BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030657Z
LOCATION:Meeting Room S426+S427\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251216T110100
DTEND;TZID=Asia/Hong_Kong:20251216T111200
UID:siggraphasia_SIGGRAPH Asia 2025_sess118_papers_1592@linklings.com
SUMMARY:DSCombiner: Double Shrinkage for Combining Biased and Unbiased Mon
 te Carlo Renderings
DESCRIPTION:Chenxi Zhou, Keheng Xu, Mufan Guo, Xianhao Yu, Zhimin Fan, Gui
 huan Feng, Yanwen Guo, and Jie Guo (Nanjing University)\n\nMonte Carlo ren
 dering often faces a dilemma, namely, whether to choose an unbiased estima
 tor or a biased one. Although different integrators have been developed to
  address various scenarios, no single method can effectively manage all si
 tuations. Thus, finding a good approach to combine different integrators h
 as always been a topic that warrants exploration. \n\nThis work proposes D
 SCombiner, a new shrinkage estimator that flexibly combines unbiased and b
 iased estimators (typically generated by different integrators) in image s
 pace into a single estimating procedure, strategically utilizing the stren
 gths of different integrators while minimizing their weaknesses. DSCombine
 r overcomes the limitation of single shrinkage combiners by introducing a 
 two-step shrinkage towards a noise-free radiance prior. We derive optimal 
 shrinkage factors for the two steps within a hierarchical Bayesian framewo
 rk, and provide a deep learning-based method to improve the results. Compr
 ehensive qualitative and quantitative validations across diverse scenes de
 monstrate visible improvements in image quality, as compared with previous
  image-space and path-space combiners.\n\nRegistration Category: Full Acce
 ss, Full Access Supporter\n\nSession Chair: Oliver Deussen (University of 
 Konstanz)\n\n
END:VEVENT
END:VCALENDAR