Material from Talks

Matthew Antone (MIT Graphics Group)   tone@graphics.lcs.mit.edu

Scalable Extrinsic Registration of Omni-Directional Image Networks

We describe linear-time algorithms for recovering scene-relative camera orientations and positions in networks of thousands of terrestrial images spanning hundreds of meters, in outdoor urban scenes, under uncontrolled lighting. Accurate registration of such image networks is currently infeasible by any other means, manual or algorithmic. Our system requires no human input or interaction, and recovers 6-DOF camera pose which is globally consistent on average to roughly $0.1^{\circ}$ and five centimeters, or about four pixels of epipolar alignment---sufficiently accurate for applications such as 3D reconstruction and image-based rendering.

The 6-DOF registration problem is decoupled into pure rotation and translation components, which take accurate intrinsic parameters, approximate extrinsic pose, and a connected camera adjacency graph as input. The algorithms estimate a local coordinate frame at each camera by classifying and combining thousands of low-level image features (lines and points) into a few robust aggregate features (vanishing points and motion baselines). These local frames are then propagated and registered through the adjacency graph. As output, the algorithms produce an accurate assignment of absolute orientation and position, and associated uncertainty estimates, to every camera.

Our principal contributions include multi-camera probabilistic extensions of classical two-camera alignment methods; new uses of the Hough transform for initialization of iterative numerical techniques; and formulation of expectation maximization algorithms for the recovery of camera pose without explicit feature correspondence. We also extend existing stochastic frameworks to handle unknown numbers of 3-D feature points, unknown occlusion, large scale (thousands of images and hundreds of thousands of features), and large dimensional extent (inter-camera baselines of tens of meters spanning areas hundreds of meters across). Finally, we introduce principled methods for the estimation and propagation of projective uncertainty, and present strong quantitative evidence of the superior utility of wide-FOV images in extrinsic calibration.

We assess the system's performance on synthetic and real data, and draw several conclusions. First, by fusing thousands of gradient-based image features into a few ensemble projective features, the algorithms achieve accurate registration even in the face of significant lighting variations, low-level feature noise, and error in initial pose estimates. Second, we show that registration of wide-FOV images is fundamentally more robust against failure, and more accurate, than is registration of ordinary imagery. Finally, the system surmounts the usual tradeoff between speed and precision: it is both faster and more accurate than manual bundle adjustment.

(Joint work with Seth Teller).

Chandrajit Bajaj (Department of Computer Sciences and Texas Institute of Computational and Applied Mathematics (TICAM) The University of Texas at Austin Austin, TX 78712-1188)    bajaj@cs.utexas.edu    http://www.cs.utexas.edu/users/bajaj  http://www.ticam.utexas.edu

Time Critical Scalable Rendering of Massive Triangular Meshes

Interactive browsing of massive triangular meshes is impossible with current single desktop graphics workstations. In this talk I shall describe a comprehensive approach, including the integration of a new specialized piece of image compositing hardware called the Metabuffer, in concert with back end compute + rendering cluster of graphics workstations, to achieving scalable interactive rendering. Our software solution is an end-to-end parallel and progressive platform, from the initial data access to the final display. I shall also provide performance details of our implemented pipeline of parallel, and progressive image compositing solutions coupled with parallel and progressive rendering of progressively accessed triangular meshes.

David Banks (Florida State University) banks@mailer.csit.fsu.edu

Theory of Rendering: Part 1

Rendering techniques borrow heavily from the physics of light transport, which is ultimately explained by quantum electrodynamics (QED). This talk bridges the first part of the gap, from QED to Maxwell's equations, as part of a 4-part program to establish a theoretical framework for rendering.

Baoquan Chen (University of Minnesota)

POP: A Hybrid Point and Polygon Rendering System for Large Data

We introduce a simple but important extension to the existing point rendering systems. Rather than using only points, we use both points and polygons to represent and render large meshe models. We start from triangles as leaf nodes and build up a hierarchical tree structure with intermediate nodes as points. During the rendering, the system determines whether to use point (of certain level node) or triangle (leaf node) for display depending on the screen contribution of each node. While points are used to speedup the rendering for distant objects, triangles are used to ensure the quality for close objects. We also perform pre-texturing on points and triangles when we build up the tree. Our hybrid representation facilitates effective antialiasing for texture mapping.

