Denoising your Monte Carlo renders: recent advances in image-space adaptive sampling and reconstruction

Monte Carlo integration is firmly established as the basis for most practical realistic image synthesis algorithms because of its flexibility and generality. However, the visual quality of rendered images often suffers from estimator variance, which appears as visually distracting noise. The current shift in the computer graphics industry towards Monte Carlo rendering has sparked renewed interest in effective, practical noise reduction techniques that are applicable to a wide range of rendering effects, and easily integrated into existing production pipelines. In this course, we survey recent advances in image-space adaptive sampling and reconstruction (filtering) algorithms for noise reduction, which have proven effective at reducing the computational cost of Monte Carlo techniques in practice. These techniques reduce variance by either controlling the sampling density over the image plane, and/or aggregating samples in a reconstruction step, possibly over large image regions in a way that preserves scene detail. To do this, they apply statistical techniques to sets of samples to drive the adaptive sampling and reconstruction process. In some cases, they use the statistical analysis to set the parameters for filtering. In others, they estimate the errors of several reconstruction filters, and select the best filter locally to minimize error. In this course, we aim to provide an overview for practitioners to assess these approaches, and for researchers to identify open research challenges and opportunities for future work. In an introduction, we will first situate image-space adaptive sampling and reconstruction in the larger context of variance reduction for Monte Carlo rendering, and discuss its conceptual advantages and potential drawbacks. In the next part, we will provide details on five specific state-of-the-art algorithms. We will provide visual and quantitative comparisons, and discuss advantages and disadvantages in terms of image quality, computational requirements, and ease of implementation and integration with existing renderers. We will conclude the course by pointing out how some of these techniques are proving useful in real-world applications. Finally, we will discuss directions for potential further improvements. This course brings together speakers that have made numerous contributions to image space adaptive rendering, which they presented at recent ACM SIGGRAPH, ACM SIGGRAPH Asia, and other conferences. The speakers bridge the gap between academia and industry, and they will be able to provide insights relevant to researchers, developers, and practioners alike.