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:20250110T023309Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241203T154300
DTEND;TZID=Asia/Tokyo:20241203T155400
UID:siggraphasia_SIGGRAPH Asia 2024_sess107_papers_826@linklings.com
SUMMARY:Local Gaussian Density Mixtures for Unstructured Lumigraph Renderi
 ng
DESCRIPTION:Technical Papers\n\nXiuchao Wu (State Key Laboratory of CAD&CG
 , Zhejiang University); Jiamin Xu (Hangzhou Dianzi Univeristy); Chi Wang (
 State Key Laboratory of CAD&CG, Zhejiang University); Yifan Peng (Universi
 ty of Hong Kong); Qixing Huang (University of Texas at Austin); James Tomp
 kin (Brown University); and Weiwei Xu (State Key Laboratory of CAD&CG, Zhe
 jiang University)\n\nTo improve novel-view synthesis of curved surface ref
 lections and refractions, we revisit local geometry-guided ray interpolati
 on techniques with modern differentiable rendering and optimization.\nIn c
 ontrast to depth or mesh geometries, our approach uses a local or per-view
  density represented as Gaussian mixtures along each ray. \nTo synthesize 
 novel views, we warp and fuse local volumes, then alpha-composite using in
 put photograph ray colors from a small set of neighboring images. \nFor fu
 sion, we use a neural blending weight from a shallow MLP. \nWe optimize th
 e local Gaussian density mixtures using both a reconstruction loss and a c
 onsistency loss. \nThe consistency loss, based on per-ray KL-divergence, e
 ncourages more accurate geometry reconstruction. \nOn scenes with complex 
 reflections captured in our LGDM dataset, experimental results show that o
 ur method outperforms state-of-the-art novel-view synthesis methods by 12.
 2\%--37.1\% in PSNR, thanks to its ability to maintain sharper view-depend
 ent appearance.\n\nRegistration Category: Full Access, Full Access Support
 er\n\nLanguage Format: English Language\n\nSession Chair: Hongzhi Wu (Zhej
 iang University; State Key Laboratory of CAD&CG, Zhejiang University)
URL:https://asia.siggraph.org/2024/program/?id=papers_826&sess=sess107
END:VEVENT
END:VCALENDAR