Back to Research Groups Page

Computer Vision and Pattern Recognition

We work in the areas of computer vision, image processing and analysis, and, statistical and structural pattern recognition. Specific research areas include object recognition (including face analysis and recognition); 3D shape analysis from 2D images; brain image analysis (including tensor MRI and MEG); tracking; watermarking; quantum information processing; spectral graph theory; machine learning and reflectance modelling.

Group Aims

The group undertakes research in the mathematical development of novel algorithms for computer vision and pattern recognition. This rigorously underpinned work draws on methods from differential geometry, information theory, probability and statistics, optimisation theory and optical physics. Two important themes running through the research are the development of novel pattern analysis algorithms for non-vectorial data in the form of graphs, trees and strings, and the development of methods for recovering 3D scene structure from 2D images. The resulting algorithms have found application in a number of challenging real world problems including radar image analysis, brain imaging, chemical structure matching and image retrieval from large image databases.

Senior Member: Edwin Hancock

Contact Person: Edwin Hancock

Back to Top

Research Areas

  • Computer vision: pattern recognition, image processing, machine learning, biometrics.
  • Machine learning with non-vectorial data: using ideas from spectral graph theory, description length and information theory to develop generative and discriminative models to data represented using graphs, trees or strings.
  • Object recognition: using shape and structure to recognise objects in complex scenes, object recognition from large data-bases and video streams.
  • Shape-from-shading and texture: developing statistical methods to recover 3D object shape from single images containing shading or texture patterns.
  • Motion analysis: robust recovery of structure from motion, motion field analysis, object tracking using particle filters.
  • Face image analysis: recovery of 3D facial shape from 2D images, face recognition, facial image synthesis, gender and ethnicity determination.
  • Reflectance modelling: development of theoretical and empirical models for light reflectance from complex non-Lambertian surfaces, the used of polarisation information for 3D shape recovery, modelling skin reflectance.
  • Image watermarking: statistical and geometric methods for watermarking 2D and 3D images, steganography.
  • Image analysis: diffusion smoothing of non-Euclidean image data, analysis of vector and tensor fields.
  • Brain imaging: algorithms for fibre tractography from diffusion tensor MRI data, MEG data analysis, MEG shape perception and recognition experiments.
  • Quantum computing: algorithms based on quantum walks for graph isomorphism and graph clustering.

Sponsorship is received from a variety of sources including EPSRC, EU and Qinetiq.

Back to Top

List of Collaborators

  • Australian National University
  • Delft University of Technology
  • ETHZ
  • IST Lisbon
  • Moscow State University
  • University of Utah
  • University of the Balearic Islands
  • University of Venice
  • University of Verona

Back to Top

Academic Members of the Group

Back to Top

Facilities

We have access to a Cyberware 3030 head scanner, reflectometetry, polarimetry and MRI/MEG via the York Neurimaging Centre.

Back to Top

Recent Prizes

Pattern Recognition Society Medal and Outanding paper award, best paper awards at ACCV 2002, CAIP 2001, ICPR 2006 (Piero Zamperoni Prize) and BMVC 2007 (Siemens Security Paper Prize).

Editorial Boards

IEEE TPAMI, IEE TNN, Pattern Recognition, IET Computer Vision.

Back to Top