Modeling of light reflection, transmission, and subsurface scattering for realistic image synthesis

This dissertation studies the modeling of light reflection, transmission and subsurface scattering. On reflection, we present a framework of spectral bidirectional reflectance distribution function (BRDF) measurement, data processing, compact representation, and spectrally-based image rendering. We describe two approaches for experimental measurement, and discuss their advantages and disadvantages. Given the measured data, we develop a numerical method for noise-filtering, data resampling and interpolation. Then we propose a method to represent compactly the spectral BRDFs in both spectral and spatial domains. Finally, we implement the represented and original spectral BRDF data for image rendering and discuss the error evaluation. On transmission, we present a physically-based analytical model for light transmission at rough surfaces and develop a Monte Carlo method to simulate light transmission processes. We consider two types of light transmission processes: single and multiple scattering. We derive an analytical expression for single scattering process. For the numerical simulation, we derive an analytical expression of individual visibility function for a ray starting at any surface height, and compare it with the numerical simulation. Based on individual visibility function, we develop a Monte Carlo method to simulate both the single and multiple scattering processes for two-dimensional random surfaces. We compare the analytical model with the previous simulation and our own simulation. On subsurface scattering, we present an approach to integrate the surface and subsurface scattering. Based on the splitting rule, we decompose a translucent object (inhomogeneous medium) with a rough surface into a homogeneous medium with a rough surface and an inhomogeneous medium with a plate surface. The scattering at a rough surface is described by a physically-based surface-reflection model; the scattering from the plate surface by the dipole point light source approximation of the diffusion theory. We implement this approach for rendering translucent objects based on our improved two-pass rendering technique, and discuss the contributions of surface and subsurface scattering, respectively