Ronald Fedkiw (Stanford Computer Science)  fedkiw@cs.stanford.edu

Physics Based Modeling for Computer Graphics

Scientists and engineers have used numerical techniques to simulate physical phenomena for many years. More recently, these numerical techniques have worked their way into a variety of new areas including computer graphics. Some key techniques will be briefly discussed including the Level Set Method for tracking interfaces and discontinuities, the Ghost Fluid Method for accurate modeling of boundary conditions at these interfaces and discontinuities, and Vorticity Confinement as a method of removing excess numerical dissipation on coarse grids. More importantly, these techniques will be shown "in action" for visual simulations of smoke and water.

Hugues Hoppe (Microsoft Research)   http://research.microsoft.com/~hoppe/

Surface Parametrizations    Slides:   html    pdf

This talk will present an overview of lapped textures and its subsequent applications in real-time fur rendering and hatching. Lapped textures cover an arbitrary surface by repeatedly instancing a small example texture onto a set of overlapping surface patches. The surface patches define local parametrizations that align to a globally consistent direction field. Because they use a small texture footprint, lapped textures reduce texture memory and bandwidth. This memory savings is particularly important when the texture is volumetric (for fur rendering), or when it is tone-dependent (for hatching). For additional information, refer to http://research.microsoft.com/~hoppe .

Chris Johnson (University of Utah)

Computational Multi-Field Visualization

Computational field problems; such as computational fluid dynamics (CFD), electromagnetic field simulation, and weather modeling -- essentially any problems whose physics can be modeled effectively by ordinary and/or partial differential equations--constitute the majority of computational science and engineering simulations. The output of such simulations might be a single field variable (such as pressure or velocity) or a combination of fields involving a number of scalar fields, vector fields, and/or tensor fields. As such, scientific visualization researchers have concentrated on effective ways to visualize large-scale computational fields. Much current and previous visualization research has focused on methods and techniques for visualizing a computational field variables (such as the extraction of a single scalar field variable as an isosurface). While single variable visualization often satisfies the needs of the user, it is clear that it would also be useful to be able to effectively visualize multiple fields simultaneously.

In this talk I will describe some of our recent work in scalar, vector, and tensor visualization techniques as applied to the domain of computational field problems. I will end with a discussion of ideas for the integration of techniques for creating computational multi-field visualizations.

Leif Kobbelt (Computergraphik & Multimedia RWTH-Aachen)   kobbelt@cs.rwth-aachen.de

Feature Sensitive Mesh Processing

Many mesh processing algorithms assume the actual geometry of a triangle mesh to be characterized by the vertex positions only. From the manifold point of view however, triangle meshes have to be considered as continuous piecewise linear surfaces.

In sufficiently smooth and flat regions of the surface this observation doesn't really matter since any triangulation will yield a decent approximation to the underlying geometry. In the presence of sharply curved features however, this is not true. Here, severe alias artifacts can affect the perceived surface quality and can lead to quite bad approximation behavior.

In my talk I will discuss several consequences of this observation and present recently developed algorithms for feature sensitive mesh generation and re-meshing techniques. I will report recent results in feature sensitive surface extraction from volume data, surface anti-aliasing by remeshing of blend regions in technical data sets, and diffusion based remeshing of triangle meshes.

Jos Stam (Alias/Wavefront)   jstam@aw.sgi.com

Stable Fluids: Towards interactive visual simulations of fluids

Building animation tools for fluid-like motions is an important and challenging problem with many applications in computer graphics. The use of physics-based models for fluid flow can greatly assist in creating such tools. Physical models, unlike key frame or procedural based techniques, permit an animator to almost effortlessly create interesting, swirling fluid-like behaviors. Also, the interaction of flows with objects and virtual forces is handled elegantly. Until recently, it was believed that physical fluid models were too expensive to allow real-time interaction. This was largely due to the fact that previous models used unstable schemes to solve the physical equations governing a fluid. In this talk we propose an unconditionally stable model which still produces complex fluid-like flows. As well, our method is very easy to implement. The stability of our model allows us to take larger time steps and therefore achieve faster simulations. We have used our model in conjunction with advecting solid textures to create many fluid-like animations interactively in two- and three-dimensions.

