Information

  • Publication Type: Master Thesis
  • Workgroup(s)/Project(s):
  • Date: July 2020
  • Date (Start): 21. February 2010
  • Date (End): 27. July 2020
  • Diploma Examination: 27. July 2020
  • Open Access: yes
  • First Supervisor: Stefan BrucknerORCID iD
  • Pages: 110
  • Keywords: Molecular Visualization, Real-Time Rendering, Computer Graphics

Abstract

The complexity of biomolecular data sets is both high, and still rising. Three-dimensional models of molecules are used in research to test and investigate their properties. Such models can consist of several millions of atoms. Additionally, visual enhancement methods and molecular surface models are helpful when visualizing molecules. There is therefore a demand for efficient and flexible data structures to accommodate such large point-based data sets. Existing solutions in the field of molecular visualization for large data sets include the use of, in most cases, regular grid-based data structures, as well as levels of detail. Other papers focus on repeating structures or improving the efficiency of surface models. We propose an octree-based data structure that divides space into areas of similar density, and provides several levels of detail. Our approach is optimized for a single time-step, moving much of the computational overhead into a pre-processing step. This allows us to speed up frame rates for interactive visualizations using visibility culling, least recently used caching based on the pre-built octree data structure, and level of detail solutions such as depth-based level of detail rendering. In our evaluation, we show that level of detail rendering significantly improves frame rates, especially in the case of distance-based level of detail selection while keeping the amount of details in the foreground high. Both the possibility to reduce the resolution and the caching strategy that allows us to only upload visible parts of the data set make it possible to render data sets that previously exhausted the capacities of our test set-up. We found the main advantage of a density based octree, instead of a regular division of space, to be in neighbourhood-based calculations, such as the clustering algorithm required to build levels of detail. This could prove particularly useful for the implementation of a Solvent Excluded Surface (SES) representation model, which would be an important feature to include when developing the framework further.

Additional Files and Images

Additional images and videos

Additional files

Weblinks

BibTeX

@mastersthesis{Nowak_2020,
  title =      "Hierarchical Multi-resolution Data Structure for Molecular
               Visualization",
  author =     "Milena Nowak",
  year =       "2020",
  abstract =   "The complexity of biomolecular data sets is both high, and
               still rising. Three-dimensional models of molecules are used
               in research to test and investigate their properties. Such
               models can consist of several millions of atoms.
               Additionally, visual enhancement methods and molecular
               surface models are helpful when visualizing molecules. There
               is therefore a demand for efficient and flexible data
               structures to accommodate such large point-based data sets.
               Existing solutions in the field of molecular visualization
               for large data sets include the use of, in most cases,
               regular grid-based data structures, as well as levels of
               detail. Other papers focus on repeating structures or
               improving the efficiency of surface models. We propose an
               octree-based data structure that divides space into areas of
               similar density, and provides several levels of detail. Our
               approach is optimized for a single time-step, moving much of
               the computational overhead into a pre-processing step. This
               allows us to speed up frame rates for interactive
               visualizations using visibility culling, least recently used
               caching based on the pre-built octree data structure, and
               level of detail solutions such as depth-based level of
               detail rendering. In our evaluation, we show that level of
               detail rendering significantly improves frame rates,
               especially in the case of distance-based level of detail
               selection while keeping the amount of details in the
               foreground high. Both the possibility to reduce the
               resolution and the caching strategy that allows us to only
               upload visible parts of the data set make it possible to
               render data sets that previously exhausted the capacities of
               our test set-up. We found the main advantage of a density
               based octree, instead of a regular division of space, to be
               in neighbourhood-based calculations, such as the clustering
               algorithm required to build levels of detail. This could
               prove particularly useful for the implementation of a
               Solvent Excluded Surface (SES) representation model, which
               would be an important feature to include when developing the
               framework further.",
  month =      jul,
  pages =      "110",
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Research Unit of Computer Graphics, Institute of Visual
               Computing and Human-Centered Technology, Faculty of
               Informatics, TU Wien",
  keywords =   "Molecular Visualization, Real-Time Rendering, Computer
               Graphics",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2020/Nowak_2020/",
}