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 (1)\, B Block\, Level 5 DTSTART;TZID=Asia/Tokyo:20241204T113100 DTEND;TZID=Asia/Tokyo:20241204T114300 UID:siggraphasia_SIGGRAPH Asia 2024_sess112_papers_389@linklings.com SUMMARY:3D Reconstruction with Fast Dipole Sums DESCRIPTION:Technical Papers\n\nHanyu Chen, Bailey Miller, and Ioannis Gki oulekas (Carnegie Mellon University)\n\nWe introduce a method for high-qua lity 3D reconstruction from multi-view images. Our method uses a new point -based representation, the regularized dipole sum, which generalizes the w inding number to allow for interpolation of per-point attributes in point clouds with noisy or outlier points. Using regularized dipole sums, we rep resent implicit geometry and radiance fields as per-point attributes of a dense point cloud, which we initialize from structure from motion. We addi tionally derive Barnes-Hut fast summation schemes for accelerated forward and adjoint dipole sum queries. These queries facilitate the use of ray tr acing to efficiently and differentiably render images with our point-based representations, and thus update their point attributes to optimize scene geometry and appearance. We evaluate our method in inverse rendering appl ications against state-of-the-art alternatives, based on ray tracing of ne ural representations or rasterization of Gaussian point-based representati ons. Our method significantly improves 3D reconstruction quality and robus tness at equal runtimes, while also supporting more general rendering meth ods such as shadow rays for direct illumination.\n\nRegistration Category: Full Access, Full Access Supporter\n\nLanguage Format: English Language\n \nSession Chair: Michael Wimmer (TU Wien) URL:https://asia.siggraph.org/2024/program/?id=papers_389&sess=sess112 END:VEVENT END:VCALENDAR