Symmetry is a ubiquitous concept and repetitions are very common in many man-made objects because of its impact on economical and functional considerations and aesthetic concerns. The goal of this project is to explore the notion of symmetry as a means to provide non-local coupling between geometric computations and to extract semantic information in order to improve the 3D reconstruction quality, specifically in the context of urban reconstruction.
Symmetry is regularly used as an organizing principle in urban planning and design, while use of repeating structures is reinforced to ease the construction process. Therefore, it is extremely important to explore such relations to enable a better understanding and processing of relevant data. Specifically, 3D reconstruction of urban scenes is one particular application that can benefit from the explicit knowledge of symmetry as each repeating element provides multiple observations of the same geometric piece. Recently, the success of virtual navigation tools like Google Earth and Microsoft Visual Earth have attracted the attention of researchers towards this popular problem. Image-based modeling methods which aim to provide fast and accurate reconstruction of buildings have gained significant attention due to the advances in camera technology and the flexible and economic data acquisition possibilities they provide. Many challenges arising from lighting variations, insufficient textures, occluding objects, however, are still unsolved, especially in the reconstruction of clean, accurate, and detailed 3D models, which are often desirable for further interactive edits. Our goal is to introduce an image-based 3D reconstruction approach that addresses these challenges by using symmetry priors inherent in urban scenes.
We present a structure-from-motion framework that detects and conforms to structural regularities, while simultaneously recovering 3D geometry starting from a set of facade images. A novel graph based global analysis yields a globally consistent 3D geometry reconstruction with explicit encoding of the facade regularities.
We present a coupled formulation for detecting symmetric line arrangements and 3D reconstruction for producing factored facade models. Unlike most competing approaches, we benefit from large-scale model repetitions, and can robustly handle inputs with strong reflections, shadow elements, or outlier objects, which are difficult to disambiguate with only local reasoning.
Given a polygonal tree model this projects deals with the problem of generating intermediate stages of the plant. A biological plausible growth model allows to animate the development of the tree.
Duygu Ceylan, Minh Dang, Boris Neubert, Niloy J. Mitra, Mark Pauly
Computer Graphics Forum, under review
Soeren Pirk, Till Niese, Oliver Deussen, Boris Neubert
ACM Transactions on Graphics (Proceedings of Siggraph Asia), 2012