I am going to give a brief overview of three papers from this year's Siggraph that I found imressive/interesting. The titles of the papers are: 1) Plenoptic Modeling: An Image-based Rendering System by Leonard McMillan and Gary Bishop 2) Intelligent Scissors for Image Composition by Eric Mortensen and William Barrett 3) Physically Based Glare Effects for Digital Images by Greg Spencer, Peter Shirley, Kurt Zimmermann and Donald Greenberg
MASTER'S THESIS PRESENTATION
-Friday, August 25, 1995
Wayne Liu, graduate student, Dept. Comp. Sci., Univ.
Waterloo, will speak on "Programming Support For
Blossoming".
TIME: 10:30 a.m. - 11:30 a.m.
ROOM: DC 1331
ABSTRACT
In this talk, I will present a system that uses
blossoming as a programming tool for implementing
geometric modeling systems. I will discuss 1) why a
programming tool based on blossoming is desirable, 2)
how such a tool can be created and previous work on
creating such tools, 3) what are the features of the
tool, and 4) how it is used to perform curve and
surface computations.
Blossoming has been shown to be a powerful technique
for analyzing many curve and surface paradigms. Not
only has it been used to derive properties of these
paradigms, but it has also been used to describe their
algorithms in a simple and unified manner. For
example, many algorithms have been shown to be special
cases of the generalized de Casteljau algorithm for
evaluating blossoms.
This ability to unify and simplify algorithms suggests
that blossoming can be made into a useful programming
tool. Such a tool would allow blossoming analysis to
be translated directly into code. Thus, the code would
be easier to understand, to write, and to update.
Several researchers have proposed methods for creating
such a tool. The idea is to create a datatype for
blossoms. This datatype would allow programmers to
perform operations on blossoms without needing to worry
about how the operations are implemented. The blossom
operations would then be used by the programmers to
perform more complicated computations.
Previous systems created by these researchers have
shown that using a blossom datatype simplifies code.
Efficiency was not considered when designing these
blossom packages, as the code complexity was the
primary concern. However, efficiency is of great
importance when building real systems. (The efficiency
of these previous systems were limited because only the
blossoming operation provided was evaluation.)
A new system has been created based on similar ideas,
but with the goal of providing an efficient and general
tool to build geometric modeling systems. This system
is a C++ library with datatypes for blossoms. It
provides new basic operations that are more efficient,
including setting and extracting the control points of
a blossom, and an operation similar to knot insertion.
These basic operations makes it possible to perform
general computations simply and efficiently.