Index:
Home Page
Education
Research
Sport Arena
Feature Articles
Contacts
Welcome !
To start your journey, you can find a brief history of my education on this page or read newspaper articles about me. I've completed a Ph.D in Pure Mathematics and Computer Science at the University of Adelaide, working in particular in the field of computer vision. My interests lie in theories of differential geometry and algebraic topology with an inclination to develop software for robotic and visual system applications.
Research Objectives
A common task in computer vision is the estimation of
parameters that describe a relationship between image features
across two or more views. Parameters and image data model the
geometry of the problem at hand. My project is to develop a numerical method to
estimate parameters for which the relationship with the image data
is expressed as a system of equations. It extends the
Fundamental Numerical Scheme (FNS) which
applies when the relationship is given by a single equation.
Accurately solving the underlying system of equations is vital to
ensure (a sound geometry and therefore) results of high quality.
This task can be reduced to minimising a cost function which
measures the extent to which a particular choice of parameters fails
to satisfy the system of equations for the given data. This problem
is highly non-trivial [Read more]. On a second
level, my work involves studying the mathematical foundations of
other parameter estimation methods and implementing
them to compare performance.
Example Applications
Analysing videos means dealing with multiple images. Deriving the
geometry that relates all the views one to another is fairly complex
(e.g: use Grassman-Cayley algebras). Not only the relationship
between unknown parameters and data is described by a system of
equations but many ancillary constraints apply, and it's not clear
how to enforce them in the most effective way. In general, people
examine image sequences by looking at pairs (stereovision) or
triplets of views (trinocular
vision).
Parameter estimation underlies many applications which involve
camera calibration, video analysis and tracking, virtual reality
animation, automatic techniques for movie special effects or PC
games. It also appears in structure from motion problems. For
instance, estimating the homography parameters to compose a
panoramic mosaic, or the trifocal tensor parameters to construct a
3D model of a scene by analysing groups of three images at a time.
These last two applications provide the main testing ground for my
numerical method, see links below.
Publications
Received
"Best publication award"