On-line Learning of Parametric Mixture Models for Light Transport Simulation

被引:5
|
作者
Vorba, Jiri [1 ]
Karlik, Ondrej [1 ]
Sik, Martin [1 ]
Ritschel, Tobias [2 ]
Krivanek, Jaroslav [1 ]
机构
[1] Charles Univ Prague, Prague, Czech Republic
[2] MPI Informat, Saarbrucken, Germany
来源
ACM TRANSACTIONS ON GRAPHICS | 2014年 / 33卷 / 04期
关键词
light transport simulation; importance sampling; parametric density estimation; on-line expectation maximization;
D O I
10.1145/2601097.2601203
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
Monte Carlo techniques for light transport simulation rely on importance sampling when constructing light transport paths. Previous work has shown that suitable sampling distributions can be recovered from particles distributed in the scene prior to rendering. We propose to represent the distributions by a parametric mixture model trained in an on-line (i.e. progressive) manner from a potentially infinite stream of particles. This enables recovering good sampling distributions in scenes with complex lighting, where the necessary number of particles may exceed available memory. Using these distributions for sampling scattering directions and light emission significantly improves the performance of state-of-the-art light transport simulation algorithms when dealing with complex lighting.
引用
收藏
页数:11
相关论文
共 50 条
  • [21] ON-LINE SIMULATION EXPERIMENTS
    ARMSTRONG, JS
    SIMULATION, 1972, 18 (04) : 127 - +
  • [22] Intelligent control using multiple models based on on-line learning
    Junyong Zhai
    Shumin Fei
    Feipeng Da
    Journal of Control Theory and Applications, 2006, 4 (4): : 397 - 401
  • [23] On-line regression algorithms for learning mechanical models of robots: A survey
    Sigaud, Olivier
    Salauen, Camille
    Padois, Vincent
    ROBOTICS AND AUTONOMOUS SYSTEMS, 2011, 59 (12) : 1115 - 1129
  • [25] ON-LINE LEARNING DYNAMIC MODELS FOR NERVE DETECTION IN ULTRASOUND VIDEOS
    Hadjerci, Oussama
    Hafiane, Adel
    Vieyres, Pierre
    Conte, Donatello
    Makris, Pascal
    Delbos, Alain
    2016 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2016, : 131 - 135
  • [26] On-line learning of language models with word error probability distributions
    Gretter, R
    Riccardi, G
    2001 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, VOLS I-VI, PROCEEDINGS: VOL I: SPEECH PROCESSING 1; VOL II: SPEECH PROCESSING 2 IND TECHNOL TRACK DESIGN & IMPLEMENTATION OF SIGNAL PROCESSING SYSTEMS NEURALNETWORKS FOR SIGNAL PROCESSING; VOL III: IMAGE & MULTIDIMENSIONAL SIGNAL PROCESSING MULTIMEDIA SIGNAL PROCESSING - VOL IV: SIGNAL PROCESSING FOR COMMUNICATIONS; VOL V: SIGNAL PROCESSING EDUCATION SENSOR ARRAY & MULTICHANNEL SIGNAL PROCESSING AUDIO & ELECTROACOUSTICS; VOL VI: SIGNAL PROCESSING THEORY & METHODS STUDENT FORUM, 2001, : 557 - 560
  • [27] Enhancing Robustness of On-Line Learning Models on Highly Noisy Data
    Zhao, Zilong
    Birke, Robert
    Han, Rui
    Robu, Bogdan
    Bouchenak, Sara
    Ben Mokhtar, Sonia
    Chen, Lydia Y.
    IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING, 2021, 18 (05) : 2177 - 2192
  • [28] Structural results about on-line learning models with and without queries
    Auer, P
    Long, PM
    MACHINE LEARNING, 1999, 36 (03) : 147 - 181
  • [29] Structural Results About On-line Learning Models With and Without Queries
    Peter Auer
    Philip M. Long
    Machine Learning, 1999, 36 : 147 - 181
  • [30] On-line Evolutionary Exponential Family Mixture
    Zhang, Jianwen
    Song, Yangqiu
    Chen, Gang
    Zhang, Changshui
    21ST INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE (IJCAI-09), PROCEEDINGS, 2009, : 1610 - 1615