A Compact Tucker-Based Factorization Model for Heterogeneous Subsurface Scattering

This paper presents a novel compact factored subsurface scattering representation for optically thick, heterogeneous translucent materials. Our subsurface scattering representation is a combination of Tucker-based factorization and a linear regression method. We first apply Tucker factorization on the intensity profiles of the heterogeneous subsurface scattering responses. Next, we fit a polynomial model for characterizing the differences between the different color channels with a linear regression procedure. We show that our method achieves good compression while maintaining visual fidelity. We validate our heterogeneous subsurface scattering representation on various real-world heterogeneous translucent materials, geometries and lighting conditions.