Efficient Modeling And Rendering High Fidelity Surface Appearance

As one of the key research problems in computer graphics, appearance modeling and rendering aims to model how light interacts with objects and reproduce realistic surface appearance under given lighting and viewing conditions. To this end, high dimensional appearance models are introduced for faithfully modeling the interactions between light and object material. Although these models can well reproduce high fidelity surface appearance, they also bring several challenges for appearance capturing and rendering. E?cient modeling and rendering high fidelity surface appearance is still a challenge problem in graphics.This dissertation mainly focuses on real-time rendering translucent materials and e?cient modeling surface reflectance of real world materials. Specifically,Chapter 3 introduces a real-time rendering algorithm for heterogenous translucent objects with deformable geometry. The proposed method starts by rendering the surface geometry in two separate geometry bu?ers- the irradiance bu?er and the splatting bu?er- with corresponding mipmaps from the lighting and viewing directions,respectively. A fast adaptive sampling algorithm is presented to select irradiance samples from the irradiance bu?er according to geometric and material properties. A novel multiresolution GPU splatting algorithm is developed to splate irradiance samples to splatting bu?er without any precomputations. The method can render subsurface scattering e?ects of heterogeneous material in real time and support material editing and geometry deformation.Chapter 4 presents a generalized linear light source solution to estimate the anisotropic surface reflectance properties of a planar spatially varying material sample. We generalize the linear light source reflectometry by modulating the intensity along the linear light source, and show that a constant and two sinusoidal lighting patterns are su?cient for estimating the local shading frame and anisotropic surface reflectance. Based on the key observation that after factoring out the tangent rotation, the anisotropic surfacereflectance lies in a low rank subspace, the proposed algorithm successfully reconstruct the anisotropic surface reflectance and the local shading frame for each surface point from sparse measurements.Chapter 5 describes a novel method for recovering the spatially varying isotropic surface reflectance from a video of a rotating subject, with known geometry, under unknown natural illumination. The appearance recovery is formulated as an iterative process that alternates between estimating surface reflectance and estimating incident lighting. To regularize the recovery of the incident lighting, the algorithm relies on the observation that natural lighting is sparse in the gradient domain. The proposed solution significantly simplifies the capturing process, and greatly expand the application range.Finally, the dissertation discusses the limitations of existing methods for appearance modeling and rendering and explore the possible solutions and possible research directions for the future work.