Richard Szeliski (Interactive Visual Media Group, Microsoft Research)  szeliski@microsoft.com

Representations for Image and Video-Based Modeling and Rendering

Obtaining photo-realistic geometric and photometric models is an important component of image-based rendering systems that use real-world imagery as their input. Applications of such systems include novel view generation and the mixing of live imagery with synthetic computer graphics. In this talk, I review a number of image-based representations (and their associated reconstruction algorithms) we have developed in the last few years. I begin by reviewing some recent approaches to the classic problem of recovering a depth map from two or more images. I then describe some of our newer representations and reconstruction algorithms, including volumetric representations, layered plane-plus-parallax representations (including the recovery of transparent and reflected layers), and multiple depth maps. Each of these techniques has its own strengths and weaknesses, which I will address. I will also present our work in video-based rendering, in which we synthesize novel video from short sample clips by discovering their (quasi-repetive) temporal structure.

About the Speaker

Richard Szeliski is a Senior Researcher in the Vision Technology Group at Microsoft Research, where he is pursuing research in 3-D computer vision, video scene analysis, and image-based rendering. His current focus is on constructing photorealistic 3D scene models from multiple images and video. He received a Ph.D. degree in Computer Science from Carnegie Mellon University, Pittsburgh, in 1988. He joined Microsoft Research in 1995. Prior to Microsoft, he worked at Bell-Northern Research, Schlumberger Palo Alto Research, the Artificial Intelligence Center of SRI International, and the Cambridge Research Lab of Digital Equipment Corporation.

Dr. Szeliski has published over 80 research papers in computer vision, computer graphics, medical imaging, and neural nets, as well as the book Bayesian Modeling of Uncertainty in Low-Level Vision. He is a Program Committee Chair for ICCV'2001, and is on the Editorial Board of the International Journal of Computer Vision. He has served as co-chair of the SPIE Conferences on Geometric Methods in Computer Vision, the 1999 Vision Algorithms Workshop, and as an Associate Editor of the IEEE Transactions on Pattern Analysis and Machine Intelligence.

Gabriel Taubin (IBM T.J.Watson Research Center and CalTech)  taubin@caltech.edu http://mesh.caltech.edu/taubin   or  taubin@us.ibm.com http://www.research.ibm.com/people/t/taubin

New Mesh Signal Processing Algorithms

Several closely related methods have been proposed in recent years to smooth, denoise, edit, compress, transmit, and animate very large polygonal models, based on signal processing techniques, constrained energy minimization, and the solution of diffusion differential equations. A number of these have been proposed to fix some of the drawbacks of the simple Laplacian smoothing algorithm, such as shrinkage, tangencial drift, and ridge over-smoothing. Almost none of these extensions beat Laplacian smoothing in simplicity and ease of implementation. In this talk I will describe new signal processing algorithms, all based on various extension and modifications of the discrete Laplacian operator defined on a polygonal mesh. These new algorithms, solve some of the existing problems while sharing the simplicity of Laplacian smoothing.

Demetri Terzopoulos (Courant Institute-New York University)    dt@cs.nyu.edu    http://www.mrl.nyu.edu/~dt 

Computational Learning Techniques for Animation

We have been investigating the promising role that computational theories of learning can play in the domain of computer graphics. I will present the motivations and results of our research in the context of physics and biology based modeling for animation. In particular, the presentation will cover:

1. Efficient neural network emulators of physical dynamics, dubbed "NeuroAnimators", that may be trained to produce physically realistic motions by observing simulated physical systems in action.

2. Biomechanically modeled artificial animals capable of acquiring, through sensor-guided reinforcement learning, motor controllers that effectuate lifelike, muscle-actuated locomotion.

3. Support vector machine learning methods to determine the functional competencies of composable motor controllers for the physics-based animation of articulated, anthropomorphic figures.

(References: 1. Grzeszczuk/Terzopoulos/Hinton SIGGRAPH 98. 2. Grzeszczuk/Terzopoulos SIGGRAPH 95. 3. Faloutsos/van de Panne/Terzopoulos SIGGRAPH 01. See also Terzopoulos CACM 42(8):1999.)

Material from Talks     Computer Graphics

2000-2001 Program: Mathematics in Multimedia