Projects per year

### Abstract

Most global illumination algorithms today solve the light transport problem
using Monte Carlo ray tracing. These algorithms are capable of producing
photo-realistic imagery and in addition have few limitations with respect to the
kind of input (geometry, re
flection models, etc.) they support. The downside to
using these algorithms is that they can be slow to converge. Due to the nature of
Monte Carlo methods, the results are random variables subject to variance. This
manifests itself as noise in the images, which can only be reduced by generating
more samples.
The reason these methods are slow is because of a lack of eeffective methods of
importance sampling. Most global illumination algorithms are based on local
path sampling, which is essentially a recipe for constructing random walks.
Using this procedure paths are built based on information given explicitly as
part of scene description, such as the location of the light sources or cameras,
or the re
flection models at each point.
In this work we explore new methods of importance sampling paths. Our idea is
to analyze the scene before rendering and compute various statistics that we use
to improve importance sampling. The first of these are adjoint measurements,
which are the solution to the adjoint light transport problem. The second is a
representation of the distribution of radiance and importance in the scene. We
also derive a new method of particle sampling, which is advantageous compared
to existing methods. Together we call the resulting algorithm englightened local
path sampling and demonstrate how the algorithm improves efficiency in some
hard scenes.

Original language | English |
---|

Place of Publication | Kgs. Lyngby, Denmark |
---|---|

Publisher | Technical University of Denmark |

Number of pages | 188 |

Publication status | Published - 2011 |

Series | IMM-PHD-2010-240 |
---|---|

ISSN | 0909-3192 |

## Projects

- 1 Finished

## Lighting Design and Real-time Global Illumination

Kristensen, A. W., Bærentzen, J. A., Henriksen, K., Myszkowski, K. & Christensen, N. J.

01/02/2006 → 30/03/2011

Project: PhD

## Cite this

Kristensen, A. W. (2011).

*Efficient Unbiased Rendering using Enlightened Local Path Sampling*. Technical University of Denmark. IMM-PHD-2010-240