Thèse soutenue

Illumination cohérente pour la simulation d'éclairage

FR  |  
EN
Auteur / Autrice : Václav Gassenbauer
Direction : Kadi BouatouchJaroslav Křivánek
Type : Thèse de doctorat
Discipline(s) : Informatique
Date : Soutenance en 2011
Etablissement(s) : Rennes 1 en cotutelle avec Université technique tchèque
Ecole(s) doctorale(s) : École doctorale Mathématiques, télécommunications, informatique, signal, systèmes, électronique (Rennes)
Partenaire(s) de recherche : autre partenaire : Université européenne de Bretagne (2007-2016)

Mots clés

FR

Mots clés contrôlés

Résumé

FR  |  
EN

Simulation of light transport in a scene is an essential task in realistic image synthesis. However, an accurate simulation of light as it bounces in the scene is time consuming. It has been shown that a key to speeding up light transport simulation algorithms is to take advantage of the high degree of spatial, angular, and temporal coherence. In this thesis we make three contributions in this area. First, we propose spatial directional radiance caching (SDRC) for accelerating the light transport simulation in scenes with glossy surfaces. The SDRC algorithm takes advantage of the smoothness of shading on glossy surfaces by interpolating the indirect illumination from a set of sparsely distributed radiance samples that are both spatially and directionally close. In the next part of the thesis, we propose an efficient and accurate local principal component analysis (LPCA) algorithm for dimensionality reduction and data compression of large data sets. To achieve efficiency our new algorithm, called SortCluster-LPCA, passes various information from previous iteration to the next. Improved accuracy is achieved through better initial seeding of cluster centroids in LPCA. Finally, we describe a work in progress focusing on the development of an algorithm for interactive relighting of animation sequences with indirect illumination. We formulate the relighting problem as a large 3D array expressing light propagation in a scene over multiple frames. We suggest an adaptive algorithm to make the pre-computation tractable exploiting coherence in light transport.