Sept. 30
Thomas J. Peters
Abstract
Graphics has matured so that geometrically flawed images are largely
a distant memory. Experimentation has led to understanding of
the geometric accuracy needed to transform floating-point
representations into pleasing images.
This success achieved has not generalized to other applications.
Visualization is one type of model simulation. Others are
computational fluid dynamics, computational electromagnetics,
and finite element analysis. Much tedious, human intensive
modification of geometry is still required to support these
applications. This talk introduces two supporting concepts
Nov. 4
Jack Tumblin (Cornell )
Abstract
Almost all our picture-making processes still have a fundamental
unsolved problem: we are immersed in a world of high contrast scenes
that we cannot directly reproduce in displayed images. Contrast, the
ratio between large and small light intensities, can easily exceed
100,000:1 in night scenes, sunsets, or any scene with visible light
sources, specular highlights, and deep shadows. Commonly used displays
may never cover this huge range because typical CRTs, printers and
slide projectors offer no more than about 50:1 displayed contrast.
Outside of this small displayable range, cameras and computer graphics
renderings often "clip" or "saturate" scene intensities for display,
causing empty black shadows and blank white highlights, but our eyes
almost never do this. Instead, local adaptation processes adjust visual
sensitivity as we glance around a scene to capture low-contrast details
and textures. Few dark shadows permanently appear as featureless black,
even in scenes lit by moonlight. In a room lit only by a bare
lightbulb, local adaptation allows us to read the tiny lettering on the
bulb's dazzling surface, and yet at the same time we can see the dim
room surrounding us. How can we construct a low contrast displayed
image that preserves revealed by local adaptation? If we severely
compress all large scene contrasts for display, how can we avoid
shrinking the small contrasts to invisibility?
I will present three workable methods for detail-preserving contrast
reduction. The first uses an S-shaped function, similar to the response
of film or retinal ganglia, to compress only the illumination
components of a computer graphics rendering, while preserving scene
reflectances and transparencies. The second method interactively
adjusts the displayed image according to the user's direction of gaze
to imitate foveally-dominated visual adaptation; this method requires a
computerized display. Third, the "LCIS" (Low Curvature Image
Simplifier) method uses a variant of anisotropic diffusion to
progressively separate fine details of a scene from its large features
and most important boundaries. It forms an ordered hierarchy (a scale
space) of scene boundaries and shadings, then compresses only the large
feature contrasts for display. The LCIS method is the least
restrictive; it works for any image from any source for any display.
I will demonstrate each method with example images (several are
available at [http://www.gvu.gatech.edu/people/jack.tumblin]), and will
discuss how each method might be useful to capture, edit and display
both synthetic and real-world imagery. The first two methods were
published in "Two Methods for Display of High Contrast Images" in ACM
Transactions on Graphics Vol. 18 #1 (January 1999) pp. 56-94, and
supplied the cover images for the previous issue. The third method was
presented in "LCIS: A Boundary Hierarchy for Detail-Preserving Contrast
Reduction" at ACM SIGGRAPH`99, pp. 83-90 and supplied the title-page
image for this year's Proceedings.
Nov. 11
Igor Guskov (Caltech)
Abstract
Recent progress in 3D acquisition techniques and mesh simplification
methods has made triangulated mesh hierarchies of arbitrary topology a
basic geometric modeling primitive. In this talk I will introduce a new
non-uniform relaxation technique which lets us build a Burt-Adelson type
detail pyramid on top of a mesh simplification hierarchy. The resulting
multiresolution framework makes it easy to perform a full range of
standard signal processing tasks such as smoothing, enhancement,
filtering and editing of arbitrary surface triangulations.
Jan. 27
Eric Haines (Cornell )
Abstract
A graphics accelerator including a Z-buffer is becoming standard issue
for new graphics PCs. Inexpensive chipsets for 3D rendering are bundled
with medium and high-end machines, with low end machines and portables
soon to follow. Clocking a sustained rate of over 20 million polygons a
second, game consoles such as the Playstation2 make incredible effects
possible. But, will the hardware Z-buffer always make sense as the
primary interactive rendering technology?
