Reproduce “Monte Carlo Rendering to Diffusion Curves with Differential BEM” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ryusuke Sugimoto, Christopher Batty, Siddhartha Chaudhuri, Iliyan Georgiev, Toshiya Hachisuka, Kevin Wampler, Michal Lukáč
A method directly converts noisy Monte Carlo renderer outputs into diffusion curves using a novel differential Boundary Element Method (BEM) and Levenberg–Marquardt optimization. This approach bypasses the intermediate rasterization stage, demonstrating robustness to input noise and complex geometric scenarios.
Key insights
- The BEM's fixed system matrix structure, independent of optimizable handle parameters, allows for efficient iterative optimization by pre-factorizing the matrix once.
- Direct optimization from noisy Monte Carlo samples is feasible and robust, eliminating the need for a clean, intermediate raster image and its associated computational and robustness issues.
- Levenberg–Marquardt optimization provides stable and faster convergence for this inverse problem compared to stochastic gradient descent methods like AdamW, effectively handling varying parameter scales.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.