Research
Ilios
Research and Design have keen research interests in a number of areas
including video compression, image analysis, and natural language
programming and are continually pursuing leading-edge research in these
areas.
We are always looking at opportunities for potential
commercial applications for our research.
Our
current research directions include :
Image
segmentation
Extraction
of objects and object boundaries from image sequences, primarily by using
dense motion fields, and assisted by other information such as luminance
discontinuities and colour information. The question of how to combine
these disparate sources of information is another interesting problem we
are investigating.
Stereoscopic
Imaging
Extraction
of 3D structure from a stereoscopic image pair. This is a promising
research area. It involves comparing features in each image, and
calculating depth by determining the horizontal translation of the feature.
The problem of deriving a complete 3D image from a dense set of features, remains
a difficult problem, and one which we are currently researching. This has
potential application in many areas including robotic vision, and driver
assistance in the automotive industry.
Natural
Language Programming
This
research topic has grown from our recent experience in the telemetry /
industrial automation industry. The
industrial automation industry has standardised on a number of languages
for programming PLCs (as specified in IEC61131-5 standard) including
function block diagrams and ladder logic diagrams. While this has
simplified the programming task from previous proprietary solutions, it
still remains a difficult task only understood by well trained field
engineers. We have been investigating the applicability of natural language
for such programming tasks. This is an old research topic dating back over
a number of decades; however recent advances in automated grammar
processing and data mining technology have once again made this a viable
proposition.
James Wiles
BE (Hons) (Monash) 1985
Master
of Engineering Science (Monash) 2001
This
involved part-time research study at Monash University,
in the field of digital vision and image analysis. My thesis title was
“Towards Automatic, Accurate Motion Estimation in Video
Sequences”.
The
research team at Monash
University was, and
continues to be, a world leader in the field of video compression, and
included a number of academics and students who are actively involved in
the MPEG steering and technical committees, with whom I worked closely.
My
work built on my previous experience working with digital video and the
MPEG 1/2/4 coding standards. It involved research into image segmentation
which was an important field of research for the MPEG-4 standard, which was
under development at the time. The MPEG-4 standard has a mode which allows
for individual objects in a scene to be coded separately; however it
provides no guidance for how object segmentation should be performed in a
video sequence. My research directly addressed this issue, and resulted in
a promising technique for using a block-based motion estimation field to
determine consistent regions of parametric motion, and then accurately
extracting the objects which closely conformed to this parametric motion.
Selected Research Interests :
Stereoscopic Vision for 3D modelling, Digital Video
Coding, Image Segmentation, Automated Image Analysis for enhanced
understanding of image content, Intuitive Logic Programming
Selected Publications :
J. Wiles and K. Pang, "A Motion-Based Image
Segmentation Scheme", 6th IEEE International Workshop on Intelligent
Signal Processing and Communication Systems, Nov. 1998
J. Wiles and K. Pang, "Towards True Motion
Detection and Compensation in Video Coders", Globecom: IEEE Global
Telecommunications Conference, 2000
|