Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

Query-Driven Analysis and Visualization for Large-Scale Scientific Dataset using Geometry Summarization and Bitmap Indexing

Abstract Details

2017, Doctor of Philosophy, Ohio State University, Computer Science and Engineering.
The computational power of modern supercomputers grows rapidly, and it facilitates scientists to produce high-resolution datasets when simulating physical or weather models, which generate extreme scale data with multiple variables most of the time. However, storage, transmission, or exploration of such large-scale data is challenging. In the past decades, several visualization approaches have been developed to effectively explore datasets by displaying underlying information of datasets. Query-driven visualization is one of the prominent approaches, as it significantly reduces visual exploration time by only focusing on interesting or important features for further analysis and decision making. However, as the size of scientific datasets becomes too large, traditional data exploration approaches become ineffective. An emerging approach is to create data summarizations to first reduce the size of the dataset, and then perform data exploration on the data summarization. An ideal data summarization aims at preserving the characteristics of the raw data as much as possible while keeping the size small. However, to retrieve salient features from the raw data and create such importance-based data summarizations is challenging. In this dissertation, we address the issues that need to be solved when applying query-driven analysis and visualization using data summarizations. First, we focus on the problem of identifying salient features and evaluating selected features for creating a data summarization. To analyze a volumetric dataset, displaying isosurface is typically used to reveal the locations of values that the user focuses on. We propose a novel algorithm to select salient features described by isosurfaces and evaluate the chosen isosurfaces quantitatively. Our approach applies information theory to examine how much information left in an enclosed volume between two selected isosurfaces and then determine whether any isosurfaces should be included to enhance the information of surface summaries. Second, we focus on the task to efficiently search for the user-queried features using existing data summarization. We propose a novel algorithm to detect local histogram-based features with high performance by leveraging bitmap index. Rather than exhaustively searching for the target local histograms at all data voxels, our approach quickly determines the search space which is much less than all data points and detect the voxels whose local histograms match the user-defined histogram. Furthermore, we also propose two algorithms to solve the performance issues when extending the approach of efficient histogram-based feature search to multi-field datasets. Third, we aim to improve an existing data summarization in order to provide a better one in terms of the quality of the content and the size of the data. We propose two approaches to tackle the limitations of the bitmap data summarization. The first one is an adaptive sampling approach using bitmap index called information guided stratified sampling (IGStS) for creating a sampling-based bitmap that preserves the important characteristics of the raw data. Furthermore, we propose a novel data recovery approach to reconstruct the irregular subsampled dataset into a volume dataset with regular grid structure for qualitative post-hoc data exploration and visualization. We transform the data recovery problem to an optimal assignment problem and solve the value assignment problem by the Hungarian algorithm. The quantitative and visual efficacy of our proposed data sampling and recovery approaches are demonstrated through multiple experiments and applications.
Han-Wei Shen (Advisor)
Gagan Agrawal (Committee Member)
Yusu Wang (Committee Member)
164 p.

Recommended Citations

Citations

  • Wei, T.-H. (2017). Query-Driven Analysis and Visualization for Large-Scale Scientific Dataset using Geometry Summarization and Bitmap Indexing [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1512051705269695

    APA Style (7th edition)

  • Wei, Tzu-Hsuan. Query-Driven Analysis and Visualization for Large-Scale Scientific Dataset using Geometry Summarization and Bitmap Indexing. 2017. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1512051705269695.

    MLA Style (8th edition)

  • Wei, Tzu-Hsuan. "Query-Driven Analysis and Visualization for Large-Scale Scientific Dataset using Geometry Summarization and Bitmap Indexing." Doctoral dissertation, Ohio State University, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=osu1512051705269695

    Chicago Manual of Style (17th edition)