Volumetric Inverse Rendering via Neural Radiative Transfer

2026-07-15

AI Summary

This work proposes a neural field–based volumetric inverse rendering method that jointly recovers spatially varying, spectrally resolved optical properties of participating media—including scattering, absorption, and phase functions—from multi-view images, while explicitly modeling global illumination effects. The approach represents both the medium’s optical parameters and the full light field within a unified neural radiance field framework. By jointly optimizing the residual of the local differential form of the radiative transfer equation and the primary view ray volume rendering objective, it achieves physically complete and efficient reconstruction without resorting to complex differentiable path tracing. This study presents the first integration of a full light transport physics model with general-purpose neural optimization, enabling high-fidelity, physically consistent reconstruction and synthesis of participating media.

Abstract

Volumetric inverse rendering seeks to recover the optical properties of participating media from images. Existing approaches either rely on differentiable stochastic light transport simulation, which require substantial algorithmic effort, or use simplified models that fail to capture global illumination. We propose a formulation that reconciles physically complete light transport with general-purpose neural optimization. The optical properties of the medium and the full light field are represented as neural fields and estimated through a joint optimization process. Global illumination is enforced via a residual objective derived from the Radiative Transfer Equation in local differential form, complemented by a volume rendering term along primary viewing rays to mitigate \rev{low-frequency} bias. We demonstrate reconstruction of spatially varying, color-resolved scattering, absorption, and phase function parameters from multi-view images. Beyond reconstruction, the same framework supports learning generative models of participating media with physical optical properties under global illumination.

Problem

• volumetric inverse rendering
• participating media
• optical properties
• global illumination
• radiative transfer

Innovation

• Neural Radiative Transfer
• Volumetric Inverse Rendering
• Global Illumination
• Neural Fields
• Radiative Transfer Equation