Monday 9 March 2009
Mathematical Challenges in Computer Vision
Organisers: Ignacio de Gregorio (Transvisual media, Warwick University Science Park) and David Mond (Mathematics Institute)
All talks will be in Lecture Room B3.02 Mathematics Institute, Zeeman Building
- 10.00-11.00 David Hogg (Leeds) Activity analysis from video ( pdf of talk)
- 11.00-11.30 Coffee in common room
- 11.30-12.30 Peter Giblin (Liverpool) Curves and surfaces in computer vision ( pdf of talk)
I shall describe some or all of the following applications of geometry and singularity theory to computer vision: medial axes and their big brother, symmetry sets; apparent contours and geometrical features of surfaces, including shading, as a clue to shape; higher order differential invariants of surfaces. I shall attempt to give an overview of some recent work.
- 12.30-13.45 Buffet lunch in common room
- 13.45-14.45 Tardi Tjahjadi (Warwick) Recovery of 3-dimensional scene information from video images ( pdf of talk)
The talk will address practical issues associated with producing a geometrical model of a scene from images grabbed using a video camera. The issues include camera calibration, features extraction and correspondence analysis. The issues are illustrated in the following approaches to 3-dimensional object reconstruction from video images: structure from motion, the combined method of optical flow and stereo vision, and shape from silhouettes. Two example applications that will be presented are collision avoidance for autonomous vehicle and reconstruction of volumetric model of sculptures.
- 14.45-15.45 Mark Nixon (Southampton) Advances in image processing and feature extraction
This talk will survey the current state-of-art in Feature Extraction and Image processing, especially as applied in new biometrics like gait and ear. The talk will survey current approaches, highlighting their advantages and limitations, with much coverage derived from the new Second Edition of my Computer Vision Textbook (www.ecs.soton.ac.uk/~msn/book/ ). The talk will then progress to the development of new physical analogies for image feature extraction, namely the use of force, water and heat as feature extraction paradigms with comparison to extant approaches. The talk will use medical image analysis and biometrics as example applications needing feature extraction techniques. This will largely be in principle though the talk will also cover the identification of people, biometrics, by the way they walk (their gait) and by their ear structure.
- 15.45-16.15 Tea in common room
We are grateful for financial support for the meeting from Transvisual Media