Real-time Level-of-Detail Strand-based Hair Rendering

TL;DR

This work introduces an innovative real-time framework for strand-based hair rendering that ensures seamless transitions between different levels of detail (LOD) while maintaining a consistent hair appearance.

Abstract

Strand-based hair rendering has become increasingly popular in production for its realistic appearance. However, the prevailing level-of-detail solution employing hair cards for distant hair models introduces a significant discontinuity in dynamics and appearance during the transition from strands to cards. We introduce an innovative real-time framework for strand-based hair rendering that ensures seamless transitions between different levels of detail (LOD) while maintaining a consistent hair appearance. Our method uses elliptical thick hairs that contain multiple hair strands at each LOD to maintain the shapes of hair clusters. In addition to geometric fitting, we formulate an elliptical Bidirectional Curve Scattering Distribution Functions (BCSDF) model for a thick hair, accurately capturing single scattering and multiple scattering within the hair cluster, accommodating a spectrum from sparse to dense hair distributions. Our framework, tested on various hairstyles with dynamics as well as knits, shows that it can produce highly similar appearances to full hair geometries at different viewing distances with seamless LOD transitions, while achieving up to a 3x speedup.

Figures (17)

• Figure 1: Illustration of the single, aggregated ZhuZJYA23, and our BCSDF for longitudinal (top) and azimuthal (bottom) components. Our model replaces $F$ lobe with $\widehat{TT}$ lobe and distributes part of its contribution to $\widehat{R}$, $\widehat{TT}$, and $\widehat{D}$ with larger variances. Additionally, our $\widehat{B}$ lobe is designed to only account for the outgoing paths toward the back hemisphere.
• Figure 2: Compared to prior aggregated model ZhuZJYA23, ours is closer to avg. ref. that averages 100 different ply instances (ref.) for both frontlit and backlit with high and low hair densities. Both avg. ref. and ref. use single hair BCSDF $f^\text{single}$.
• Figure 3: Illustrations of light pass in aggregated cross-section: (a) the blue $A_1$ and pink $A_3$ paths are considered in the traditional local scattering with one and three backward scattering events by $B$ lobe. The green path illustrates our added $A_{1_+}$ paths, which encounter two deflections before returning to the shading hair. (b) when the incident and outgoing directions are close, the observed hair along the incident direction may not be occluded by others along the outgoing direction. All hairs with radius $r$ are contained by an elliptical cross-section, with major axis $\mathbf{r}_A$ and minor axis $\mathbf{r}_B$.
• Figure 4: Our approach (offline ray shooting and real-time OpenGL rasterizer) achieves close results to the path tracing reference, while our real-time implementation is 36000$\times$ faster for close view.
• Figure 5: Four hairstyles, ranging from straight to curly hair, under near, middle, and far views. Compared to path tracing with full hair geometry (full w/ PT), our approach (LoD w/ ours) utilizing reduced hair geometry through level-of-detail techniques can outperform existing dual scattering employing full geometry (full w/ DS) in both memory size and appearance. The FLIP error of ours increases for the far view due to the geometry simplification. Pink and cyan denote the number of segments and FLIP error of screen-space bounding box of hairs.