Computer Vision Research Projects

Omnidirectional Vision
Omnidirectional vision involves the capture and analysis of images with full 360 degrees view of the surroundings. It has a number of significant benefits. In particular, not losing sight of objects from rotating vehicles in dynamic environments makes omnidirectional vision very useful for Autonomous Navigation, Guidance, Mapping, and Robotics. For more details, see my Omnidirectional Vision Project.
Edge Detection
Edge detection is a fundamental part of computer vision. It needs to provide good quality input to numerous other vision methods. Despite its long history, it contiues to be of interest and probably will always remain so. Some issues: optimality, efficiency, the best use of colour and texture.
Sample Images:An edge map of human face., An edge map of urban landscape.
General Purpose Recognition and Face Recognition
Face Location and Recognition using low level features descriptors derived from edges, textures, colour and curvatures. We are particularly interested in the invariance and stability of visual descriptors.
Sample Publications: D.Hond, L.Spacek: `Distinctive Descriptions for Face Processing', 8th BMVC, England, pp.320-329, 1997.
Face Recognition Data: I am making available here a large collection of facial images (7900 in total) for research and testing purposes. Go to the description and download of the face recognition data.
Historical Notes: teaching and research in Computer Vision was started here in mid 1970's by Dr Mike Brady (later at MIT and professor of Information Engineering at Oxford).
Hall of Fame: some PhD students supervised by Dr Spacek in the past.