ITEE Home » Seminars » Archive » 2003 » Seminar: Image Segmentation

Print version

ITEE Ph.D confirmation seminar: Ben Appleton, 11.00AM, Tue 11 Mar 2003

Image Segmentation

Speaker: Ben Appleton, ITEE

When: 11.00AM, Tuesday 11 Mar 2003

Venue: 78-622

Host: Associate Professor Brian Lovell

Abstract:

  Segmentation is the fundamental first step toward analysing and
  understanding images. It may be defined as decomposing the image
  into its constituent parts, extracting the location and outline of
  objects of interest in an image. However the definition of an
  "interesting object" is dependent both upon the application and the
  image modality.  Additionally, many applications demand speed at the
  expense of accuracy.  As a result there is a wide spectrum of
  segmentation methods, each suited to a subset of segmentation
  problems and application environments.

  In this thesis we focus on segmentation problems presented by
  biomedical data. We consider planar data acquired by x-ray,
  ultrasound and optical microscopy, as well as volumetric and
  time-series data from Magnetic Resonance and Computed Tomography
  images.

  Biomedical data is characterised by high geometric variation both
  within and between patients or samples. Organs deform between
  observations due to physiological changes or simple displacement,
  cells grow and split, and often we are searching for pathological
  instances which by their very nature defy classification. In many
  cases even human experts differ in their judgement of a good
  segmentation. Despite the bleak outlook, it is these difficulties
  which are being overcome at the forefront of medical image
  segmentation.

  We will investigate a class of segmentation methods suited to many
  biomedical image analysis problems. These geometric optimisation
  methods are designed to use minimal structural information,
  instead aiming to find coherent curvilinear features which are
  deemed likely to be the boundary of an object of interest. Commonly
  known as active contours, they include the discrete active contours,
  snakes and level sets. Of particular interest is the geodesic active
  contour, which has been shown to give reliable segmentation
  results. We build on previous work in geodesic active contours to
  give optimal solutions to the associated minimisation problem,
  producing an algorithm that is significantly more robust and
  accurate on a range of segmentation problems. We also extend the
  geodesic active contour to extract open and closed geodesics in
  spaces with tensor-valued metrics.  

Biography:

 
  Ben Appleton is currently undertaking a PhD in the area of Image
  Analysis at the University of Queensland, Brisbane, Australia.  He
  received his bachelor degrees in Science and Electrical Engineering
  from the University of Queensland in 2001.  His research interests
  include active contours and energy minimisation algorithms.

Type:

Ph.D confirmation

Contact:

Associate Professor Brian Lovell, seminar host (lovell@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator) (guido@itee.uq.edu.au)

ITEE seminar web page: http://www.itee.uq.edu.au/~seminar

---

[All seminars]