|
Bob Dony: Research
Much of my research interests focus on the use of neural network
and other adaptive techniques for image and signal processing.
A novel neural network, the mixture of principal components
(MPC), has been successfully used in image compression within a
block transform framework. Because of its locally adaptive
nature, it can out-perform the globally optimal Karhunen-Loeve
transform (KLT). The network has also been applied to colour
image segmentation and in a neuro-fractal image compression
scheme. Other work in this vein has seen the development of a
nonlinear, neural network based wavelet coding method.
Another focus of my work has been medical imaging. Some of the
applications for the above compression methods has been on
medical images. Other work includes multimodal image
registration, and the analysis of equine tendon sonograms.
Extending the neural approaches from images to audio signals has
yielded some promising results. We have been using neural
network architectures and other nonlinear, adaptive techniques
for speech enhancement and noise reduction. Of particular
interest is the application of such advanced DSP techniques to
digital hearing aids.
A final interest of mine is in image mosaicing and 3D
reconstruction from moving monocular video. This interest stems
from work I did for the National Geographic Magazine in
2000, the result of which was the publishing of a mosaic of the
USS Arizona shipwreck sunk at Pearl Harbor in the June,
2001 issue. My current work in this regard is the extraction of
a three-dimensional relief from unconstrained moving monocular
video.
Projects
Biomedical Image and Signal Processing
|
Medical Image Registration
There has recently been considerable interest in the medical
image processing community, in the use of voxel similarity
measures for automatic registration of images from the same, or
different modalities. Currently we are using mutual
information as the error metric for 3D multimodal image
registration for both rigid and elastic transformations.
|
|
Ultrasound Texture Segmentation
|
|
Speech Enhancement and Noise Reduction
|
|
Hardware/Software Co-design for Intelligent Noise Reduction
|
|
Vsualization of Heart and Body Cavity Sounds
|
3D Vision
|
Real-time Corner Detection on a Single FPGA
|
|
Improved Calculation of Essential and Fundamental Matrices
|
People
Current Graduate Students
Reggie Rowlandson B.Sc. (University of Guelph)
|
Ying Huang B.Eng. (Zhejiang University, Zhejiang, China)
Ying Huang was born in 1971 in Fujiang, China. He recieved
B.Eng. degree in electrical and electronics engineering from
Zhejiang University, Zhejiang, China in 1993. Following his
graduation, he worked for 7 years as research engineer in
various companies, mainly focused on medical signal processing
and embedded system design. From 2002, he has been pursuing
M.Sc. degree in the Engineering System and Computing program in
the School of Engineering at University of Guelph, Guelph,
Canada. His current research interests are in time frequency
analysis, blind signal separation, neural network and speech
signal enhancement.
|
Qi (Chee) Huang B.Eng. (Zhejiang University, Hanzhou,
China)
Qi Huang received the B.Eng. degree in Biomedical engineering
& instrumentation from Zhejiang University, Hanzhou, China,
in 1991. Since 2002 he has been studying in the Engineering System &
Computing program in the School of Engineering at University of
Guelph for the M.Sc. degree. His main research interests include
signal/image processing, medical imaging, theory and
applications of neural network, and machine vision.
|
Wenxin (Wendy) Wang B.Sc. (School of Aeronautics,
Engineering of Aeronautics and Astronautics University, China)
Wenxin Wang received the B.Sc.. degree in Aeronautics
engineering Engineering of Aeronautics and Astronautics
University, China. From 1991-1998, she was with Beijing
Institute of Space Machinery and Electricity, Beijing, China,
working on systems analysis and image quality control using MTF
for sensing satellites. Fom 1999-2000, she was with TTS
Computer Company, North Carolina, USA and from 2001-2002 she was
with Batacom Crop. Toronto, Canada working on constrast
enhancement for video image processing and CMSO smart glass
design. Since 2002 she has been studying in the Engineering
System & Computing program in the School of Engineering at
University of Guelph for the M.Sc. degree. Her main research
interests include hardware implementation in FPGA of feature
extraction algorithms for real-time machine vision. |
Xiaoyan (Anna) Xu B.Sc., M.Sc. (Beijing Normal
University, Beijing, China)
Xiaoyan Xu received her B.Sc. and M.Sc degrees in Radio
electronics and Telecommunation from Beijing Normal University
in 1998 and 2001 respectively. Since 2002 she has been persuing
her PhD studies under the direction of Dr. Robert Dony at
University of Guelph. The emphasis of her research work is
multimodal medical image registration. Her interests include:
digital signal/image processing, computer vision and pattern
recognition, reconfigurable computing, telecommunication,
fractals.
|
Past Graduate Students
|
Email
