< Terug naar vorige pagina

Publicatie

Acquiring the World through Photographs

Boek - Dissertatie

Applications in entertainment, augmented/virtual reality, architecture and digital museums require the creation, or rendering of realistic computer-generated images of real-world objects from novel viewpoints and with novel illumination. These images should be indistinguishable from real photographs taken under similar conditions. In this dissertation, we will explore methods to acquire accurate input models, for photorealistic rendering, from the real-world using cameras and photographs. The capture step should be as simple as possible. We will describe a system for capturing detailed geometry within a cinematographic setting. The system uses a compact and simple set-up of cameras and a projector that illuminates the scene with specific light patterns and allows real-time depth reconstruction in a single shot. After, we will broaden the scope and look into methods to acquire accurate models for geometry, lighting and reflective properties of materials from photographs taken of real-world objects. We focus on approaches that are flexible and do not require specialized lighting or capture equipment. We will start off with an overview of inverse rendering and analyze the well-posedness and conditioning of such methods and assess which inverse problems are well-posed or solvable versus ill-posed. We will present an image-based technique for recovering 3D shape and spatiallyvarying reflectance properties from a sparse set of photographs, taken under varying illumination. This technique models the reflectance with a set of low-parameter BRDFs without knowledge of the location of the light-sources or camera. This results in a flexible and portable system that can be used in the field. The technique can successfully recover shape and reflectance of several objects (synthetic and real) and the acquired information can be used to rendering the objects with modifications to the geometry and lighting via traditional rendering methods. Finally, we propose an approach for recovering the reflectance of a static scene with known geometry from a collection of images taken under distant, unknown illumination. In contrast to previous work, we allow the illumination to vary between the images, which greatly increases the applicability of the approach. Using an allfrequency relighting framework based on wavelets, we are able to simultaneously estimate the per-image incident illumination and the per-surface point reflectance. The wavelet framework allows for incorporating various reflection models. The quality of the results is demonstrated for synthetic test cases as well as for several datasets captured under laboratory conditions. Combined with multi-view stereo reconstruction, we are even able to recover the geometry and reflectance of a scene solely using images collected from the Internet.
Aantal pagina's: 146
Jaar van publicatie:2015
Toegankelijkheid:Open