Reconstruction of 3D models of real objects from digital images
V. Lempitsky, A. Shokurov, D. Ivanov, Ye. Kuzmin
Moscow State University, Department of Mathematics and Mechanics
Moscow, Russia
Object reconstruction from photographs and video has been a topic of intensive research for many years. This is motivated by the need for highly-realistic models, which can be employed in many virtual reality applications (computer games, virtual tourism). Image-based techniques can also be employed in CAD-systems, non-intrusive metrology systems, and augmented reality applications. Image-based modeling starts with registration of input images in a global coordinate system. After registration it becomes possible to perform surface-based reconstruction or volumetric reconstruction, obtaining a model of the depicted object. Alternatively, a virtual object can be brought into the global coordinates and reprojected onto the images (augmented reality, CAD-systems) or some measurements of the depicted objects can be done (non-intrusive metrology). Recently, several commercial and non-commercial image-based modeling systems have been developed. Almost all of them require significant amount of user interaction. At the same time a number of automatic methods have been proposed in academic papers. This work briefly surveys some of the available image-based algorithms and applications. It is partially based on the experience obtained from the development of an image-based modeling system at CGG MSU.
|