KIT | KIT-Bibliothek | Impressum | Datenschutz

Detecting Bias in Monte Carlo Renderers using Welch’s t-test

Jung, Alisa; Hanika, Johannes; Dachsbacher, Carsten

Abstract (englisch):
When checking the implementation of a new renderer, one usually compares the output to that of a reference implementation. However, such tests require a large number of samples to be reliable, and sometimes they are unable to reveal very subtle differences that are caused by bias, but overshadowed by random noise. We propose using Welch’s t-test, a statistical test that reliably finds small bias even at low sample counts. Welch’s t-test is an established method in statistics to determine if two sample sets have the same underlying mean, based on sample statistics. We adapt it to test whether two renderers converge to the same image, i.e., the same mean per pixel or pixel region. We also present two strategies for visualizing and analyzing the test’s results, assisting us in localizing especially problematic image regions and detecting biased implementations with high confidence at low sample counts both for the reference and tested implementation.

Open Access Logo


Verlagsausgabe §
DOI: 10.5445/IR/1000120188
Veröffentlicht am 16.06.2020
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Visualisierung und Datenanalyse (IVD)
Publikationstyp Zeitschriftenaufsatz
Publikationsdatum 13.06.2020
Sprache Englisch
Identifikator ISSN: 2331-7418
KITopen-ID: 1000120188
Erschienen in Journal of Computer Graphics Techniques (JCGT)
Band 9
Heft 2
Seiten 1–25
Schlagwörter Rendering, Bias, Welch, t-test
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page