| Photo-realistic image synthesis is the process of creating or rendering images that are indistinguishable from real photographs. Applications of photo-realistic image synthesis play an important role in a wide range of areas including cultural heritage preservation, entertainment, and architecture. Accurate modeling of scene elements such as material properties and illumination is a vitally important and essential part of photo-realistic image synthesis. This thesis builds on and extends prior research in appearance modeling to enable the acquisition of more detailed appearance and to allow for easy editing.;This thesis presents two novel contributions. First, We propose a novel photometric stereo solution to jointly estimate surface orientations and translucent material properties from a globally planar homogeneous translucent object. Similar to classic photometric stereo, our method only requires as few as three observations of the translucent object under directional lighting. Second, we propose a novel measurement-based approach for recoloring regions of a scene without any knowledge of geometry and lighting of the scene while maintaining physically accurate interreections. |
