Parametric Integration with Neural Integral Operators

2025-07-23

AI Summary

In real-time rendering, noise in lighting transport simulation remains a persistent challenge due to strict sampling budget constraints. Conventional denoisers operate post-shading, struggling to simultaneously ensure material-agnostic generalization and high-fidelity noise suppression. This paper introduces Material-Agnostic Denoising (MAD), a novel paradigm that models the unoccluded lighting integral *before* shading. Its core innovation is a parameterized neural integration operator that treats light transport as a learnable continuous function mapping—jointly generalizing across scene geometry, light configurations, and BRDF parameters. MAD requires only single-frame supervision, enables efficient training, and natively supports temporal antialiasing and integration into existing denoising pipelines. Experiments demonstrate substantial improvements in image fidelity and material consistency while maintaining real-time performance, providing a seamless, robust denoising solution for physically based rendering workflows.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Real-time rendering imposes strict limitations on the sampling budget for light transport simulation, often resulting in noisy images. However, denoisers have demonstrated that it is possible to produce noise-free images through filtering. We enhance image quality by removing noise before material shading, rather than filtering already shaded noisy images. This approach allows for material-agnostic denoising (MAD) and leverages machine learning by approximating the light transport integral operator with a neural network, effectively performing parametric integration with neural operators. Our method operates in real-time, requires data from only a single frame, seamlessly integrates with existing denoisers and temporal anti-aliasing techniques, and is efficient to train. Additionally, it is straightforward to incorporate with physically based rendering algorithms.

Problem

• Reducing noise in real-time rendering before material shading
• Achieving material-agnostic denoising via neural integral operators
• Integrating with existing denoisers and temporal anti-aliasing techniques

Innovation

• Material-agnostic denoising before shading
• Neural network approximates light transport
• Real-time single-frame data processing