Conferences de Demetri Terzopoulos lors de son sejour au CEREMADE

comme professeur invite les 12, 14, 19, 25 et 28 Septembre 2006.

Bio: Demetri Terzopoulos is the Chancellor's Professor of Computer

Science at UCLA. He graduated from McGill University and was awarded

the PhD degree by MIT in 1984. He is a Fellow of the IEEE, a Fellow of

the Royal Society of Canada, and a member of the European Academy of

Sciences. His many awards include an Academy Award for Technical

Achievement (a Technical Oscar) from the Academy of Motion Picture

Arts and Sciences for his pioneering work on physics-based computer

animation. He is one of the most highly-cited computer scientists and

engineers in the world, with approximately 300 published research

papers and several volumes, primarily in computer graphics, computer

vision, medical imaging, computer-aided design, and artificial

intelligence/life.

Professor Terzopoulos is one of the most well known scientists in the domains of computer vision

and computer graphics. He is among the top 100 scientists in terms of citations in all domains of science and engineering.

Mardi 12 Septembre 2006, 14h30, Salle A 711

A Tensor Algebraic Framework for Image Synthesis, Analysis and Recognition

Demetri Terzopoulos

University of California, Los Angeles

We introduce a multilinear (tensor) algebraic framework for image

synthesis, analysis, and recognition. Natural images result from the

multifactor interaction between the imaging process, the illumination,

and the scene geometry. Numerical multilinear algebra provides a

principled approach to disentangling and explicitly representing the

essential factors or modes of image ensembles. Our multilinear image

modeling technique employs a tensor extension of the conventional

matrix singular value decomposition (SVD), known as the N-mode SVD.

This leads us to a multilinear generalization of principal components

analysis (PCA) and a novel multilinear generalization of independent

components analysis (ICA). As example applications, we tackle

currently important problems in computer graphics, computer vision,

and pattern recognition, in particular, image-based rendering,

specifically the multilinear synthesis of images of textured surfaces

for varying viewpoint and illumination, as well as the multilinear

analysis and recognition of facial images under variable face shape,

view, and illumination conditions.

Jeudi 14 Septembre 2006, 14h30, Salle A 711

Biomechanical Modeling and Neuromuscular Control of the Face-Head-Neck

System

Demetri Terzopoulos

University of California, Los Angeles

Facial animation has a lengthy history in computer graphics. To date,

most efforts have concentrated either on labor-intensive keyframe or

motion capture animation schemes. As an alternative, we advocate the

highly automated animation of faces using physics-based and behavioral

animation methods. To this end, we develop a biomechanical model of

the face, which includes synthetic facial soft tissues with embedded

muscle actuators. Despite its sophistication, our facial model can

nonetheless be simulated in real time on a high-end PC. The model

incorporates a motor control layer that automatically coordinates eye

and head movements, as well as muscle contractions to produce natural

expressions. We augment the synthetic face with a perception model

that affords it a visual awareness of its environment, and we provide

a sensorimotor response mechanism that links percepts to meaningful

actions. Unlike the human face, the neck has been largely overlooked

in the computer graphics literature, this despite its complex

anatomical structure and the important role that it plays in

supporting the head in balance while generating the controlled head

movements that are essential to so many aspects of human behavior. We

introduce a biomechanical model of the human head-neck system.

Emulating the relevant anatomy, our model is characterized by

appropriate kinematic redundancy (7 cervical vertebrae coupled by

3-DOF joints) and muscle actuator redundancy (72 neck muscles arranged

in 3 muscle layers). This anatomically consistent biomechanical model

confronts us with a challenging motor control problem, even for the

relatively simple task of balancing the mass of the head in gravity

atop the cervical spine. We develop a neuromuscular control model for

human head animation that emulates the relevant biological motor

control mechanisms. Employing machine learning techniques, the

controller's neural networks are trained offline to efficiently

generate online control signals for the autonomous behavioral

animation of the human head and face.

Mardi 19 Septembre, 14h dans le cadre du congres MIA06
Mathematics  and Image Analysis 2006
http://www.ceremade.dauphine.fr/~cohen/mia2006/

Deformable and Functional Models in Medical Image Analysis

The modeling of biological structures and the model-based interpretation of medical images present many challenging problems. I will present a powerful paradigm known as deformable models, which combines geometry, computational physics, and estimation theory. Deformable models evolve in response to simulated forces as dictated by the continuum mechanical principles of flexible materials, expressed mathematically via variational principles and PDEs. The talk will focus on several biomedical applications, including image segmentation using dynamic finite element and topologically adaptive deformable models, as well as recent work on "deformable organisms" which aims more fully to automate the segmentation process by augmenting deformable models with behavioral and cognitive control mechanisms. I will also discuss the recent trend towards functional modeling, such as craniofacial models that include the biomechanical modeling of facial tissues and muscles of facial expression.

Lundi 25 Septembre 2006, 14h, Salle A 709

Artificial Animals and Humans: From Physics to Intelligence

Demetri Terzopoulos

University of California, Los Angeles

The confluence of virtual reality and artificial life, an emerging

discipline that spans the computational and biological sciences, has

yielded synthetic worlds inhabited by realistic, artificial flora and

fauna. Artificial animals are complex synthetic organisms that possess

functional biomechanical bodies, perceptual sensors, and brains with

locomotion, perception, behavior, learning, and cognition centers.

Artificial humans and lower animals are of interest in computer

graphics because they are self-animating graphical characters that can

dramatically advance the state of the art of production animation and

interactive game technologies. More broadly, these biomimetic

autonomous agents in realistic virtual worlds also foster deeper

computationally oriented insights into natural living systems. In

addition, they engender interesting applications in computer vision,

sensor networks, and other domains.

Jeudi 28 Septembre 2006, 14h, Salle A 709

Virtual Vision: Human Simulation and Visual Sensor Networks

Demetri Terzopoulos

University of California, Los Angeles

Virtual vision is a fledgling paradigm for computer vision research,

which exploits computer graphics and realistic virtual worlds. This

lecture will be in two parts. First, we address the challenging

problem of emulating the rich complexity of real pedestrians in urban

environments. Our artificial life approach integrates motor,

perceptual, behavioral, and cognitive components within a

comprehensive model of pedestrians as individuals, yielding

unprecedented fidelity and complexity for fully autonomous multi-human

simulation in a large urban environment. We represent the environment

using hierarchical data structures, which efficiently support the

perceptual queries that influence the behavioral responses of the

autonomous pedestrians and sustain their ability to plan their actions

on local and global scales. Second, we explore the use of this

visually and behaviorally realistic simulator in the development and

testing of visual surveillance systems. Our research would be more or

less infeasible in the real world given the impediments to deploying

and experimenting with an appropriately complex camera sensor network

in a large public space the size of, say, an airport or train station.

In particular, we develop and experiment with surveillance systems in

a virtual train station environment populated by autonomous, lifelike

virtual pedestrians, wherein easily reconfigurable virtual cameras

generate synthetic video feeds that emulate those generated by real

surveillance cameras monitoring richly populated public spaces.