 |
|
Event(s) on October 2005
- Tuesday, 25th October, 2005
| Title: |
Tracking Edges, Corners and Vertices in an Image. |
| Speaker: |
Dr. Christian Rau, Department of Mathematics, Hong Kong Baptist University and, Mathematical Sciences Institute, Australian National University, Australia |
| Time/Place: |
11:30 - 12:30
FSC1217
|
| | |
| Abstract: |
An ever-recurring task in image processing and spatial statistics
is the recovery of edges. We interpret an edge as a curve of
discontinuity points in an otherwise smooth regression surface
observed with noise, with the design points representing planar
observation sites. In earlier work, we developed an estimator
of a single (connected) edge, which is based on the idea of
'tracking' the edge through the plane once a point near enough
to it has been found. This method uses a locally parametric model
and least-squares to estimate the tangent at a given point on
the edge. In most cases of practical interest, however, the
edges may join at points called 'corners' and 'vertices' at which
the edge has no unique tangent, and tracking may either introduce
bias or miss part of the edge altogether. In this talk, we extend
the tracking approach to cater for this situation. A theoretical
analysis as well as artificial and real-world numerical examples
are given, along with some open problems for future research.
|
|
|
|
|
