3D Reconstruction

Our environment is more and more shaped by the influence of digital media which itself evolves at tremendous pace. A small example: 15 years ago, offthe- shelf computers were only able to cope with flat 2D graphics. Improvements in system design made it possible to process 3D scenes, leading to a powerful contribution to our everyday life. However, such scenes need to be created by experienced artists employing difficult and expensive tools. The increasing demand for realism makes things not easier. At the same time, the rapid improvement of commodity hardware enables the development of low cost systems for 3D photography. How can systems for volumetric scene reconstruction be efficiently designed? How may high quality be combined with real-time processing? And how is it possible to create a lightweight architecture using a single computer? After presenting an introductory overview and sketching a novel approach to scene reconstruction, the author Christian Nitschke gives a survey of related work in high-performance reconstruction by shape from silhouette and shape from photo-consistency. To convey a profound understanding to the reader he explains the fundamentals of scene and camera geometry, image formation, light and color as well as 3D reconstruction from multiple views. The novel approach is then introduced in two steps by mapping a basic algorithm to an advanced algorithm using graphics hardware acceleration. Performance and quality issues are analyzed and discussed in detail. The book aims at software and system engineers, scientists and researchers, technical directors as well as the interested public.