In this talk I will outline areas where the Z-buffer is weak, and some
possible solutions for these problems. The talk will also discuss some
myths about Z-buffering, such as it being doomed in the long run because
of its linear performance. Alternate rendering strategies such as
parallel ray tracing will be touched upon and compared.
Author Bio
Eric Haines is a graduate of the Program of Computer Graphics at
Cornell and is coauthor of the recent book 'Real-Time Rendering. He has worked 15 years in the computer graphics industry,
designing and implementing ray tracing, radiosity, depth buffer, and
other rendering algorithms.
He is currently an editor for the "journal
of graphics tools", the editor of the "Ray Tracing News", archivist for
the "Graphics Gems" code base, and webmaster for ACM TOG. He is
currently employed by Autodesk and is a part-time member of the staff at
the PCG at Cornell.
Feb. 3
Michael Capps (Naval Postgraduate School)
Abstract
Sharing information in virtual worlds is awfully difficult. Like
most networked systems, this problem becomes exponentially less
pleasant with any increase in the number of users.
However, it turns out that the real roadblocks in the delivery of
cyberspace are often political: graphics researchers treat the network
as a "black box"; network researchers think of graphical 3-D content
as "just more bits"; and it seems no one wants to accept a solution Not
Invented Here.
In this presentation, I will discuss three approaches to sharing
graphical content that will (hopefully!) excite and amaze:
Author Bio
Michael V. Capps
Michael Capps is a professor in the Modeling, Virtual Environments,
and Simulation curriculum at the Naval Postgraduate School. His
research involves techniques for optimization of networked graphics,
and software engineering for interoperable, scalable shared virtual
environments. Michael has degrees in Creative Writing, Mathematics,
and Computer Science from the University of North Carolina and MIT.
For his work in multi-user virtual environments, he was chosen as one
of 50 computer graphics pioneers featured in the SIGGRAPH documentary
"The Story of Computer Graphics." Michael actively consults for such
cloak-and-dagger entities as NSA, NRO, and Disney Imagineering. He is
currently serving as Technical Program Chair for the 2000 ACM
Web3D / VRML Symposium this February, which you shouldn't miss.
Feb. 10
John Hughes (Brown)
Abstract
The linear algebra of real n-space (n = 2,3,4) has been used for years as a foundation for representing things in computer graphics -- points, vectors, normal vectors, color-triples, etc. -- and matrix multiplication has been used as a fundamental building block in transforming things. The free interchangeability of these objects -- "everything's just three floats!" -- leads to a wide variety of programming errors, errors that can be prevented by a slightly more sophisticated view and a type-checking system that supports that view. Stephen Mann, Nathan Litke and Tony DeRose have developed a "coordinate free geometry" system that achieves this; I'll describe that system, and a richer hierarchy of mathematical objects that can help prevent higher-level programming errors. The talk will bounce between (relatively simple) mathematical concepts, hints at how an object-oriented language can be used to implement these concepts, and code-fragments illustrating the application of such a system.
The ideas in this talk are a foundation I'm considering for a new approach to presenting the mathematical material at the start of computer graphics, and I seek feedback from the audience about the approach.
Feb. 17
Renato B. Pajarola (UC Irvine)
Abstract
Most systems that support the visual interaction with 3D models use
shape representations based on triangle meshes. The increasing size of
these models combined with the commonly used expensive representation
formats lead to immense storage consumption for maintaining large
collections of shapes. Furthermore, it imposes limits on applications,
where complex 3D models must be accessed remotely. Techniques for
simplifying and compressing 3D models help reducing the storage space
and transmission time. Multiresolution models provide explicit access
to a specific level-of-detail (LOD) and often provide incremental
reconstruction of the LODs starting with an initial crude model.
Unfortunately, compared to the best non-progressive mesh compression
methods, most progressive refinement techniques impose a significant
overhead for storing and transmitting the full resolution model.
The proposed Compressed Progressive Meshes (CPM) approach eliminates
this overhead. It uses a new "Implant Sprays" technique, which refines
the mesh topology between LODs in batches. An Implant Sprays batch
increases the number of vertices in a triangular mesh by up to 50%.
