BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035312Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221209T103000
DTEND;TZID=Asia/Seoul:20221209T120000
UID:siggraphasia_SIGGRAPH Asia 2022_sess173_papers_390@linklings.com
SUMMARY:Neural James-Stein Combiner for Unbiased and Biased Renderings
DESCRIPTION:Technical Communications, Technical Papers\n\nNeural James-Ste
 in Combiner for Unbiased and Biased Renderings\n\nGu, Iglesias-Guitian, Mo
 on\n\nUnbiased rendering algorithms such as path tracing produce accurate 
 images given an infinite number of samples, but in practice, the technique
 s often leave visually distracting artifacts (i.e., noise) in their render
 ed images due to a limited time budget. A favored approach for mitigating 
 the noise problem is applying learning-based denoisers to unbiased but noi
 sy rendered images and suppressing the noise while preserving image detail
 s. However, such denoising techniques typically introduce a systematic err
 or, i.e., the denoising bias, which does not decline as rapidly when incre
 asing the sample size, unlike the other type of error, i.e., variance. It 
 can technically lead to slow numerical convergence of the denoising techni
 ques. We propose a new combination framework built upon the James-Stein (J
 S) estimator, which merges a pair of unbiased and biased rendering images,
  e.g., a path-traced image and its denoised result. Unlike existing post-c
 orrection techniques for image denoising, our framework helps an input den
 oiser have lower errors than its unbiased input without relying on accurat
 e estimation of per-pixel denoising errors. We demonstrate that our framew
 ork based on the well-established JS theories allows us to improve the err
 or reduction rates of state-of-the-art learning-based denoisers more robus
 tly than recent post-denoisers.\n\nRegistration Category: FULL ACCESS, ON-
 DEMAND ACCESS\n\nLanguage: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_390&sess=sess17
 3
END:VEVENT
END:VCALENDAR