 |
[a] Dr. Farag research focus is image understanding with
practical applications. His contributions have been in image modeling,
segmentation, registration, and object reconstruction. Specifically,
- Image modeling: This problem deals with describing the
appearance (gray level) and spatial interaction (textures) of an image.
Dr. Farag’s contributions include: i) introducing a generalized linear
model for density estimation from the gray level histogram using a novel
linear combination of Gaussian kernels, with consistent estimation of
the parameters, including the number of modes and the kernel parameters;
ii) developing a local estimation approach for the Gibbs distribution in
which the optimal neighborhood topology and the corresponding clique
potentials are automatically estimated from the observed image; iii)
developing a new vector level set distance representation for describing
the surface evolution; iv) developing an accurate solution of the
Eikonal Equation on the Cartesian domain; v) developing a new centerline
extraction approach for shape representation that links the objects
topological nodes; vi) developing a new shape representation approach
using vector level set function; vii) developing a new functional
representation function for the external energy in the deformable model
(snake) representation; and viii) developing a new approach for
estimation of the Support Vector Machine (SVM) classifier, and a novel
approach for estimating the weights based on the mean field theory.
These contributions have been manifested in novel and robust approaches
for multimodality image segmentation and registration, and have been
used in a number of computer vision problems in the past 17 years.
- 3-D Surface Reconstruction: This problem is known in the
computer vision literature as ``Shape from X''. For example,
depth-inferring techniques like stereovision, Space Carving, and shape
from shading (SFS) have been used to generate 2D to 3D mappings of
visible surfaces. Dr. Farag has contributed to the use of stereovision,
range finders and structured light for robust 3-D surface
reconstruction. Dr. Farag has also designed and implemented a well-known
trinocular active vision system (CardEye) for the studies of the shape
from X paradigm. Our contribution to the SFS problem includes
generalization of the image formation problem by removing the
constraints on the object with respect to the camera and light source,
and removing the Lambertian restriction on the objects’ surface model.
We have also devised an approach to calibrate the surface generated by
the SFS using range information at landmarks; a US patent has been
issued to this invention.
- Surface and Volume Registration: This problem deals with how
to correlate a data set to another one sharing some common attributes.
This is a fundamental problem in remote sensing, computer vision and
medical imaging. Our contributions include the Surface Point Signature (SPS)
approach, and a novel elastic registration framework based on partial
differential equations.
- Computer-Assisted Diagnosis (CAD) Systems: For over a decade
The CVIP Lab has been engaged in developing and validating CAD systems
in collaboration with several biomedical scientists and clinicians
worldwide. In particular, we developed comprehensive CAD system for
automatic detection of lung nodules from low dose CT scans; CAD system
for quantification of the Kidney function after a transplant using
dynamic contrast enhanced magnetic resonance imaging (DCE-MRI); and a
CAD system for virtual colonoscopy. These systems have been published in
the relevant literature including MICCAI, IPMI, CARS and ISBI.
The above problems are among the most popular and active in the field
of computer vision, image processing, machine learning and medical
imaging. Dr. Farag has authored or co-authored over 250 articles on the
above subjects with good citations.
[b]
Dr. Farag is the founder (1994) and director (1994-pesent) of the
Computer Vision and Image Processing Laboratory (CVIP Lab). Among the
impacts of the CVIP Lab are the following:
- The CVIP Lab is the vehicle that brought supercomputing, immersive
visualization, autonomous robotics, high bandwidth networking, and
multidisciplinary bioimaging research to the University of Louisville.
Facilities at the CVIP Lab are used by university wide audience;
researchers from engineering, medicine, psychology, physics and
mathematics collaborate with Dr. Farag and use the facilities of the
CVIP Lab.
- The CVIP Lab is a popular visiting site for local high schools,
which impacts positively on student recruiting to the University and in
making younger students develop interests in science and engineering.
- In addition, the research at the CVIP Lab was cited in an official
media release by NSF http://www.eng.nsf.gov/engnews/1998_News/computer_vision_for_better_med.htm,
Silicon Graphics, Inc. http://www.sgi.com/features/2003/jan/louisville/
, and the Louisville Science Museum.
[c]
Dr. Farag's funded research projects and initiatives have a very
positive impact on the entire University of Louisville's research
infrastructure. He was responsible for bringing the first supercomputer
(24CPU SGI ONYXR10000) to the University of Louisville in 1996. In
collaboration with the University of Louisville's Office of Information
Technology, Dr. Farag is responsible for linking the University of
Louisville to the NSF high-speed research network vBNS. In addition, Dr.
Farag's funded projects are also responsible for linking the University of
Louisville's Belknap Campus and the Health Science Campus with higher
bandwidth network. Enormous commitments from the University of Louisville
helped to make these initiatives a reality.
[d]
Dr. Farag has graduated 25 MS students, 15 Ph.D. students (as of Spring
2008), and trained eight postdoctoral fellows in the past 16 years, and
attracted over $8M of funded research from the NSF, DoD, NIH, Norton
Hospital, Whitaker Foundation and private organizations. Currently, Dr.
Farag is advising 8 PhD graduate students, one MS student and two
undergraduate students. In addition, Dr. Farag is supervising three
technical staff. Dr. Farag is an inventor (has one official patent and
four provisional patents), and an educator (one text book on DSP is
pending and two edited book on level set methods, Springer 2007, in
addition to three other textbooks in various planning stages). |