Combined with an optimized encoding of the local connectivity changes,
less than 4 bits per triangle encode where and how the topological
refinements should be applied. Furthermore, we apply prediction-error
compression to quantized vertex coordinates. We estimate the position
of new vertices from the positions of their topological neighbors in
the less refined mesh using a new estimator that leads to
representations of vertex coordinates that achieve a compression ratio
of 2:1. We also provide a time efficient coding method for vertex
coordinates that allows real-time progressive mesh decompression over
limited bandwith communication.
The Implant Sprays method provides a bit-efficient and progressive
multiresolution representation of triangular mesh connectivity through
compact encoding of a series of vertex split refinement operations.
Furthermore, it also supports efficient vertex-coordinates compression
based on the topology and geometry information available from previous
LODs.
Feb. 24
Ulrich Neumann (USC)
Abstract
Numerous technical and practical challenges are encountered in producing
realistic animated models of human faces. This talk surveys these problems
and some specific solutions that arise from a "performance-driven" approach
developed in our research group. We model a human head and face from a
series of images of a real person. Video sequences of the facial
expressions produced by the same person are analyzed to extract a model of
their specific facial deformations and appearance changes. The models are
then used in an interactive session to control an animated 3D model that
mimics the appearance and behavior of the real person. Similar methods are
used to create and parameterize hair models from images. Through future
evolutions of these methods we hope to produce more realistic humans in
virtual environments, personal avatars for interactive 3D teleconferences,
and perceptive human-computer interfaces.
Apr. 11 (Tuesday)
Bruce Gooch (Utah), Matt Kaplan (Utah), Lee Markosian (Brown), J. D. Northrup (Brown), and Jonathan Cohen (Brown)
Abstracts
This talk will provide a short overview of non-photorealistic
rendering and then give a snapshot of Center activities in this area
with presentations from the NPAR conference, the first ever conference
devoted to non-photorealistic rendering (http://www.annecy.org/npar/).
Speaker: Bruce Gooch, Utah - Overview
Speaker: Matt Kaplan, Utah.
"Interactive Artistic Rendering" by Matthew Kaplan, Bruce Gooch, and Elaine Cohen
We present an algorithm for rendering subdivision surface models of
complex scenes in a variety of artistic styles using an interactively
editable particle system. The algorithm is suitable for modeling
artistic techniques explicitly by the user, or automatically by the
system. Our approach can simulate a large number of artistic effects
due to the fact that almost any type of mark made on paper or canvas
can be imitated. Any of our artistic effects is customizable by the
user through a particle editing interface. The algorithm maintains
complete frame-to-frame coherence, a characteristic required for good
animation, and runs at interactive rates on current computer graphics
workstations.
Speaker: Lee Markosian, Brown
"Art-based Rendering with Continuous Levels of Detail" by Lee Markosian, Barbara J. Meier, Michael A. Kowalski, Loring S. Holden, J. D. Northrup and John F. Hughes.
In previous work we presented an algorithm for rendering virtual scenes using
art-based styles. We demonstrated the ability to render fur, grass, and trees
in a stylized manner that evoked the complexity of these textures without
representing all their components explicitly. We achieved this with
stroke-based procedural textures that generated detail elements, or
graftals, just as needed.
Our implementation had several drawbacks. First, each new graftal texture
required a procedural implementation that included writing code. Also, graftals
were regenerated in each frame in a way that led to excessive introduction and
elimination of graftals even for small changes in camera parameters. Lastly,
our system provided no way to continuously vary the properties of graftals,
including color, size, or stroke width. Such an ability could be used to
achieve better frame-to-frame coherence, or more generally to animate graftals.
In this paper, we present a new framework for graftal textures that addresses
these issues. Our new framework allows all major decisions about graftal look
and behavior to be specified in a text file that can be edited by a designer.
We have achieved greater frame-to-frame coherence by using graftals that remain
in fixed positions on the model surface. The look and behavior of graftals as
they appear or disappear can now be animated to create smooth transitions.
Finally, we introduce the concept of tufts which manage the multiresolution
behavior of graftals according to the specifications of the scene designer.
