Alla Sheffer is an associate professor in the Computer Science department at the University of British Columbia. She investigates algorithms for geometry processing focusing on computer graphics applications. Her work had been published at venues such as Siggraph, Siggraph Asia, Eurographics,and the Symposium on Geometry Processing. She is the recipient of a Killam Junior Research Fellowship (2009) and an IBM Faculty Award (2006). Alla co-chaired the Symposium on Geometry Processing, the leading geometry processing conference, and served on the PC of major graphics conferences including SIGGRAPH and Eurographics. Alla Sheffer received her PhD from the Hebrew University of Jerusalem (1999), she then did a postdoc at the University of Illinois at Urbana-Champaign and was an assistant professor at Technion, Israel.

AG 1, AG 3, AG 5, SWS, AG 2, AG 4, RG1, MMCI   Info

AG Audience

English

60 Minutes

Saarbrücken

Realistic virtual worlds contain scores of man-made objects. Models of such shapes are usually created using a variety of softwares and vary significantly from the typical scanned meshes targeted by the vast majority of geometry processing algorithms. Man-made objects have a lot more structure than the natural shapes targeted so far. At the same time their models do not satisfy the commonly expected rules of mesh "niceness", and in general are represented by a polygon soup. Consequently, editing, manipulation and visualization of such models require the development of new processing methodologies.

In this talk I'll discuss two method for processing of man-made shapes, one designed for resizing of such objects and one for shape abstraction. Resizing can be very useful when creating new models or placing models inside different scenes. However naive resizing can destroy some model structure and lead to serious visual artefacts. We introduce a resizing method based on a non-homogeneous space-deformation approach which detects and protects the vulnerable structures.

While models of man-made shapes can be very detailed, capturing every small feature, they are often identified and characterized by a small set of defining curves. Compact, abstracted shape descriptions based on such curves are often visually more appealing than the original models, which can appear to be visually cluttered. We introduce an algorithm for abstracting 3D models using characteristic curves or contours as building blocks. In our algorithm, we use a two-step
procedure that first approximates the input model using a manifold envelope surface and then extracts from it a hierarchical abstraction curve network.

One common denominator of these methods is the focus on protection of the structure and regularity of the underlying shapes. The other is the choice of suitable representations and processing techniques which are independent of the underlying model mesh quality allowing for proceeding of poor quality meshes and even polygon soups.

Michael Wand

--email hidden

passcode not visible

logged in users only