Simplification algorithms for large virtual worlds
Abstract/Contents
- Abstract
- Metaverses are virtual worlds where users create the entire world: all objects, their models and behavior, are user-generated. Rich, large-scale metaverses have been imagined in the realm of fiction for decades. Such worlds, however, present significant technical challenges due to the limitations of network and graphics resources. Since the world is user-generated, the content has to be stored in a shared, networked resource such as the cloud. Since they contain hundreds of thousands of 3D objects, it is too expensive for a graphics card to render them at interactive frame rates. Finally, due to their dynamic, user-generated nature, we cannot rely on hand-crafted or pre-computed levels of detail. Combined, these factors necessitate a dynamic and automatic approach to aggregate and simplify 3D content. This dissertation describes a simplification pipeline for displaying large metaverses. The basic approach is to group objects into simplified aggregates, and then display a small subset of objects and aggregates that complete the view of the world. Building on this approach, the key contributions are a set of algorithms that optimize this approach for low download and rendering cost. These algorithms are based on the insight that metaverses generally possess a high degree of spatial coherence: similar and even identical objects are often placed closely together. Using a technique called instancing, such a group of identical objects can be efficiently represented as one object and a set of transformations to create its duplicates. The dissertation describes three main algorithms. First, it presents an algorithm to group visually similar objects into aggregates, reducing the download cost of virtual worlds by up to 25%. Second, it presents a technique to deduplicate highly similar objects within each aggregate, which further increases instancing and reduces download cost. Third, it proposes a new instance-aware algorithm for simplifying these aggregates' models. On highly instanced models, this algorithm results in simplified versions that are orders of magnitude smaller compared to existing simplification algorithms. Compared to existing techniques, these algorithms decrease the network transfer cost of displaying large metaverses by up to 80%. Furthermore, they reduce the rendering cost of displaying the metaverses by up to 23%. Combined with texture atlasing, an implementation detail crucial for rendering performance, this makes it possible to display these metaverses at interactive frame rates. Finally, we validate our results using a large, distributed metaverse rendered at interactive frame rates over a wide area network.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2013 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Azim, Tahir |
|---|---|
| Associated with | Stanford University, Department of Computer Science. |
| Primary advisor | Levis, Philip |
| Thesis advisor | Levis, Philip |
| Thesis advisor | Freedman, Michael J, 1979- |
| Thesis advisor | Hanrahan, P. M. (Patrick Matthew) |
| Advisor | Freedman, Michael J, 1979- |
| Advisor | Hanrahan, P. M. (Patrick Matthew) |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Tahir Azim. |
|---|---|
| Note | Submitted to the Department of Computer Science. |
| Thesis | Ph.D. Stanford University 2013 |
| Location | electronic resource |
Access conditions
- Copyright
- © 2013 by Tahir Azim
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...