报告题目: Structuring 3D Geometry based on Symmetry and Instancing Information

主讲人: Fran?ois Sillion

(joint work with Aurélien Martinet (principal author), Nicolas Holzschuch, Cyril Soler)

时间: 2007年4月3日4:00-5:00pm

地点: FIT楼1区-515房间

报告摘要:

The growing complexity of the 3D data generated by various modeling techniques requires efficient algorithms to process and visualize these data. The efficiency of such algorithms is related to the underlying structure of the data: without any structuring, an algorithm has no hope to achieve sub-linear complexity which is a huge problem when processing large amount of 3D data with millions of primitives.

We define structural information as a two-scale notion, namely the object and the scene levels: tructural information at the object level captures the way an object has been modeled i.e. this object is a revolution surface, a Nurbs surface, . . . . Conversely, structural information at the scene level contains information about how objects are organized in the scene i.e. the hierarchy of objects in the scene or their semantic representation. We specially focus on two specific types of structural information and propose a way of structuring the scene at both the object and scene levels using the information of symmetry and instancing.

We first propose an original method to structure the geometry at the object level based on the information of symmetry. We show how to compute the whole symmetry group of a 3D mesh i.e. either discrete (such as planar or rotational symmetries) or continuous (such as cylindrical or spherical symmetries). Finding all symmetries of a shape is much more difficult than simply checking whether a given transform actually is a symmetry. In particular the naive approach that would consist of checking as many potential candidates as possible to find a symmetry is far too costly. We thus need a deterministic method to find good candidates. Inspired by the work on principal component analysis, we introduce for this purpose the generalized moment functions of a 3D shape.

In a second part, we propose a method to structure the geometry at the scene level based on instancing information i.e. the structure of what is repeated in the scene at multiple levels of scale. The knowledge of such information can be used in many tasks such as rendering efficiency for ray-tracing or radiosity, or for mesh-editing. From the symmetry information that we have computed on the objects of the scene, we show that computing such a hierarchy is a non-trivial process and divide it in two steps: a first step which aims at discovering frequent geometric patterns i.e. objects or set of objects that occur multiple times in a scene and a second step that organizes these patterns into a hierarchy.

Sillion教授简介:

Fran?ois Sillion is a senior researcher with INRIA (Institut National de Recherche en Informatique et Automatique), and the director of the Rh?ne-Alpes research unit in Grenoble. He obtained a graduate degree in Solid State Physics and a PhD in Computer Science as a student of the “Ecole Normale Supérieure” in Paris. Before joining CNRS, he spent two years as a postdoctoral research associate in the Program of Computer Graphics of Cornell University (USA). He joined INRIA in 1998.

Fran?ois Sillion currently chairs the Eurographics working group on rendering, and is a member of the editorial board of the “Computer Graphics Forum” and ?ACM Transactions on Graphics ? journals. He co-chaired the technical program of the Eurographics'96 conference, the 1992 Eurographics workshop on rendering, and organized the 1993 Eurographics workshop on rendering. He was a member of several conference program committees, including SIGGRAPH , Eurographics, EG workshop on rendering, Pacific graphics, ACM Symposium on Interactive 3D graphics etc. He delivered tutorials in several SIGGRAPH and Eurographics conferences, and invited presentations in several international conferences.

Fran?ois Sillion published many papers in the fields of lighting simulation, real-time rendering, visibility and shadow techniques, focusing on the complementary issues of image realism and interactive control of visualization. He wrote a comprehensive book on the radiosity method, with Claude Puech (“Radiosity & Global Illumination”, Morgan Kaufmann Publishers, 1994).