I am a member of the
Visual Systems Lab in the
Group at York. This page summarizes my research interests. Most work is
done in collaboration with others. Links that look like this:
[*] take you to a relevant publication reference
where the students and colleagues who have worked with me are co-authors.
Other links (the ones not in square brackets) take you deeper into this
section of the site.
Face Image Processing
- A new scheme for covariance matrix regularization
gives excellent results when applied to appearance-based face processing.
Using conditional density estimation, 3D face structure can be recovered
from mugshots without
explicit modelling of shape
- We are investigating methods for speeding up the training phase of
cascaded classifiers that use Haar basis functions for face detection.
- Earlier activity in face image processing is summarized
here and related pages (follow the links under "IMAGING HUMANS").
Motion Picture Parsing and Summarization
[*] is a system for
real-time projective transform estimation and mosaicing
between frames of a video sequence.
We are developing efficient plane+parallax 3D estimators that
will extend SAM to make frame-to-frame depth estimates.
The long-term objective
is a smart camcorder that forms models of scenes as you record them.
We are applying SAM plus hierarchical classification to
automated [*] and
semi-automated [*] movie analysis,
and developing reverse storyboarding
methods [*] for video summarization.
Image and Video Coding
- Adaptive Prediction Trees (APT)
[*] is a
high-performance still-image compression scheme, based on interpolative
predictive coding. As well as continuing to make
improvements to the core method, we are applying APT to 3D-image
[*] and video coding.
- In object-based coding our recent publications
[*] are about motion
segmentation and region description.
- We have developed a number of
Video-Augmented Environments including
[*] and the Audiophoto Desk
[*]. We're continuing
to develop VAE applications and the supporting
technology, particularly reliable image analysis algorithms for all lighting
conditions, and robust marker recognition for customizable augmented objects.
Most of my research is done using