Animated Smooth Levels of Detail


Coarse View-Dependent Multi-Resolution Models for Hierarchical and Deformable Polygonal Objects
Project duration: 1997-1999
Contact: Dieter Schmalstieg, Anton Fuhrmann

Description

The ASLOD (Animated Smooth Level Of Detail) approach presented in this paper is capable of animating and rendering scenes consisting of mixtures of conventional and multi-resolution models, and apply deformation to objects.

Application

Large-scale virtual environments and other interactive graphical applications which require both real-time rendering with constant high frame rates and animation of rigid as well as organically deformable models.

Problems

While a large amount of work is geared towards multi-resolution modeling and rendering, the presented methods work almost exclusively with static models and scenes. New approaches are needed to include animation and deformation as well.

Approach

ASLOD features coarse view-dependent multi-resolution rendering by a consistent decomposition of large models into arbitrary regions together with an analytic solution to optimize polygon allocation to such assemblies of multi-resolution models. The multi-resolution model can be used together with animated hierarchical scene graphs. Finally, the multi-resolution model can be combined with a hierarchical skeleton and a real-time deformation algorithm for animating organic shapes rendered at any desired level of detail.

Publications

[PDF] Coarse View-Dependent Levels of Detail for Hierarchical and Deformable Models (D. Schmalstieg and A. Fuhrmann). Technical report TR-186-2-99-20, Vienna University of Technology, 1999.


Here are some results:



Coarse View-Dependent Rendering

(a) (b)
(c) (d)
Large terrain model with coarse view-dependent rendering - (a) original terrain model with 32K triangles sen from corner, (b) same model seen from top, (c) coarse-view dependent model with 10K triangles seen from corner, (d) coarse view-dependent model seen from top. Note how the lower right corner - the one closest to the viewpoint - has been assigned significantly more triangles.


Animation and Deformation


A deformable mesh of an animated mannequin obtained from a laser range scanner (courtesy of Cyberware). The mesh was decomposed into 19 regions (10 deformable regions at the joints and 9 rigid regions). The left view show the logical decomposition of the mesh, the right view shows the regions color coded.


The two images above show two snap shots from a simultaneous deformation and region-based view-dependent simplification procedure. Note that all computation is done in real-time.


This page is maintained by Dieter Schmalstieg. It was last updated on October 4, 1999.
If you have any comments, please send a message to dieter@cg.tuwien.ac.at.