Authors
Beibei Wang, Wenhua Jin, Miloš Hašan, Ling-Qi Yan
Nankai University; Nanjing University of Science and Technology; Adobe Research; University of California, Santa Barbara
Portals
Summary
In this indoor scene, we model the window shade using our SpongeCake model, with a two-layer configuration: a specular fiber-like microflake layer on the inside and a rougher fiber-like microflake layer at bottom. Three different light settings are shown: exterior sunlight gives a diffuse (but non-Lambertian) transmission effect, while interior lighting leads to a very different specular fabric sheen effect; finally, we combine the two lighting configurations. Our layered model is able to design these kinds of appearances easily, while offering fast analytic evaluation, including an effective multiple scattering approximation; there is a lack of comparable analytic material models with similar benefits.
Abstract
In this paper, we propose SpongeCake: a layered BSDF model where each layer is a volumetric scattering medium, defined using microflake or other phase functions. We omit any reflecting and refracting interfaces between the layers. The first advantage of this formulation is that an exact and analytic solution for single scattering, regardless of the number of volumetric layers, can be derived. We propose to approximate multiple scattering by an additional single-scattering lobe with modified parameters and a Lambertian lobe. We use a parameter mapping neural network to find the parameters of the newly added lobes to closely approximate the multiple scattering effect. Despite the absence of layer interfaces, we demonstrate that many common material effects can be achieved with layers of SGGX microflake and other volumes with appropriate parameters. A normal mapping effect can also be achieved through mapping of microflake orientations, which avoids artifacts common in standard normal maps. Thanks to the analytical formulation, our model is very fast to evaluate and sample. Through various parameter settings, our model is able to handle many types of materials, like plastics, wood, cloth, etc., opening a number of practical applications.
Contribution
- Definition of the SpongeCake model, and demonstration of the wide range of appearances it can achieve: plastics, metals, fabrics, plant leaves, and polished hardwood with secondary specular highlights
- An analytic single-scattering solution, including reflection and transmission of multiple layers, together with an importance sampling scheme. This is a generalization of many similar previous single-scattering derivations
- An accurate analytic approximation to multiple scattering, based on the same analytic form as the single-scattering BSDF, with modified parameters predicted by a neural network
Related Works
Layered material models; Microflake models; Microfacet multiple scattering; Single scattering in volumetric layers; Neural networks for material appearance
Overview
Since our SpongeCake model consists of only volumetric layers, all capabilities of this model can be evaluated with Monte Carlo random walk simulation of light transport, which we use as ground truth. However, our goal is to achieve much more efficient analytic evaluation and sampling. We first derive an analytic single scattering formula for our BSDF model (Sec. 4). We show that for a microflake phase function with flake distribution ?, the resulting BSDF has a form similar (but not identical) to a standard microfacet BSDF [Walter et al. 2007], with ? becoming analogous to the microfacet normal distribution function. We discuss extensions to transmission and multiple layers. We also discuss the optional unscattered delta transmission component, which we may or may not choose to include, depending on the desired appearance. While single scattering is already sufficient for a variety of appearances, multiple scattering is important for some materials. We approximate multiple scattering by noting that the ground truth multiple-scattering component of the BSDF often looks similar to the single-scattering component, except with modified parameters; an observation already made by others [Heitz et al. 2016]. We propose a multiple-to-single mapping approach, finding the parameters via a small neural network (Sec. 5). Thus, the multiple scattering term has the same form and the same evaluation cost as the single-scattering term.