David Squire

• Image Processing and Computer Vision
  
  • use of Bayesian models, combining Gaussian mixtures and Markov random fields, for image segmentation, applied to dermatoscopic images and aerial photographs
           Code to generate samples from a multidimensional Gaussian mixture
  • application of independent component analysis to the learning of texture features, with a focus on dermatoscopic images
  • invariant pattern recognition
  
  
• Content-Based Multimedia Retrieval
  In collaboration with the Vision Group at the University of Geneva, I work on the Viper project, in which we investigate algorithms, data structures and image representations for content-based retrieval of images and video sequences from large databases. There is a particular emphasis in this project on the interaction of the user with the database, and the use of user-provided information (i.e. relevance feedback) for the design and incremental improvement of the database system, on a variety of time-scales. We are also interested in the application of techniques inspired by text retrieval to content-based image retrieval. Areas of particular interest include:
  • Performance evaluation - find out about the Benchathlon CBIR benchmarking contest
  • Communication and querying protocols , in particular MRML
  • Relevance feedback
  • Term-weighting schemes
  • Comparison of human and machine judgments of image similarity
  
  The Viper project has now given rise to The GIFT, the GNU Image-Finding Tool. The GIFT is an open framework for content-based image retrieval. If you are interested in using this tool or contributing to its development, please visit The GIFT homepage. You can also try out a demo.
  
  
• Plagiarism Detection
  I am interested in the application of web searching and text indexing techniques to the detection of plagiarism in documents. This work has led to the development of the Damocles system.
  
  
  
  
  
  
  
  
• Formal Context Analysis
  This project aims to develop a model for context, termed Formal Context Analysis. This model will be based on the notion of context as formalised in Formal Concept Analysis (FCA)—a branch of applied lattice theory. In FCA, contexts are collections of objects, given a priori, with a fixed static characterisation of attributes. Our extension will make contexts first-class objects, will allow objects to be characterised in multiple incomplete contexts simultaneously, and in rapidly changing contexts. Statistical similarity of contexts will be modelled to cater for handling of very large numbers of objects in large numbers of overlapping contexts and for dealing with context independently of meta data.

Recent:

Nabeel Mohammed and David McG. Squire, Efficient and Accurate Independent Component Filter-based Features for Texure Similarity, In Proceedings of the 20th IEEE International Conference on Image Processing (ICIP), Melbourne, Australia, September 15-18 2013. (to appear)

A full list of publications is also available.

• PhD
  

  Loke Kar Seng (co-supervised with Poh Aun Lee) 2006-
  Image Recognition and Retrieval

  Robyn McNamara 2007-
  (co-supervised with Ian Mitchell)
  Assessing Programming Ability in Introductory Computer Science
  Joint Publications

  Bin Liu 2009-
  (co-supervised with Geoff Webb)
  Learning Semi-Naïve Bayesian Classifiers from Numeric Data

  Nabeel Mohammed 2010-
  (co-supervised with Peter Tischer)
  Development of Collection-specific Texture Features for Image Retrieval
  Joint Publications

• Past graduate students
  

• Current
  

  2011, 2nd semester: FIT2043 Technical Documentation
  2008-2011, 1st and/or 2nd semester: FIT3077 Software Engineering: Architecture and Design
  2007-2011, 1st and/or 2nd semester: FIT2024 Software Engineering Practice
  

• Past
  

  2007-2008, All year: FIT4002 Software Engineering Studio Project Supervisor
  2000-2005 & 2007, 1st semester: CSE3308 Software Engineering: Analysis and Design
  2000-2005, All year: CSE3200 Industrial Experience Project Supervisor
  2000-2005: Introduction to LaTeX and citation practices for honours students, as part of CSE417
  2000-2004, 2nd semester: CSE5230 Data Mining

  I have taught and tutored physics, mathematics, artificial intelligence, software engineering, and computer science subjects at other institutions.

• Computer Vision Group, Université de Genève, Switzerland
• Vision Research Group, Curtin University of Technology, Australia
• Computer Vision and Machine Intelligence Laboratory, University of Melbourne, Australia