MultiGrid Methods in Image Processing
Speaker: Simon Long, ITEE
When: 2004-12-07 14:00:00
Venue: 78-420
Host: Brian Lovell
Abstract:The field of image segmentation is concerned with deriving the
boundaries of regions of interest in an image, and is of particular
importance in narrowing the range of data that must be processed in
higher level applications, reducing the overall computation in those
stages. By its very nature it operates on large data sets, and with
the increase in resolution and dimensionality of modern data sets
segmentation requires techniques that operate with low orders of
computational complexity.
The implementation of algorithms as multi-resolution methods is
already recognised to exhibit less complexity than those operating
on a single resolution. Many such combinations are thrown together
as an afterthought and accept as a tradeoff for speed increases a
decided loss in accuracy in their converged solution when compared
to that yielded by the single-resolution method. Multigrid methods
offer a rich theoretical basis in numerical analysis that provides
the means to implement a multi-resolution approach without a loss of
accuracy, with considerable improvements in the rate of convergence.
The process of anisotropic diffusion-reaction is one preprocessing
method for segmentation for its denoising and image simplification
properties. From the investigation of implementing multigrid
diffusion-reaction, the research aims to generalise a
multigrid-style, improved multi-resolution method for general
optimisation problems in image processing.
Biography:Simon Long is a PhD candidate in the IRIS group of the EMI division.
Type: Ph.D confirmation
Contact:Brian Lovell, seminar host (lovell@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)