以及一个统一的、概念验证的可视化系统，该系统配备了一个拓扑分析用户界面，使科学家能够在他们的台式PC上对比计算机主内存中大得多的数据集进行有效的特征提取和高效的可视化。该项目将促进包括结构力学、计算流体动力学和冲击物理学在内的主题领域的科学数据的可视化分析。

\Out-of-Core Simplification and Multiresolution Visualization of Large Volume Data Exploring Topological FeaturesYi-Jen Chiang, Polytechnic University of New York In recent years, new challenges for scientific visualization emerged as the size of large-scale data grew exponentially. The sheer size of data often makes the task of interactive exploration impossible, and it is no longer feasible for the scientists to come up with the desired visualization parameters by a brute-force try-and-error process on the raw data; additional feature-analysis information is needed to guide them through the data exploration process. There are two important and promising research directions towards solving large-scale visualization problems: multiresolution techniques and out-of-core approaches. Also, one of the most important features of a scalar-field dataset is the topology of all isosurfaces embedded in the volume data, which plays a central role in understanding the behavior of the scalar field.This research involves the development of out-of-core simplification algorithms that smoothly simplify the geometry of the volume mesh as well as the topology of the embedded isosurfaces to build a multiresolution volume hierarchy, and out-of-core level-of-detail (LOD) visualization algorithms, including isosurface extraction and direct volume rendering, that use the multiresolution volume hierarchy as a unified infrastructure to support LOD visualization satisfying both the topology and geometry error bounds specified. The techniques being developed also provide a road map of the topological features of all isosurfaces in order to guide the user through the data exploration process. The investigators work on both steady-state (i.e., single time step) and time-varying data and typically consider the class of irregular-grid volume datasets represented as tetrahedral meshes. The project is exploring the rich interplay between theory and practice by applying approaches from Morse theory and computational topology to the design of visualization algorithms. The aim is the development of a collection of novel, out-of-core visualization algorithms that explore topological features, together with a unified, proof-of-the-concept visualization system equipped with a topology-analysis user interface that enables scientists to perform effective feature extraction and efficient visualization on their desktop PCs for much larger datasets than can fit in the computers main memory. This project will facilitate the visual analysis of scientific data for topic areas that include structural mechanics, computational fluid dynamics, and shock physics.