DiffusionShadow: Diffusion-based Shadow Caching for Neural Volume Rendering

2026-09-24

AI Summary

This study addresses the prohibitive computational overhead of shadow evaluation in neural volumetric rendering and the storage bloat associated with independent implicit neural representations (INRs). We propose a diffusion-based shadow caching framework that compresses an extensive collection of shadow INRs into a single unified model. By leveraging the memorization capacity of diffusion models over densely precomputed illumination conditions, our method predicts network weights conditioned on light directions to encode shadow coefficient volumes. This formulation integrates seamlessly into standard renderers without requiring additional runtime sampling. Compared to conventional approaches, the proposed framework substantially accelerates rendering while eliminating storage redundancy. Furthermore, the generated shadows exhibit high fidelity, closely matching reference results.

Abstract

Implicit neural representations (INRs) have gained momentum in scientific visualization due to their compactness and scalability to large datasets, making them well suited for integration with direct volume rendering (DVR). However, real-time volume rendering of INR with advanced illumination effects, such as shadows, remains computationally expensive, as evaluating shadow terms via ray marching is costly. Alternatively, precomputing and storing shadows for many lighting directions is prohibitive in both memory and storage. To address this, we introduce a diffusion-based shadow caching framework that compresses a vast set of pre-calculated shadow INRs into a single diffusion model. Rather than focusing on generalizing to unseen directions, our method effectively memorizes and reconstructs a dense set of pre-trained lighting conditions on the fly. We first encode a collection of shadow coefficient volumes as shadow INRs, and then train a diffusion model conditioned on lighting direction to predict the corresponding shadow INR weights at inference time. This design integrates directly with standard INR renderers without additional runtime sampling. Experiments show that our approach achieves faster rendering than traditional methods while bypassing the massive storage bloat of independent INRs, producing shadows that closely match most of the reference results.

Problem

• Implicit Neural Representations
• Volume Rendering
• Shadow Caching
• Real-time Rendering
• Computational Cost

Innovation

• Diffusion Model
• Neural Volume Rendering
• Shadow Caching
• Implicit Neural Representations
• Direct Volume Rendering