Computer Graphics Seminar—Fall 2006

CS 718
Mondays, 12:15–1:15 pm  (.... no longer Fridays)
551 Rhodes Hall (PCG Conference Room)

The talks in the Graphics Seminar stem from current research, and they can be of various types:

Talks about research that is already written up and submitted are generally not allowed; you should present work in seminar when you are still ready to get feedback on it.

Please sign up to the mailing list for future seminar announcements:

Schedule for Fall 2006

Fri September 1
Doug James
Introduction to Graphics Seminar
Fri September 8
Ganesh Ramanarayanan
Color Correction and Google Maps / Earth

ABSTRACT: Recently, companies such as Google, Yahoo, and Microsoft, as well as local companies in various countries, have been putting together sophisticated mapping tools allowing users to view and interact with the world at a large number of scales, from images of continents/countries to satellite/aerial photography showing people's houses and backyards. There are two main challenges with this imagery: 1) acquisition of good data on a global scale, and 2) integration of the data into a seamless whole. I will talk about these issues as they relate to Google Maps/Earth, and show some comparisons with Yahoo Maps Beta, Microsoft TerraServer, and other products. Then, I will talk briefly about addressing challenge 2) through color correction algorithms, focusing on recent colorization / color transfer papers that do not depend on models of the image capture device. There may not be time but I will try to cover the following:

Fri September 15
Jon Kaldor
Discrete Differential Geometry and You

ABSTRACT: Despite its ominous sounding name, discrete differential geometry (DDG) is actually an attempt to make it easier to think about and simpler to implement certain physical simulations (for instance, cloth simulators).  Rather than formulating equations over some idealized continuous surface and then only solving on a mesh as a last step, DDG argues that the equations should be generated directly for the mesh itself.  This results in a convincing simulation with some surprising properties and fewer mathematical hoops to jump through.

In this talk, I'll be discussing the basic ideas behind DDG, including some of the intuitive basis for why it works and how it builds on previous cloth simulations, all the while hopefully keeping the intimidating equations to a minimal level (I hope to achieve a cute-bunny-picture-to-incomprehensible-equation ratio of greater than 1). The talk will primarily center around the Discrete Shells paper, with several
references to other work both prior to and since then.

Fri September 22
Christopher Cameron
Toward fast frictional dynamics for rigid and deformable models

ABSTRACT: Fast algorithms for dynamics simulations with contact and friction using rigid bodies have been studied extensively in computer graphics while algorithms for simulations with deformable bodies have not received as much attention.  In this talk I will discuss rigid body simulation techniques and how I plan to extend them to deformable multibody systems.
Mon October 2 David Kaplan  and  Brendan Holt
Foundations, Flaws, and Fish in Computer Animation

ABSTRACT: 3-D computer animation draws heavily on traditional animation concepts. We present these concepts as well as the pipeline for creating a 3D-animated short using industry-standard software.  We will use our own work-in-progress, a four minute film about a fish, to provide insight into some of the benefits and limitations of modern animation techniques, and outline solutions to some of the obstacles we encountered.
Mon October 9
No seminar
Fall break
Mon October 16
Adam Arbree

ABSTRACT:  In 1997, Veach and Guibas introduced a novel rendering technique for Monte Carlo rendering.  To this day, this rendering technique is unique because it samples the space of possible light paths based on their contribution to the input image rather than on their power.  It does this by a process of mutation.  When a high contribution path is found, it is mutated slightly; hopefully discovering nearby strongly contributing paths.  By careful construction of the mutation rules and their selection probabilities, the samples will have a probability distribution exactly proportional to their image contribution.

In this talk I will review the development of the Metropolis Light Transport system beginning with importance sampling, then the basic Metropolis Sampling algorithm developed in 1953, the requirements for unbiased sampling, and a discussion of the mutation rules.  Finally, as time permits, I will discuss more recent uses of Metropolis for volumetric rendering and photon mapping.
Mon October 23
Todd Alan Harvey
Fur and Hair: Structural representation in feature film

ABSTRACT:  Fur and Hair-like structures are a common part of living organisms.  Visual representation of all types of creatures, both human and non-human require a hair structure implementation in order for the representation to be believable to the human perception system.  Researchers and artists working in the field of computer graphics have sought over the years to define working models for describing the character of hair.  a brief overview of working models of fur and hair in motion picture production will be presented.  the main body of the presentation will draw upon fur and hair development at weta digital for the production of king kong (2005).   i will include a review of steve marschner's research in the context of its application in the film industry and specifically to the production of the digital characters, King Kong and Ann Darrow.  i will also present some limitations of current models employed and practical compromises reached during production.
Mon October 30
No seminar
DreamWorks visit: 
Mon November  6
No seminar
Mon November 13
Reynald Dumont
Overview of TEAMLOG/PRISM (Physics, Radiation, Image, Simulation, Modeling)

ABSTRACT:  At TEAMLOG/PRISM, our know-how is the physical modelling and simulation of an environment seen by a sensor (Camera, Electro-Optic systems, Radar, SAR, Sonar, etc).

These simulations consist in calculating the propagation of electromagnetic waves in a digital geometric and physical model, meant to represent as faithfully as possible the reality of a given scene.

A thorough knowledge of the geometry, including buildings, objects, vehicles, and targets, as well as the characteristics and physical
properties of the materials, are indispensable.

Based on several R&D studies, and many years of experience, our experts (PhD, MSc), have developed new technologies and software, dedicated to scene generation, scenario editing, image/signal computation and analysis.

Our research efforts concentrate particularly on understanding the interactions between electromagnetic waves and the atmosphere / scene

In this talk I will detail some fundamentals of our simulation processes and their similarities with computer graphics (in particular Ray Tracing and Global Illumination). Finally, I will review the development of our simulators and provide some insights of our work in progress/future work
(with respect to the computer graphics literature).

Mon November 20 & 27
No seminar
Thanksgiving recess

Mon December 4
Milos Hasan

Previous semesters:

Doug James (