Information

  • Publication Type: PhD-Thesis
  • Workgroup(s)/Project(s): not specified
  • Date: 2007
  • Date (Start): September 2004
  • Date (End): June 2007
  • 1st Reviewer: Werner PurgathoferORCID iD
  • 2nd Reviewer: László Szirmay-Kalos
  • Rigorosum: 12. June 2007
  • First Supervisor: Michael WimmerORCID iD

Abstract

Computer graphics today plays an important role in 3D computer and video game creation, since all the images which are presented to the player are based on principles and algorithms devised from computer graphics research.

One important aspect with regards to the believability and consecutively immersion of players within game worlds are shadows. Creating high-quality dynamic shadows in real-time is still a very open field of research. One real-time shadowing approach that is very appealing due to its apparent elegant simplicity is shadow mapping. The term "apparent" is used deliberately here, since the elegance of shadow mapping comes at a price, namely the discretization of the first-hit visibility problem to a grid storing occluder depth information with regards to the light source. In practice this often leads to shadowing artifacts, due to too little information being available to answer the visibility query with enough accuracy. In this thesis, a new family of shadow mapping algorithms is presented, which address the problem of the shadow map not containing enough information, This is done without requiring a shadow map grid too large to be stored in memory and which cannot be filled with depth information in real time. Based on a basic brute-force approach, two smart algorithms are presented, which work in a manner adaptive to the resolution requirements of the scene, speeding up the process by at least an order of magnitude.

Another problem with relevance to the practical application of computer graphics to games is how to render a large number of objects fast enough to give the player a high level of responsiveness (and thereby, again, immersion). Noticing that due to perspective shortening objects farther away from the player do not need as much detail as objects which are nearer, computer graphics has come up with the concept of level of detail (LOD), where farther away objects use simpler representations; if done correctly, this leads to a large increase in rendering speed. While LOD has been studied extensively in computer graphics, one problem of the LOD technique dominating practical application, discrete LOD, has been largely ignored: In discrete LOD the simpler representations of an object are, as the name implies, discrete, i.e. they exist independently of each other. Switching between representations of different complexity by simply switching the representation in use leads to a disruptive discontinuity in player perception. This thesis presents a practical algorithm, that addresses this problem.

Linked to this research in the computer graphics fields of shadows and level of detail, is the case study of the GameTools Project, a Europe-wide project, funded by the European Union, which had as its agenda to bring the results of current computer graphics research done within the project to European game developing companies and companies from neighboring fields. The author holds the position of Community Manager within the GameTools Project, and presents an overview over the project and a more detailed view of his work as Community Manager.

Additional Files and Images

draft: Late draft - final version only available in print from the TU Wien library draft: Late draft - final version only available in print from the TU Wien library

Weblinks

No further information available.

BibTeX

@phdthesis{giegl-2007-pcg,
  title =      "Practical Computer Graphics Research, Game Development and
               the European GameTools Project",
  author =     "Markus Giegl",
  year =       "2007",
  abstract =   "Computer graphics today plays an important role in 3D
               computer and video game creation, since all the images which
               are presented to the player are based on principles and
               algorithms devised from computer graphics research.  One
               important aspect with regards to the believability and
               consecutively immersion of players within game worlds are
               shadows. Creating high-quality dynamic shadows in real-time
               is still a very open field of research. One real-time
               shadowing approach that is very appealing due to its
               apparent elegant simplicity is shadow mapping. The term
               "apparent" is used deliberately here, since the elegance of
               shadow mapping comes at a price, namely the discretization
               of the first-hit visibility problem to a grid storing
               occluder depth information with regards to the light source.
               In practice this often leads to shadowing artifacts, due to
               too little information being available to answer the
               visibility query with enough accuracy. In this thesis, a new
               family of shadow mapping algorithms is presented, which
               address the problem of the shadow map not containing enough
               information, This is done without requiring a shadow map
               grid too large to be stored in memory and which cannot be
               filled with depth information in real time. Based on a basic
               brute-force approach, two smart algorithms are presented,
               which work in a manner adaptive to the resolution
               requirements of the scene, speeding up the process by at
               least an order of magnitude.  Another problem with relevance
               to the practical application of computer graphics to games
               is how to render a large number of objects fast enough to
               give the player a high level of responsiveness (and thereby,
               again, immersion). Noticing that due to perspective
               shortening objects farther away from the player do not need
               as much detail as objects which are nearer, computer
               graphics has come up with the concept of level of detail
               (LOD), where farther away objects use simpler
               representations; if done correctly, this leads to a large
               increase in rendering speed. While LOD has been studied
               extensively in computer graphics, one problem of the LOD
               technique dominating practical application, discrete LOD,
               has been largely ignored: In discrete LOD the simpler
               representations of an object are, as the name implies,
               discrete, i.e. they exist independently of each other.
               Switching between representations of different complexity by
               simply switching the representation in use leads to a
               disruptive discontinuity in player perception. This thesis
               presents a practical algorithm, that addresses this problem.
                Linked to this research in the computer graphics fields of
               shadows and level of detail, is the case study of the
               GameTools Project, a Europe-wide project, funded by the
               European Union, which had as its agenda to bring the results
               of current computer graphics research done within the
               project to European game developing companies and companies
               from neighboring fields. The author holds the position of
               Community Manager within the GameTools Project, and presents
               an overview over the project and a more detailed view of his
               work as Community Manager.",
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Institute of Computer Graphics and Algorithms, Vienna
               University of Technology ",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2007/giegl-2007-pcg/",
}