Factorized Point Based Global Illumination

Abstract

The Point-Based Global Illumination (PBGI) algorithm is composed of two major steps: a caching step and a multiview rasterization step. At caching time, a dense point-sampling of the scene is shaded and organized in a spatial hierarchy, with internal nodes approximating the radiance of their subtrees using spherical harmonics. At rasterization time, a microbuffer is instantiated at the unprojected position of each image pixel (receiver). Then, a view-adaptive level-of-detail of the scene is extracted in the form of a tree cut and rasterized in the receiver's microbuffer, solving for visibility using a local variant of the z-buffer. Finally, the pixel color is computed by convolving its filled microbuffer with the surface BRDF. This noise-free indirect lighting method is widely used in the industry and captures several critical lighting effects, including ambient occlusion, color bleeding, (indirect) soft-shadows and environment lighting. However, we observe a large redundancy in this algorithm, both in cuts and receivers'microbuffers, which stems from their relatively low resolution. In this paper, we propose an evolution of PBGI which exploits spatial coherence to reduce these redundant computations. Starting from a similarity-based variational clustering of the receivers, we compute a single tree cut and rasterize a single microbuffer for each cluster. This per-cluster microbuffer provides a faithful approximation of the incident radiance for distant nodes and is composited over a receiver-specific microbuffer rasterizing only the closest nodes of the cluster's cut. This factorized approach is easy to integrate in any existing PBGI implementation and offers a significant rendering speed-up for a negligible and controllable approximation error.

High Dynamic Range Point Clouds for Real‐Time Relighting

“…The method works iteratively with the input of the current GS iteration stored in the TF buffer by the previous GS iteration. The second observation by Wang et al [WHB∗ 13] points out that the BSH cuts gathered to fill the micro‐buffers end up to be very similar for receivers which are close in position and normal. We further develop this idea by traversing H using a mipmapped version of the G‐buffer proposing a new mipmapping operator for normal vectors (see Fig.…”

“…This algorithm is free from noise, accounts for long‐range indirect lighting and reproduces an important subset of GI effects. Its evolutions demonstrate high scalability for parallel architectures [REG∗ 09, HREB11] and out‐of‐core execution [Tab12], robustness to compression [BB12] and factorization [WHB∗ 13], the ability, to a certain extent, to cope with nondiffuse effects [WMB15], and scalability to render complex scenes from a very large number of viewpoints [KBLE19]. Our key observation is that a 3D scanning colored point cloud already provides the input of a PBGI tree avoiding the significant amount of work requested at caching time.…”

MegaViews: Scalable Many‐View Rendering With Concurrent Scene‐View Hierarchy Traversal

“…Image‐space clustering is also employed in point‐based global illumination (PBGI) [WHB*13], where tiles are repartitioned using a k‐means clustering. Assuming coherence of grouped pixels, a baseline cut through the scene hierarchy is established per tile.…”

Wavelet Point‐Based Global Illumination

“…Holländer et al [HREB11] further improved fine‐grained parallelism of the adaptive cut computation. A number of approaches have been proposed to improve the cut computation, including importance‐driven point projection based on an initial clustering [MW11], cut picking algorithm for HDR imaging [Tab12], and tree‐cut/microbuffers factorization based on spatial coherence [WHB*13]. The PBGI memory issue has been tackled with an out‐of‐core framework for PBGI, providing a cache‐coherent tree construction and traversal [KTO11], and with an in‐core solution which quantizes all tree nodes against a small set of representatives, learned on‐the‐fly [BB12].…”