Hybrid mesh-volume LoDs for all-scale pre-filtering of complex 3D assets

Abstract

International audienceWe address the problem of constructing appearance-preserving level of details (LoDs) of complex 3D models such as trees. We propose a hybrid method that combines the strengths of mesh and volume representations. Our main idea is to separate macroscopic (i.e. larger than the target spatial resolution) and microscopic (sub-resolution) surfaces at each scale and to treat them differently, because meshes are very efficient at representing macroscopic surfaces while sub-resolution geometry benefits from volumetric approximations. We introduce a new algorithm that detects the macroscopic surfaces of a mesh for a given resolution. We simplify these surfaces with edge collapses and we provide a method for pre-filtering their normal distributions and albedos. To approximate microscopic details, we use a heterogeneous microflake participating medium and we introduce a new artifact-free voxelization algorithm that preserves local occlusion. Thanks to our macroscopic surface analysis, our algorithm is fully automatic and it generates seamless LoDs at arbitrarily coarse resolutions for a wide range of 3D models

Real Time Multiscale Rendering of Dense Dynamic Stackings

“…Sequential point trees [DVS03] are an interesting evolution of QSplat [RL00] using an hybrid model, but it makes the same assumption that the point cloud is static. The inter-model transition was recently successfully addressed in the surface-to-volumetric context by Loubet et al [LN17]. Their setting is more general than ours but designed for off-line rendering and not tacking advantage -because not assuming -of self-similarity.…”

A non-exponential transmittance model for volumetric scene representations

“…The state-of-the-art method for general, appearance-preserving LoD is the hybrid approach proposed by Loubet and Neyret [2017]. It performs a heterogeneous simplification of a scene by labelling parts of a surface mesh to use either a geometric or a voxelized volumetric representation at every rendering scale.…”

“…On the other hand, scalable scene representations that are purely based on decimating triangle meshes cannot efficiently filter the appearance of fine structures, such as tree leaves, where accurately modeling of partial visibility is required. Ultimately, at farther distances, any geometric structure only achieves partial coverage at the sub-pixel level and can be more efficiently approximated using a volumetric representation [Loubet and Neyret 2017]. This approach has also become commonplace for modeling captured scenes [Lombardi et al 2019;Mildenhall et al 2020].…”

A new microflake model with microscopic self‐shadowing for accurate volume downsampling

“…Jakob et al [JAM*10] have generalized the radiative transfer equation to such anisotropic media and have introduced the physically‐based microflake model. This model has been used for rendering and pre‐filtering intricate shapes such as foliage [HDCD15, LN17] and fabrics [ZJMB11, ZWDR16]. In our work, we have focused on downsampling microflake volumes, and we believe that our approach can be adapted to other volume models, e.g .…”