A Survey on Gradient-Domain Rendering

Binh-Son Hua

The University of Tokyo

Adrien Gruson

McGill University

Victor Petitjean

Delft University of Technology

Matthias Zwicker

University of Maryland, College Park

Derek Nowrouzezahrai

McGill University

Elmar Eisemann

Delft University of Technology

Toshiya Hachisuka

The University of Tokyo

Monte Carlo methods for physically-based light transport simulation are broadly adopted in film productions and animation/visual effects industries. Monte Carlo methods, however, often result in noisy images and have slow convergence. As such, improving the convergence of Monte Carlo rendering remains an important open problem. Gradient-domain light transport is a recent family of techniques that can accelerate Monte Carlo rendering by up to an order of magnitude, leveraging a gradient-based estimation and a reformulation of the rendering problem as an image reconstruction. This state of the art report comprehensively elaborates the fundamentals of gradient-domain rendering, as well as the specifics behind gradient-domain uni- and bidirectional path tracing and gradient-domain photon density estimation. We also discuss various image reconstruction schemes which are important components of gradient-domain rendering. Finally, we benchmark various gradient-domain techniques against state-of-the-art denoising methods before discussing open problems.


Tuesday, May 7, 2019
9:00am - 10:30am
Centro Congressi di Genova
Porto Antico


Code & Data



9:00am: Introduction

Welcome speech and a brief introduction about the report, its objectives and schedule.

9:05am: Fundamentals of gradient-domain rendering

The key ideas of gradient-domain rendering is introduced. Particularly, the audience will learn the concepts of the modified rendering equation for gradient-domain rendering, the shift mapping function and its Jacobian. A simple gradient-domain path tracer will be introduced.

9:20am: Gradient-domain light transport algorithms

We continue with the discussion of gradient-domain rendering in more details. Advanced shift mappings, progressive photon mapping, bidirectional path tracing in the gradient domain are introduced.

9:50am: Image reconstruction

We then discuss how to reconstruct images from gradients robustly.

10:00am: Advanced topics in gradient-domain rendering

We discuss a collection of many advanced techniques in gradient-domain rendering such as volumetric rendering, temporal rendering, adaptive rendering, Metropolis light transport, path reuse, and spectral rendering.

10:20pm: Conclusions

Finally, we discuss several potential ideas for future work.

Gradient-Domain Rendering Papers

Deep Convolutional Reconstruction for Gradient-Domain Rendering (SIGGRAPH 2019)

Gradient Outlier Removal for Gradient-Domain Path Tracing (EG 2019)

A Survey on Gradient-Domain Rendering (EG 2019 STARs)

Feature Generation for Adaptive Gradient-Domain Path Tracing (Pacific Graphics 2018)

Gradient-Domain Volumetric Photon Density Estimation (SIGGRAPH 2018)

Spectral Gradient Sampling for Path Tracing (EGSR 2018)

Gradient-Domain Path Reusing (SIGGRAPH Asia 2017)

Gradient-Domain Vertex Connection and Merging (EGSR 2017)

Gradient-Domain Photon Density Estimation (EG 2017)

Temporal Gradient-Domain Path Tracing (SIGGRAPH Asia 2016)

Image-Space Control Variates for Rendering (SIGGRAPH Asia 2016)

Regularizing Image Reconstruction for Gradient-Domain Rendering with Feature Patches (EG 2016)

Gradient-Domain Bidirectional Path Tracing (EGSR 2015)

Gradient-Domain Path Tracing (SIGGRAPH 2015)

Improved Sampling for Gradient-Domain Metropolis Light Transport (SIGGRAPH Asia 2014)

Gradient-Domain Metropolis Light Transport (SIGGRAPH 2013)


Binh-Son Hua

The University of Tokyo

Adrien Gruson

McGill University