Speaker: J. D. Northrup, Brown
"Artistic Silhouettes: A Hybrid Approach" by J. D. Northrup and Lee Markosian.
We present a new algorithm for rendering silhouette outlines of 3D polygonal
meshes with stylized strokes. Rather than use silhouette edges of the model
directly as the basis for drawing strokes, we first process the edges in image
space to create long, connected paths corresponding to visible portions of
silhouettes. The resulting paths have the precision of object-space edges, but
avoid the unwanted zig-zagging and inconsistent visibility of raw silhouette
edges. Our hybrid screen/object space approach thus allows us to apply
stylizations to strokes that follow the visual silhouettes of an object. We
describe details of our OpenGL-based stylized strokes that can resemble natural
media, but render at interactive rates. We demonstrate our technique with the
accompanying still images and animations rendered with our technique.
Speaker: Jonathan Cohen, Brown.
" Harold: A World Made of Drawings" by Jonathan Cohen, John F. Hughes, and Robert Zeleznik
The problem of interactively creating 3D scenes from 2D input is a
compelling one, and recent progress has been exciting. We present our
system, Harold, which combines ideas from existing techniques and
introduces new concepts to make an interactive system for creating 3D
worlds. The interface paradigm in Harold is drawing: all objects are
created simply by drawing them with a 2D input device. Most of the 3D
objects in Harold are collections of planar strokes that are
reoriented in a view-dependent way as the camera moves through the
world. Virtual worlds created in Harold are rendered with a
stroke-based system so that a world will maintain a hand-drawn
appearance as the user navigates through it. Harold is not suitable
for representing certain classes of 3D objects, especially
geometrically regular or extremely asymmetric objects. However,
Harold supports a large enough class of objects that a user can
rapidly create expressive and visually rich 3D worlds
Apr. 13
Russell M Taylor (UNC)
Abstract
The nanoManipulator project is a multi-year effort by Computer Scientists, Physicists and Materials Scientists to develop improved visualization and control for experiments at the nanometer scale. This talk with discuss the system, features that have been particularly useful to the scientists, experiments done with the system, what it is like to work as part of an multidisciplinary team, and recent extensions of the system for educational outreach and tele-collaboration.
Apr. 20
Doug Roble (Digital Domain)
Abstract
You've been to Siggraph. You've seen the movies. What goes on
in between? This talk will concentrate on what a large-scale
effects house considers important in computer graphics. We'll
also talk about the realities of software development and
research at an effects house.
At Digital Domain we've developed many software
packages that range from a cutting edge 2D compositor to
motion capture tools to computer vision tools to fluid
simulation. I'll discuss the development of the tools and what
artists find important.
If you've ever considered a job in the effects industry and
were wondering what it would be like, this is the talk for you!
Author Bio
Doug Roble received his Bachelor's in Electrical Engineering
from the University of Colorado and his Ph.D. in Computer
Science from The Ohio State University in 1992. His first and
only "real job" is at Digital Domain as a software
developer. He started in 1993 and is now the Creative Director
of the software department. In 1999 he won a Technical Academy
Award for his work on "track", a 3D tracking and
photogrammetry program for use in film.
Geometric Accuracy: Graphics Is Not the Only Show in Town!
Department of Computer Science & Engineering, University of Connecticutt
tpeters@engr.uconn.edu, www.eng2.uconn.edu/~tpeters/Peters.html
This broader view has software architectural implications for
geometry and graphic subsystems, as will be illustrated with
examples from the aerospace industry.
Three Methods For Detail-Preserving Contrast Reduction
Multiresolution Hierarchies for Irregular Meshes
Is the Hardware Z-Buffer Doomed?
Small- and Large-Scale Virtual Environments for Collaboration
Research Assistant Professor
Department of Computer Science, Naval Postgraduate School
capps@acm.org
http://vr.edu/capps
Coordinates, geometry, graphics, and programming
Compressed Progressive Meshes
Face Modeling, Rendering, and Animation
NPR
Multidisciplinary Science at the Nanometer Scale
Applying Graphics Research in a Production Environment
 |
 |
 |
 |
 |
 |
| Home | Research | Outreach | Televideo | Admin | Education |