Modular Radiance Transfer

Abstract

Figure 1: Indirect light computed in reduced subspaces for a cave with 19 blocks and 4 lights. We derive low-dimensional transport operators, on simple proxy shapes, that are warped and combined at run-time, at > 475 FPS on high-end GPUs and > 45 FPS on mobile platforms, and can model indirect light at surfaces (with detailed normal variation) and within volumes of large-scale scene geometry. AbstractMany rendering algorithms willingly sacrifice accuracy, favoring plausible shading with high-performance. Modular Radiance Transfer (MRT) models coarse-scale, distant indirect lighting effects in scene geometry that scales from high-end GPUs to low-end mobile platforms. MRT eliminates scene-dependent precomputation by storing compact transport on simple shapes, akin to bounce cards used in film production. These shapes' modular transport can be instanced, warped and connected on-the-fly to yield approximate light transport in large scenes. We introduce a prior on incident lighting distributions and perform all computations in low-dimensional subspaces. An implicit lighting environment induced from the low-rank approximations is in turn used to model secondary effects, such as volumetric transport variation, higher-order irradiance, and transport through lightfields. MRT is a new approach to precomputed lighting that uses a novel low-dimensional subspace simulation of light transport to uniquely balance the need for high-performance and portable solutions, low memory usage, and fast authoring iteration.

Reduced Aggregate Scattering Operators for Path Tracing

“…wavelets [NRH03] or radial basis functions [TS06, WRG*09]). Additionally, we train our basis using a prior distribution of expected direct and indirect lighting signals, as proposed by Loos et al [LAM*11].…”

“…Zhou et al [ZHL*05] or Sun and Mukherjee [SM06]), where light transport representations are precomputed on some (potentially simplified) subset of the scene. We are inspired by Loos et al's “modular radiance transfer” [LAM*11, LNJS12] which precomputes compact transport operators for simple geometric shapes and then dynamically warps them to the actual scene, recomposing the total light transport on‐the‐fly. Zhao et al [ZHRB13] extended these ideas to volumetric transport in cloth expressed using exemplar voxel blocks.…”

“… we employ the lighting prior proposed by Loos et al [LAM*11] to adapt the bases to the expected lighting scenarios; we create separate bases for the direct and transported indirect light on/in an asset, each adapted to their expected signals; and, we compute a reduced transport matrix that directly computes all bounces of light—instead of relying on iterated multiplication of a one‐bounce operator—allowing us to account for arbitrary, all‐frequency internal scattering. …”

“…Several hybrid techniques isolate individual indirect effects and tailor specialized solutions, leveraging e.g., repetitive structure [ZHRB13], simplified transport configurations [LAM*11, ZW06, JMLH01], or statistical representations [MWM07]. Motivated by these approaches, we also segment costly effects; however, instead of devising solutions specialized to individual effects, we propose a representation that encodes all internal transport within an asset ; here an “asset” refers to a collection of surfaces and/or participating media.…”

Aural Proxies and Directionally-Varying Reverberation for Interactive Sound Propagation in Virtual Environments

“…In addition to accelerating the rendering of large, complex scenes, simplified proxies can also be used to significantly accelerate the computation of complex, computationally-intensive phenomena such as global illumination. Modular radiance transfer [16] decribes a method for replacing complex geometry with cubical proxies, which are then used to compute indirect illumination in response to direct illumination computed for the original, complex geometry. Our method shares some similarities with these previous methods, in that it fits a 6-sided cubical proxy to the local geometry around the listener, and uses this proxy to compute higher-reflections in response to first-order reflections computed using the original geometry.…”