Information

Abstract

This thesis focuses on improvements for an interactive lighting design approach that utilizes GPU-accelerated ray tracing and a view-independent global illumination solver. Our goal is to enable automated lighting design for a set of user-specified illumination targets in 3D scenes. Current solvers are highly effective but still have some limitations. For instance, they rely on an initial number of light sources and their respective placements in a given 3D scene and this can result in insufficient solutions when there are more target spots than provided light sources. On the other hand, if there are more light sources than needed, the resulting solution can be sub-optimal, leading to superimposed lights that can negatively impact performance and increase computational cost. In response to the limitations, we investigate several strategies for increasing the effectiveness and efficiency of the optimization algorithm by developing a dynamic light source generation approach that programmatically inserts and removes lights in the 3D scene to achieve a more refined light placement. In our results, we show that our specialized optimization approach, yields improved lighting solutions compared to established algorithms. Moreover, we also implement a light source merging technique to address the issue of light sources with overlapping areas of influence. By formulating conditions on intensity and proximity and then applying linear interpolation, we can combine overlapping light sources in a way that minimizes performance impact and computational cost. We also take measures to remove lights with a small illumination contribution to the scene during the optimization process. Evidence from our study suggests that our approach of expanding the solution space and improving the light source placement achieves superior lighting solutions for any given scene.

Additional Files and Images

Weblinks

No further information available.

BibTeX

@bachelorsthesis{Koeppl-2023-DLO,
  title =      "Gradient-based Light Optimization with Variable Light Count:
               Dynamic Generation and Merging of Light Sources",
  author =     "David K\"{o}ppl",
  year =       "2023",
  abstract =   "This thesis focuses on improvements for an interactive
               lighting design approach that utilizes GPU-accelerated ray
               tracing and a view-independent global illumination solver.
               Our goal is to enable automated lighting design for a set of
               user-specified illumination targets in 3D scenes. Current
               solvers are highly effective but still have some
               limitations. For instance, they rely on an initial number of
               light sources and their respective placements in a given 3D
               scene and this can result in insufficient solutions when
               there are more target spots than provided light sources. On
               the other hand, if there are more light sources than needed,
               the resulting solution can be sub-optimal, leading to
               superimposed lights that can negatively impact performance
               and increase computational cost. In response to the
               limitations, we investigate several strategies for
               increasing the effectiveness and efficiency of the
               optimization algorithm by developing a dynamic light source
               generation approach that programmatically inserts and
               removes lights in the 3D scene to achieve a more refined
               light placement. In our results, we show that our
               specialized optimization approach, yields improved lighting
               solutions compared to established algorithms. Moreover, we
               also implement a light source merging technique to address
               the issue of light sources with overlapping areas of
               influence. By formulating conditions on intensity and
               proximity and then applying linear interpolation, we can
               combine overlapping light sources in a way that minimizes
               performance impact and computational cost. We also take
               measures to remove lights with a small illumination
               contribution to the scene during the optimization process.
               Evidence from our study suggests that our approach of
               expanding the solution space and improving the light source
               placement achieves superior lighting solutions for any given
               scene.",
  month =      dec,
  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 ",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2023/Koeppl-2023-DLO/",
}