Kolja Kähler

Head shop: generating animated head models with anatomical structure

We present a versatile construction and deformation method for head models with anatomical structure, suitable for real-time physics-based facial animation. The model is equipped with landmark data on skin and skull, which allows us to deform the head in anthropometrically meaningful ways. On any deformed model, the underlying muscle and bone structure is adapted as well, such that the model remains completely animatable using the same muscle contraction parameters. We employ this general technique to fit a generic head model to imperfect scan data, and to simulate head growth from early childhood to adult age.

May I talk to you? : -) - facial animation from text

We introduce a facial animation system that produces real-time animation sequences including speech synchronization and non-verbal speech-related facial expressions from plain text input. A state-of-the-art text-to-speech synthesis component performs linguistic analysis of the text input and creates a speech signal from phonetic and intonation information. The phonetic transcription is additionally used to drive a speech synchronization method for the physically based facial animation. Further high-level information from the linguistic analysis such as different types of accents or pauses as well as the type of the sentence is used to generate non-verbal speech-related facial expressions such as movement of head, eyes, and eyebrows or voluntary eye blinks. Moreover, emotions are translated into XML markup that triggers emotional facial expressions.

A Head Model with Anatomical Structure for Facial Modeling and Animation

In this dissertation, I describe a virtual head model with anatomical structure. The model is animated in a physics-based manner by use of muscle contractions that in turn cause skin deformations; the simulation is efficient enough to achieve real-time frame rates on current PC hardware. Construction of head models is eased in my approach by deriving new models from a prototype, employing a deformation method that reshapes the complete virtual head structure. Without additional modeling tasks, this results in an immediately animatable model. The general deformation method allows for several applications such as adaptation to individual scan data for creation of animated head models of real persons. The basis for the deformation method is a set of facial feature points, which leads to other interesting uses when this set is chosen according to an anthropometric standard set of facial landmarks: I present algorithms for simulation of human head growth and reconstruction of a face from a skull. The creation of computer-animated human faces is a long-standing and challenging problem since the early 1970s. There are numerous approaches to facial animation, but to this day no general-purpose system exists that solves the problem in a manner satisfying the needs of all practical applications. In the medical field, highly accurate reproduction of a real head is required to enable well-informed decisions in surgery planning. Animation capabilities and computation time are not important. On the other hand, realism is of minor concern in interactive dialog systems or computer games. Here, the animation merely has to look plausible but play in real time. A real-life model does not need to be reproduced with all subtleties of facial shape and texture. An obvious approach to achieve generality is the simulation of the inner workings of a real face. High expectations are tied to physics-based systems, where the ultimate goal is to have the full range of conformation and expressiveness in the face emerge “naturally” through precise modeling of the anatomical structure and accurate simulation of tissue properties. This has so far only been realized in parts, and no current implementation catches all the intricacies of the human face. Traditionally, the computational cost of physics-based simulation has been prohibitive for real-time facial animation on consumer-class PC hardware. This has changed dramatically in recent years, making fast high quality animation possible on current desktop computers. But, apart from the run-time issues, constructing a virtual head model with the complex structure of skull, muscles, and skin is non-trivial, requiring artistic skills and time. This situation motivates my dissertation: I propose an anatomy-based virtual head model that is animatable in real time using numerical simulation techniques, driven by an advanced facial muscle model. The simulation is efficient enough to achieve real-time frame rates on current PC hardware. Manual construction of such a struc-

Dynamically refining animated triangle meshes for rendering

We present a method to dynamically apply local refinements to an irregular triangle mesh as it deforms in real time. The method increases surface smoothness in regions of high deformation by splitting triangles in a fashion similar to one or two steps of Loop subdivision. The refinement is computed for an arbitrary triangle mesh, and the subdivided triangles are simply passed to the rendering engine, leaving the mesh itself unchanged. The algorithm can thus be easily plugged into existing systems to enhance the visual appearance of animated meshes. The refinement step has very low computational overhead and is easy to implement. We demonstrate the use of the algorithm in a physics-based facial animation system.

Efficient processing of large 3D meshes

Due to their simplicity triangle meshes are often used to represent geometric surfaces. Their main drawback is the large number of triangles that are required to represent a smooth surface. This problem has been addressed by a large number of mesh simplification algorithms which reduce the number of triangles and approximate the initial mesh. Hierarchical triangle mesh representations provide access to a triangle mesh at a desired resolution, without omitting any information. In this paper we present an infrastructure for mesh decimation, geometric mesh smoothing, and interactive multiresolution editing of arbitrary unstructured triangle meshes. In particular, we demonstrate how mesh reduction and geometric mesh smoothing can be combined to provide a powerful and numerically efficient multiresolution smoothing and editing paradigm.

Subdivision surface simplification

A modified quadric error metric (QEM) for simplification of Loop subdivision surfaces is presented The suggested error metric not only measures the geometric difference but also controls the smoothness and well-shapedness of the triangles that result from the decimation process. Minimizing the error with respect to the original limit surface, our method allows for drastic simplification of Loop control meshes with convenient control over the reproduction of sharp features.

Dynamic refinement of deformable triangle meshes for rendering

We present a method to adaptively refine an irregular triangle mesh as it deforms in real-time. The method increases surface smoothness in regions of high deformation by splitting triangles in a fashion similar to one or two steps of loop subdivision. The refinement is computed for an arbitrary triangle mesh and the subdivided triangles are simply passed to the rendering engine, leaving the mesh itself unchanged. The algorithm can thus be easily plugged into existing systems to enhance the visual appearance of animated meshes. The refinement step has a very low computational overhead and is easy to implement. We demonstrate the use of the algorithm in our physics-based facial animation system.