2/9/2006 3:30pm-4:30pm ECCR 265
|
Realism and Surrealism in Computer Animation (Modeling and Simulation of Deformable Objects)
Min-Hyung Choi
University of Colorado at Denver
Over the past decades, a great stride has been made in the computer graphics
making a tremendous impact on motion picture and gaming industry. As the visual
quality of photorealistic imagery improves, attention shifts to other aspects
increase the feeling of reality, particularly for the behavioral realism in
animation and simulation. Due to the recent advancement of computer graphics
hardware and software, simulating the natural phenomena that is typically
prohibitively expensive to process in real-time has become a viable alternative
to the key frame based animations. This talk presents techniques in computer
graphics to model and simulate various deformable objects such as volumetric
human tissue model, thin shell cloth, fluid, and molecular dynamic simulation.
Two major issues pose a significant challenge. First, the conflicting demand
between computational efficiency and visual realism directs us to take
trade-offs or to sacrifice one for the other. Second, identifying the factors
that contribute the visual realism, not the overall accuracy of the simulation,
is often very difficult. While some very inaccurate and physically
non-conforming scenes may look perfectly natural, some animations based on a
rigorously tested numerical simulation look significantly abnormal. This talk
addresses techniques to reduce overall computational load while preserving
important visual realism by using domain decomposition, level of detail, and
adaptive simulation. Although the presented simulations are not strictly
accurate, they capture the essential visual, temporal, and an object's
behavioral characteristics in animation. In addition, the controllability
issues will be presented to accommodate a director's intension (even if it's
physically non-conforming) into animation and to generate compelling
surrealistic animation. Practical aspects, implementation, and future
directions on collaborative research will also be discussed.
|