BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250110T023312Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241205T090000
DTEND;TZID=Asia/Tokyo:20241205T091400
UID:siggraphasia_SIGGRAPH Asia 2024_sess125_papers_319@linklings.com
SUMMARY:A Statistical Approach to Monte Carlo Denoising
DESCRIPTION:Technical Papers\n\nHiroyuki Sakai and Christian Freude (Techn
 ical University of Vienna), Thomas Auzinger (Institute of Science and Tech
 nology Austria), and David Hahn and Michael Wimmer (Technical University o
 f Vienna)\n\nThe stochastic nature of modern Monte Carlo (MC) rendering me
 thods inevitably produces noise in rendered images for a practical number 
 of samples per pixel. The problem of denoising these images has been widel
 y studied, with most recent methods relying on data-driven, pretrained neu
 ral networks. In contrast, in this paper we propose a statistical approach
  to the denoising problem, treating each pixel as a random variable and re
 asoning about its distribution. Considering a pixel of the noisy rendered 
 image, we formulate fast pair-wise statistical tests—based on online estim
 ators—to decide which of the nearby pixels to exclude from the denoising f
 ilter. We show that for symmetric pixel weights and normally distributed s
 amples, the classical Welch t-test is optimal in terms of mean squared err
 or. We then show how to extend this result to handle non-normal distributi
 ons, using more recent confidence-interval formulations in combination wit
 h the Box-Cox transformation. Our results show that our statistical denois
 ing approach matches the performance of state-of-the-art neural image deno
 ising without having to resort to any computation-intensive pretraining. F
 urthermore, our approach easily generalizes to other quantities besides pi
 xel intensity, which we demonstrate by showing additional applications to 
 Russian roulette path termination and multiple importance sampling.\n\nReg
 istration Category: Full Access, Full Access Supporter\n\nLanguage Format:
  English Language\n\nSession Chair: Wenzel Jakob (École Polytechnique Fédé
 rale de Lausanne)
URL:https://asia.siggraph.org/2024/program/?id=papers_319&sess=sess125
END:VEVENT
END:VCALENDAR